Citation
Analysis of 360 video files across multiple resolutions and cameras

Material Information

Title:
Analysis of 360 video files across multiple resolutions and cameras
Creator:
Pelc, Nicholas James
Place of Publication:
Denver, CO
Publisher:
University of Colorado Denver
Publication Date:
Language:
English

Thesis/Dissertation Information

Degree:
Master's ( Master of science)
Degree Grantor:
University of Colorado Denver
Degree Divisions:
Department of Music and Entertainment Industry Studies, CU Denver
Degree Disciplines:
Recording arts
Committee Chair:
Grigoras, Catalin
Committee Members:
Smith, Jeff M.
Merkel, Jeff

Notes

Abstract:
Forensic experts rely on the validity and reliability of data consistent with Daubert standards and Federal rules of evidence for identifying findings and formulating opinions about evidence in legal proceedings. Digital forensic experts have an acute awareness of possible distortions or manipulations of electronically recorded or transmitted data. A critical aspect of analyzing digital evidence is to inspect the reliability, absence of, degradation, or alteration of data used in forensic evaluation. New recording devices are constantly being developed. One such new tool is the 360 video available for GoPro players and, often, for broadcast on YouTube. 360 video provides an opportunity to capture, record, and review multi-dimensional digital information. For this study, several test patterns are analyzed using the 360 video process. The purpose of the research is to identify any potential degradation, distortion, manipulation, or data loss when 360 evidence is transferred from an originating device to a secondary device.

Record Information

Source Institution:
University of Colorado Denver
Holding Location:
Auraria Library
Rights Management:
Copyright Nicholas James Pelc. Permission granted to University of Colorado Denver to digitize and display this item for non-profit research and educational purposes. Any reuse of this item in excess of fair use or other copyright exemptions requires permission of the copyright holder.

Downloads

This item has the following downloads:


Full Text
ANALYSIS OF 360 VIDEO FILES ACROSS MULTIPLE RESOLUTIONS AND CAMERAS
by
NICHOLAS JAMES PELC B.S., University of Colorado Denver, 2015
A thesis submitted to the Faculty of the Graduate School of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Recording Arts Program
2019


© 2019
NICHOLAS JAMES PELC ALL RIGHTS RESERVED
II


This thesis for the Master of Science degree by
Nicholas James Pelc has been approved for the Recording Arts Program by
Catalin Grigoras, Chair Jeff M. Smith
Jeff Merkel


Pelc, Nicholas James (M.S., Recording Arts Program)
Analysis of 360 Video Files Across Multiple Resolutions and Cameras Thesis directed by Associate Professor Catalin Grigoras
ABSTRACT
Forensic experts rely on the validity and reliability of data consistent with Daubert standards and Federal rules of evidence for identifying findings and formulating opinions about evidence in legal proceedings. Digital forensic experts have an acute awareness of possible distortions or manipulations of electronically recorded or transmitted data. A critical aspect of analyzing digital evidence is to inspect the reliability, absence of, degradation, or alteration of data used in forensic evaluation. New recording devices are constantly being developed. One such new tool is the 360 video available for GoPro players and, often, for broadcast on YouTube. 360 video provides an opportunity to capture, record, and review multi-dimensional digital information. For this study, several test patterns are analyzed using the 360 video process. The purpose of the research is to identify any potential degradation, distortion, manipulation, or data loss when 360 evidence is transferred from an originating device to a secondary device.
The form and content of this abstract are approved. I recommend its publication.
Approved: Catalin Grigoras
IV


This thesis is dedicated to my entire family. Thank you to my father who, when I was at an early age, showed me the world of forensic sciences and to my mom whose patience is unheard of.
v


ACKNOWLEDGMENTS
Thank you to Jeff and Catalin for allowing me to "cd" from the directory of recording arts and into the world of digital forensics. The information I gathered while in this Master's program is incredible and extremely relevant. I am looking forward to continuing this educational journey and learning more.
Jeff, thank you for allowing me to work in the lab and on the DARPA project. Hopefully, all the work that we have done and continue to do will help many people.
Thank you, Leah, for helping keep us all on track, and Emma in Leah's absence.
A thank you is due to Jeff Merkel for helping with your knowledge of Ambisonics in this evolving world of
audio.
Finally, thank you to my mom, dad, and dogs.
VI


TABLE OF CONTENTS
CHAPTER
I. INTRODUCTION...............................................................................1
Problem Statement.......................................................................1
Literature Review.......................................................................1
II. METHODOLOGY................................................................................3
Go Pro Fusion...........................................................................6
YouTube-dl..............................................................................6
Matlab..................................................................................6
010 Editor..............................................................................7
Medialnfo...............................................................................7
Exit tool...............................................................................8
FFmpeg..................................................................................8
AtomicParsley...........................................................................8
Go Pro VR player........................................................................8
360Fly app..............................................................................8
GoPro app...............................................................................8
Fly360 Director.........................................................................8
VLC.....................................................................................8
Paint Shop Pro..........................................................................8
III. RESULTS..................................................................................12
YouTube Up Down........................................................................23
PRNU Results...........................................................................25
Audio Results..........................................................................31
IV. DISCUSSION...............................................................................34
Conclusion.............................................................................35
REFERENCES....................................................................................36
vii


LIST OF TABLES
TABLE
1. Metadata..................................................................................12
2. Video.....................................................................................13
3. Audio.....................................................................................16
4. Other 1...................................................................................18
5. Other 2...................................................................................19
6. Other 3...................................................................................19
7. Atoms of the Original Recording.........................................................22
8. Atoms of the 360 Recompressed Video.......................................................22
viii


LIST OF FIGURES
FIGURE
1. Illustration of GoPro Fusion Different Resolutions...............................................3
2. GoPro Fusion Diagram.............................................................................4
3. Diagram of 360Fly................................................................................5
4. MDCT Equation....................................................................................6
5. Hex Analysis: Original..........................................................................20
6. Hex Analysis of the Files Rendered out of the GoPro Fusion Studio Software......................21
7. Hex Analysis of the Files Downloaded from YouTube in the MKV File Type..........................23
8. Hex Analysis of the Files Downloaded from YouTube in the MP4 File Type..........................24
9. GoPro Rendered File Comparison to GoPro.........................................................26
10. GoPro Comparison to Sony Falcon.................................................................26
11. GoPro Comparison to Canon PowerShot.............................................................27
12. Image Used for First PRNU Tests.................................................................27
13. A Screenshot from the 360 Point of View Taken from the GoPro Studio Software Before Rendering..28
14. A Screenshot from the Fish Eye Point of View Taken from VLC Upon Playback.......................28
15. Image Captured from the 360Fly as the Fisheye Perspective.......................................29
16. Image Captured from the 360Fly as the 360 Perspective, Camera Not Moved.........................30
17. Waveform, Spectrogram and MDCT Map..............................................................31
18. LCF Analysis....................................................................................32
19. LTAS Analysis...................................................................................33
20. LTA Analysis Two................................................................................33
IX


CHAPTER I
INTRODUCTION Problem Statement
360 video has become an increasingly popular multidimensional recording device, yet the 360 technology has not been subject to a thorough analysis concerning its reliability with data transfer. For the technology to meet one of the basic Daubert standards, 360 video requires testing and assessment of its consistency following transfer from the originating device to other media. Without such an analysis, the 360 video technology might not be considered acceptable forensic evidence under Daubert standards, which require the demonstrated validity and reliability of assessment tools.
Literature Review
Over the past two decades, advances in technology have created significant opportunities for and challenges to the collection and analysis of digital evidence for both solving crimes and preparing cases for court.
In a white paper prepared for the National Institute of Justice, Goodison, Davis, and Jackson (2015) provided an overview of digital evidence in the criminal justice system. They underscored major themes, including that, "Documentation of digital evidence incorporates the twin issues of authentication and chain of custody." They also cited the growing importance of video evidence in criminal investigations.
Inherent to the growing importance of digital evidence is adherence to rules of evidence and case law. Testimony by expert witnesses (Federal Rules of Evidence, 702) requires that the presentation of data meets standards of adequate reliability and validity to be considered as a basis for the findings and opinions provided by an expert. The courts have moved from a general acceptance standard, in Frye vs. US, 293F. 1013 (1923), to a scientific standard, as defined in Daubert vs. Merrell Dow Pharmaceuticals, 509 US 579 (1993). New technology, then, requires the development of research that supports these Federal standards and withstands scientific scrutiny. In the development of video evidence, data should both be recorded accurately and, in the chain of evidence, accurately transferred from one media device to another. Data transmission is the process of transferring data between two or more digital devices, and such transmission may involve either serial or parallel transfers (Melton, 2016). Robinson (2012) has identified some of the problems that can be encountered in data transmission. Others (e.g., Natarjan, 2003; Tsun-Li, 2014;) have addressed the manner by which these transfer
1


problems can be addressed. Recent research has also presented methods for verifying and authenticating
transmitted video data (Whitecotton, 2017; Harran et al., 2018).
2


CHAPTER II
METHODOLOGY
Multiple comparison paradigms were used to test this study's hypotheses. These paradigms can be summarized as follows. Understanding this comparison and analysis is facilitated by a detailed examination of the GoPro Fusion. The GoPro Fusion has two lenses attached to the body of the camera. There is a front-facing lens and a back facing lens. Both lenses have an f stop of 2.0. It is essentially two cameras in one body. The GoPro records to two separate micro SD cards, which work in tandem. While recording on the Fusion, the operator can either record using a standard electrical outlet or the camera's internal battery. The outlet on the camera is a USB-C style jack. The portable battery life of the GoPro is around 1.5 hours maximum. The system can be navigated using the button the front of the unit or by connecting to a cell phone via WiFi and using the GoPro application. Data can be written through the application, or to a mobile device, or the internal micro SD cards. Additional information can be stored to GoPro's cloud. The maximum resolution that can be recorded on the GoPro is 5.2K at 30 FPS (frames per second), NTSC or 25 FPS, PAL Or 3k at 60 FPS. Frames per second are the number of frames of video that are captured each second when recording. The term 5.2k refers to the horizontal lines and the vertical number of pixels. At 5.2K, there are 4992 lines with a width of 2496 pixels. At 3k, there are 3000 horizontal lines and a width of 1504 pixels. An illustration of the different resolutions is shown below.
Figure 1 Illustration of GoPro Fusion Different Resolutions.
The aspect ratio of the recordings is made at 16:9. Aspect ratio is the ratio of the image's width to height. In this case, the image is 16 inches wide and 9 inches high. The purpose of a Go Pro 360 camera is to capture not only a front view but a complete image of the entire surround of the camera. The field of view (FOV) refers to how
3


much of the subject is captured through the lens. The measurement is in degrees so that with the Go Pro camera,
360 degrees of the field of view. Therefore, the camera will capture the entire surrounding. The GoPro also has six
(6) axes of stabilization. This increases camera stability in order to maintain smooth and steady shots. Protune is another feature on the GoPro Fusion as well as most of the products sold by the manufacturer. This feature allows the operator to be able to adjust both the International Standards Organization (ISO) and the Exposure Value (EV) level. ISO is used to adjust the overall brightness of the image. With a lower ISO setting, the image will be darker, but there will be less noise. With a higher ISO setting, the image will be brighter but will contain more noise. For the purpose of the tests used in this study, various ISO settings were used depending on the lighting situation. The settings available were 400,1600 and 6400. EV, exposure value, ranges from -2.0 to +2.0. The default setting and what was recorded which was the EV set to (0). The higher the EV setting, the higher the brightness of the image. Protune is the ability to adjust the camera from auto settings to have more control by using variable ISO and EV settings. The GoPro Fusion, like many other digital cameras, has the capacity to adjust the date and time of the image acquisition. This information is stored within the metadata. When the device is connected either to the GoPro application or to the proprietary software, Fusion Studio, the date and time will auto adjust to whatever the actual device setting. This adjustment is relevant if the device is recovered for time offset purposes. Location of the device using GPS information is recorded as well. This information is only available when the fusion is connected to the GoPro application. This information is then stored within the metadata of the file. The GoPro features four built in microphones in order to capture audio all around the camera. The figure below depicts the various feature of the Go Pro 360 camera.
1. Speaker 5. Microphones
2. Camera Lens (Back) 6. USB-C Port
3. Status Light 7. Side Door
4. Mode Button ([o-»| ] 8. Latch Release Button
9. Camera Lens (Front) 13. Battery Door
10. Status Screen 14. Slot for microSD Card 2
11. Slot for microSD Card 1 15. Battery
12. Shutter Button (jo)) 16. Mounting Fingers
Figure 2 GoPro Fusion Diagram.
4


Another camera used in the tests in this study was the 360 Fly. This camera's original design purpose was
for aviation and military use. The camera's navigation is controlled by a single button that can be used to pair to another device for total operating control. Once paired to another device, even more control over settings of the Fly can be maintained. Adjustments can be made for variables such as saturation, contrast, f speed and brightness. These are not necessarily detailed pro controls. They all feature a single slider, and they do not give numerical values of any specific degree of change is being made. Adjustment are not finite either. For the purposes of all test recordings performed in this study therefore, all settings were left at a neutral point. The Fly, like the GoPro, can record in many different resolutions. The maximum resolution that can recorded is at 2880 x 2880 with 30 frames per second. The aspect ratio is not 16:9 unless the camera is set to POV mode. Once that setting is made, further adjustments cannot occur including any changes to frame rates either. This camera features a single lens instead of having two lenses. The lens on the Fly is a single fisheye. A fisheye lens is one that shoots extremely wide images. The single lens can go to f 2.5. The field of view is again slightly different. Vertically, it has 240 degrees, and horizontally it has 360 degrees. Instead of recording two a SD card, the 360 Fly records internally with a total space capacity of 64gb. The battery life is slightly longer than the Go Pro, with the average time being two hours of battery life. The Fly 360 charges via a docking system that uses USB 2.0 technology. Like the Go Pro fusion, the Fly 360 stores GPS data within the metadata. This feature can be turned on or off within the settings on the Fly360 application. The features if the Fly are shown above.
Features
Figure 3 Diagram of 360Fly.
5


Various definitions are available for image recording. For the purpose of this study, recordings were made
at the highest possible resolution to capture and store the most maximum amount of data possible. This procedure would allow the capture and analysis of the maximum change that occurs. All test recordings were at 5.2k on the GoPro at 30 frames per second. The Fly 360 recordings were at 2880 x 2880 at 30 frames per second. While it is possible to record with the Fly at 60 frames per second, the resolution would drop to 1728 x 1728.
Six recordings were collected and several types of data analysis were performed in order to compare any changes in the data acquisition and transfer. Files were then rendered out of the GoPro Fusion freeware, causing recompression in the files. The same process was done using the Fly360 test recordings. The recompressed files from the GoPro Fusion software were then recompressed through YouTube and analyzed.
The Go Pro Fusion software was used to create the 360 videos in multiple resolutions. Recording number one was used to create rendered file, VIDEO_0002.mov.
YouTube-dl was used to download the laundered videos from YouTube. Recording number one from the GoPro Fusion software, VIDEO_0002.mov, was used for the uploaded video to YouTube. The two YouTube files were created from this test as well, VIDEO_0002_yt-QQco8L093WY.mkv and VIDEO_0002_yt-QQco8L093WY.mp4 Matlab was used to analyze the PRNU, MDCT and ENF of the recordings. PRNU, (Photo Response Non-Uniformity) allows for identification of the unique characteristics of the camera's noise. Information using PRNU can be extracted and analyzed against the test database to see if the camera has support, limited support or no support to the images in the other camera in the database. MDCT was also analyzed. MDCT (Modified Discrete Cosine Transform) is designed to be performed on consecutive blocks of large data sets. It is not applied to an audio signal directly so that interference with the original data is minimized. The GoPro Fusion records audio in an AAC format. AAC is a ISO standard format that contains MDCT.
Figure 4 MDCT Equation.
6


LTAS (Long Term Average Spectrum) was performed on the audio data in order to record and
demonstrate the mean, max, and min or the power. This helps us to define the overall recordings in the audio channel. ENF (electrical network frequency) was also evaluated in the audio recordings occurring in this study. ENF measures where noise from the main power supply is recorded into the actual recording. LCF, (low cut filter) was also measured. This variable provided information to determine if at a certain point, the audio recording has a low-cut filter that removes data from the low-frequency range. Recordings number one, GPBK0037.mp4, was used to analyze the PRNU. A Print Screen HQ was taken for analysis against the PRNU database. A Print Screen HQ was used as that is the best possible way currently, to extract a single frame from a recording using the native player. This recording had the best exposure settings for a neutral test. Recording number three, GPBK0121.mp4, was used for the MDCT, LCF and LTAS analysis. This recording was longer in length with more background noise for more information for analysis. This simulates a more realistic environment. Recordings four, GPBK0068.mp4, five, GPBK0129.mp4, and six, GPBK0130.mp4, were used for ENF analysis. Recording four was made in an outdoor location, five was made in the same location as one and two but for a greater duration, under fluorescent lights and run off battery power. Six was made with the same conditions as five but plugged into the wall for power instead of running off battery power.
010 Editor was used to perform a hex analysis of the original recordings, recompressed recordings through YouTube and recompressed recordings through the creation of the 360 videos with the GoPro Fusion software. Hex analysis (hexadecimal) allows examination of the bytes of the file. From there, an analysis of the raw information can be made without it being encoded. Analysis of both the header and footer of this information can be completed to observe if any changes have been made from recompression, or any other changes have occurred including for example if the file has decreased or increased in size, and what is missing or has been added. Test recording one was used for this analysis as well as the first render from GoPro Studio and the two downloaded files from YouTube.
Medialnfo was used to assess the metadata and changes that occurred between the recompression of the videos in each stream. Metadata analysis was also used for this study to obtain a comparison between information is stored within the files. Test recording one, GPBK0037.mp4, the first render from the GoPro Studio, VIDEO_0002.mov, was used, the two downloaded files from YouTube were used VIDEQ_0002_yt-
7


QQco8L093WY.mkv and VIDEO_0002_yt-QQco8L093WY.mp4, test recording number one from the Fly360, FLY08401.mp4 was used and the same recording but rendered out through the Fly360 director FLY08401(CONVERTED).mp4 was used in this test.
Exif tool was used to verify the results from the Medialnfo information. All recordings were used in this
test.
FFmpeg was used to extract the audio stream from the video. All test recordings, one through six, had the audio stream extracted for analysis.
AtomicParsley was used to demonstrate the loss in atoms upon recompression of the video. Recording one, GPBK0037.mp4 and the recording rendered from the GoPro Fusion software, VIDEO_0002.mov were used.
The Go Pro VR player was used to playback the original recordings without having to render them and effectively alter the video as well as to create a print screen of the original file. The Print Screen HQ function was also utilized to extract the image for all of the PRNU tests.
The GoPro app was used to start and stop all recordings made.
The 360Fly app was used to start and stop recordings and to adjust camera settings to the highest possible resolution.
Fly360 Director was used for the creation of the 360 videos from the additional 360 camera.
VLC was used in an attempt to play back the 360 video files without the GoPro native player. All test recordings were attempted to playback in VLC.
Paint Shop Pro was utilized to resize the images for the PRNU analysis. This was due to the unique size of the frame exported from the GoPro VR player.
In addition to comparing the two cameras, an examination was completed of the files after being uploaded and downloaded to YouTube. A popular way of sharing videos is by uploading them to the YouTube site. It is important to see the change that occurs between the original video and the recompression through YouTube. When recording to the GoPro Fusion, two files are created upon clicking record which captures a separate file per each lens. When these files are played back, they are warped and not playable as 360 videos yet. These videos must be imported into the GoPro Fusion software to create the 360 FOV videos. Two pieces of software are provided with the GoPro Fusion to aid with this data transfer. Upon rendering from the GoPro Fusion Studio,
8


several options are available including Video codecs, H.264, CineForm '422 High' and Pro Res 422. For the purpose
of this study, CineForm was used because it is likely to produce the least amount of recompression. Video resolutions was a second option. From 5.2k (4992 x 2496), 4k (3840 xl920), 3k (2880x1440), 2K (2048 x 1024_ are all available options. 5.2K was used in this study because it contains maximum data for the present study. Spatial Audio is the third option. Stereo was used for this study but 360 audio (Ambix) is another available option. Finally, DWarp (Parallax Compensation) is the final option. This setting was left on for this study because DWarp removes parallax lines in the raw stitched footage. Parallax is a displacement or difference in the apparent position of an object when viewed along two different lines of sight and is measured by the angle or semi-angle of inclination between those two lines.
Finally, the files were taken and uploaded to YouTube. This was to measure the recompression of the 360 files compared to the original data set to see what changes will occur upon download of the files.
For the data analysis, best practices for measuring authenticity or recompression in transferred data were utilized for transfer from GoPro Fusion's warped original files to a 360 playable format uploaded and downloaded from YouTube. These best practices have been written as guidelines by the Scientific Working Group on Digital Evidence (SWGDE). Image authentication is necessary to determine whether image data are accurate and valid representations of subjects and events. Importantly, these guidelines do not define specific analytic techniques or tools but a process to detect staging or manipulation of images in the manner of acquisition or transfer of images from one recording device to another. This process requires that the original image should be preserved, and any transferred image should maintain the integrity of the initially acquired data image. In this study, the transfer of data from the 360 GoPro to another device for data preservation is tested. An illustration of this image transfer and integrity are described in the article from SWGDE on, "Best practices for digital video evidence"
The discussion of this analysis includes the following topics:
1. Implications for authenticity and reliability when making data transfers of 360 video;
2. Directions for future research;
3. Best practices for collecting video data from the recording device;
9


The Scientific Working Group on Digital Evidence has also published guidelines on "Minimum
requirements for testing tools used in digital and multimedia forensics." (SWGDE, 2018). These guidelines differentiate the critical forensic tools for preservation, acquisition, hashing, and wiping digital data. Multimedia tools for imagery enhancement and analog video capture are also addressed. The purpose of digital testing evidence is to assess the confidence level which can be assigned to a tool or procedure to perform correctly and reduce the risk of errors. In this study, the acquisition and transfer of video images using the GoPro Fusion 360 as well as the Fly360 was tested regarding the accuracy and validity of data moved from an original recording device to a second device. This process of transfer of video images is described in an applied legal article on the collection and production of technically sound and defensible digital data in a litigation process by Bowers (2018). Fie described several PDF pitfalls. Producing native files which hold all the metadata might be opened, corrupted, or inadvertently altered. Another problem can occur when multiple documents are produced in a single PDF without the original metadata being shown in a corresponding load file. The receiving party might then have to separate documents which could open the door to challenges from opposing counsel or even court sanctions. Bowers strongly urged the use of document specialists in the processing and management of native files. The specialist can identify common issues which are inaccurately collected and produced data. Bowers also noted that the common legal practice of Bates Numbering could produce even obvious errors in identifying where one document ends and the next one begins. Fie summarized the top four self-collection issues (Bowers, p.5):
1. A single PDF was created, which contained multiple documents without any delineation between those
documents.
2. Metadata had been altered by employees simply forwarding requested emails.
3. Emails had been printed and then scanned into PDF which removed native metadata.
4. The inability to easily identify the melody of relationships of files.
Fie concluded, "Every attorney should approach each collection with the goal of making it forensically sound." (Bowers, p.7) Fie also noted, "Under the newly amended FRE 902, metadata will not be self-authenticating unless a qualified person has inspected the data, recorded the process used, and certified that an exact copy of the data was created." (Bowers, p. 6). This study was designed to test the hypothesis that there will be a significant change in the video data resulting in a loss of data accuracy due to recompression.
10


As a starting point to eliminate misleading data, SD cards were formatted on recording devices before use. To avoid camera motion, all recordings were made using a cell phone to start and stop recording. Version 01.70.00 of GoPro Fusion was used.
11


CHAPTER III
RESULTS
The metadata and file structure of the non-combined files were analyzed. 360 videos were captured to the GoPro fusion via two micro sd cards, one for each side of the camera, which were labeled 1 - 100GBACK (BACK) and 2 - 100GFRT (FRONT). One test recording was created for this research. One recording resulted in two files: GPBK0002 and GPFR0002. The tables below are a comparison of the metadata of the data collected through test recordings. Table 1 is of the metadata. You can see the format as well as file size changes between recompression. Table 2 is of the video content. Variances occur in multiple items in this table. From the aspect ratio to the bit rate. Table 3 is of the audio content and table. Again, multiple changes occur here. From the sample rate to even the duration. Four, five and six is additional metadata. In the original GoPro recording this information is stored but in all other form of recompression, including the Fly 360's recordings, this information is not stored. Table 1 Metadata
Informa tion Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002_ yt- QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_yt QQco8L093WY. mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4)
Format MPEG-4 MPEG-4 Matroska MPEG-4 MPEG-4 MPEG-4
Format Version n/a n/a Version 4 / Version 2
Format Profile N/A Quicktime n/a Base Media Base Media / Version 2 Base Media
Codec ID mp41 (mp41) qt 2005.03 (qt ) n/a isom (isom/iso2/avcl /mp41) mp42 (mp42/iso m) isom (isom/iso2/avcl/mp 41)
File size 228 MiB 2.26 GiB 15.5 Mb/S 52.2 Mib 54.4 Mib 24.4 MiB
Duratio n 42 s 309 ms 42s 42ms 42 s 98ms 42 s 98ms 9 s941ms 9 s 963 ms
Overall bit rate mode Variable Variable Variable
12


Table 1 Cont
Metadata
Overall bit rate 45.2 Mb/s 461 Mb/s 15.5 Mb/s 10.4 Mb/s 45.1 Mb/s 20.6 Mb/s
Writing application n/a n/a Lavf57.83.100 Lavf57.83.100 Lavf57.25.100
Writing Library n/a n/a Lavf57.83.100 n/a
Error Detection Type n/a n/a Per Level 1 n/a
xyz +00.0000+000.0000/
Table 2 Video
Information Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002 _yt- QQco8L093W Y.mkv) YouTube Up Down MP4 (VIDEO_0002_ yt- QQco8L093W Y.mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVER TED.mp4)
ID 1 1 1 1 1 1
Format AVC CineForm VP9 AVC AVC AVC
MultiView_ Count 2
Format/lnfo Advanced Video Codec Advanced Video Codec Advanced Video Codec Advanced Video Codec
Format Profile High@L5.1 High(®L5.1 Baseline@ L5.1 Baseline@L5.1
Format Settings CABAC/1 Ref Frames 2 Ref Frames 1 Ref Frames 1 Ref Frames
Format settings, CABAC Yes No No No
Format settings, RefFrameS 1 Frame 2 Frames 1 Frame 1 Frame
Format settings, GOP M=l, N=15 M=l, N=30
13


Table 2 Cont
Video
Codec ID Avc 1 CFHD V_VP9 Avc 1 Avc 1 Avc 1
Codec ID/Info Advanced Video Coding CineForm High-Definition (HD) wavelet codec Advanced Video Coding Advanced Video Coding Advanced Video Coding
Duration 42 s 309 ms 42 s 9ms 42s 9ms 42s 9ms 8 s 876 ms 9s 527ms
Source Duration 9 s 491 ms
Bit Rate Mode Variable Variable
Bit Rate 45.0 Mb/s 460 Mb/s 10.3 Mb/s 46.4 Mb/s 21.3 Mb/s
Width 2704 pixxels 5120 pixels 3840 pixxels 5120 piels 2880 pixels 3840 pixels
Height 2624 pixxels 2560 pixels 2160 pixels 2560 pixels 2880 pixels 1920 pixels
Display Aspect Ratio 1.030 2.000 16:9 2.000 1.000 2.000
Frame Rate Mode Constant Constant Constant Constant Variable Constant
Frame Rate 29.970 (30000/1001) FPS 29.970 (29970/1000) FPS 29.970 (30000/1001) FPS 29.970 (30000/1001) FPS 28.027 FPS 28.027 FPS
Minimum Frame Rate 1.542 FPS
Maximum Frame Rate 37.943 FPS
Color Space YUV YUV YUV YUV YUV
Chroma subsampling 4:2:0 4:2:0 4:2:0 4:2:0
Bit Depth 8 Bits 8 Bits 8 Bits 8 Bits
Scan Type Progressive Progressive Progressive Progressive Progessive
Bits/(Pixel*Frame) 0.212 1.171 0.026 0.200 0.103
Stream Size 227 MiB (100%) 2.25 Gib (100%) 52.5 MiB (99%) 49.0 MiB (98%) 24.1 MiB (99%)
Source Stream Size 52.4 Mib (98%)
Title GoPro AVC Video Handle
Language English English English English
Default n/a n/a Yes
Forced n/a n/a No
Encoded Date UTC 2018-08-23 11:37:56 UTC 2018-10-08 18:56:14 UTC 2018- 10-24 21:56:54
Tagged Date UTC 2018-08-23 11:37:56 UTC 2018-10-08 18:56:14 UTC 2018- 10-24 21:56:54
Color Range Full n/a Limited Limited
14


Table 2 Cont
Video
Color primaries BT.709 BT.709 BT.709 BT.709
Transfer characteristics BT.709 N/a Bt.709 BT.709
Matrix coefficients BT.709 BT.709 Bt.709 BT.709
Mdhd_Duration 8876
Writing library x264 core 148 r2597 e86f3al
*Only found on Fly rendered : Encoding settings : cabac=0 / ref=l / deblock=0:0:0 / analyse=0:0 /
me=dia / subme=0 / psy=l / psy_rd=1.00:0.00 / mixed_ref=0 / me_range=16 / chroma_me=l / trellis=0 / 8x8dct=0 / cqm=0 / deadzone=21,ll / fast_pskip=l / chroma_qp_offset=0 / threads=12 / lookahead_threads=2 / sliced_threads=0 / nr=0 / decimate=l / interiaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=0 / weightp=0 / keyint=250 / keyint_min=25 / scenecut=0 / intra_refresh=0 / rc=crf / mbtree=0 / crf=23.0 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=0
15


Table 3
Audio
Informati on Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002_ yt- QQco8L093W Y.mkv) YouTube Up Down MP4 (VIDEO_0002_ yt- QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4)
ID 2 2 2 2 2 2
Format AAC PCM AAC AAC AAC AAC
Format/I nfo Advanced Audio Codec N/a Advanced Audio Codec Advanced Audio Codec Advanced Audio Codec Advanced Audio Codec
Format Settings n/a Little/Signed n/a
Format Profile LC n/a n/a LC LC LC
Codec ID mp4a-40-2 sowt A_AAC-2 mp4a-40-2 mp4a-40-2 mp4a-40-2
Duration 42 s 304 ms 42s 42ms 42s 98ms 42 s 98 ms 9 s 941 ms 9 S 963 ms
Source Duration n/a 42s 9ms n/a
Bit Rate Mode Constant Constant Constant Variable
Bit Rate 128 kb/s 1536 kb/s 126 kb/s 96.0 kb/s 126 kb/s
Channel) s) 2 Channels - - - - -
Channel Positions Front: L R - - - - -
Sampling Rate 48.0 kHz - 44.1 kHz 44.1 khz 48 kHz 48khz
Bit Depth 16 bits
Frame Rate 46.875 FPS (1024 SPF) 43.066 FPS (1024 SPF) 43.066 FPS (1024 SPF) 46.875 FPS (1024 (SPF) 46.875 FPS (1024 SPF)
Compres sion Mode Lossy Lossy Lossy Lossy Lossy
Stream Size 662 KiB(0%) 7.70 Mib (0%) 645 kib (1%) 117 kiB (0%) 153 Kib (1%)
Title GoPro AAC 7.69 MiB (0%0
Languag e English English English
Default Yes Yes Yes
Alternat e group 1
16


Table 3 Cont
Audio
Forced No
Encoded Date UTC2018- 08-23 11:37:56
Tagged Date UTC 2018- 08-23 11:37:56
17


Table 4
Other 1
Informa tion Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002. mov) YouTube Up Down MKV (VIDEO_0002_ yt- QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt- QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4)
ID 3 n/a
Type Time code n/a
Format QuickTime TC n/a
Duratio n 42 s 309 ms n/a
Time code of first frame 12:01:17:2 1 n/a
Time code, striped Yes n/a
Title GoPro TCD n/a
Languag e English n/a
Encoded Date UTC 2018- 08-23 11:37:56 n/a
Tagged Date UTC 2018- 08-23 11:37:56 n/a
Bit Rate Mode CBR n/a
18


Table 5
Other 2
Informa tion Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002. mov) YouTube Up Down MKV (VIDEO_0002_ yt- QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt- QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4)
Type meta -
Duratio n 42 s 42ms -
Bit Rate Mode CBR -
Table 6 Other 3
Informatio n Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002_ yt- QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt- QQco8L093W Y.mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4)
Type meta n/a
mdhd_Dur ation 42309 n/a
Bit rate mode VBR n/a
Below, is the first hex analysis. The file was imported into 010 editor. You can see information such as the version number, name of the camera and the file type. This matches what we have found in the mediainfo data above.
19


♦ tdit As: Hex v Run bcript v Run I emplate:
A 6 A B E p . • -as •.
h: 00 00 00 14 66 74 79 70 6D 70 34 31 20 13 10 18 . ...fcypmp41 ...
D01C h: 6D 70 34 31 OE 3C DF 84 6D 64 61 74 47 50 52 4F np41. h: 8$ 04 00 00 46 53 31 2E 30 34 2E 30 31 2E 37 30 ’...FS1.04.C1.70
JO 3C hi 2E 30 30 4E 41 46 37 31 31 32 36 30 30 35 30 34 .00NAF7112600504
304C â– h: 3 3 38 34 00 00 00 00 00 00 00 00 00 00 00 00 00 384
hi 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
0C6C h: 00 00 00 53 78 25 A9 CE 9A 6D 9B CA 66 29 A 4 D2 ...Sx3ctSm >j £)u6
hi DF CD FC 43 33 32 30 31 31 32 34 35 34 35 35 32 610C320112454552
h: 31 00 4 h: 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00A< h: 00 00 00 82 00 44 00 00 53 78 25 A9 CE 9A CD 9B ..... D. .Sx*Cl Sin >
DOBC h: 6 A 66 29 A4 D2 DF CD FC CB 85 E2 07 78 09 03 00 jl) fOlitOE.-A. x. . .
h: 00 00 11 DO 00 00 00 00 00 00 00 00 00 CO 00 00 . . .€■
hi 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
oc?:c :h: 00 00 00 00 32 34 35 34 35 35 32 31 03 00 46 00 24545521..F.
:• o Fi hi 00 02 00 01 32 34 35 34 35 35 32 31 17 CB 25 38 24545521..38
h: 00 00 00 00 01 00 00 00 90 5F 01 00 90 OA 00 00
one hi 40 OA 00 00 00 00 00 00 90 CA 00 00 40 OA 00 00 3 3. . .
h: 01 00 00 00 80 BB 00 00 64 00 00 00 10 CO 00 00 e». .d
0 i 3 hi 02 00 00 00 00 F4 01 00 94 9C 7E 5B 94 A3 00 00 0. ."or* .
J 1 h: C2 03 00 00 00 00 00 00 00 00 00 00 00 CO 00 00 A
hi 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
i . r i h: CO 00 00 oo CO OO 00 OC 44 45 56 4 J 00 01 01 08 DEVC
h: 44 56 49 44 4C 04 00 01 00 00 00 01 44 56 4E 4D DVIDL DVNM
h: £3 01 00 15 47 65 6r 6D 65 74 72 79 20 43 61 CC c...Goometiy Cal
j 1 9C h: 69 62 72 61 74 69 6F 6E 73 00 00 00 53 54 52 4D ibrationu...STEM
j 4 A; )h: CO 01 00 D4 53 48 46 53 64 08 00 01 40 30 42 44 ...OSHFXd...3CBD
V 4 28 66 7E 53 48 46 59 64 08 00 01 3F D2 E7 46 t (1-SHFYd. . . ?0 â– h: C2 A1 73 7E 53 48 46 bA 64 08 00 01 40 40 D8 CE AiS-SHFCd...3301
h: 2D B7 03 39 41 4E 47 58 64 08 00 01 BF CB 00 F2 - •.9ANGXd...iE.6
h: D7 ED FB C5 41 4E 47 59 64 08 00 01 40 66 82 FA -iCAANGYd...3l.u
01 FC hi 0 b F2 03 04 41 4E 47 5A 64 08 00 01 BF E5 B4 20 .6..ANGZd...ii
h: DD 2C 87 CB 43 41 4C 57 4C 04 00 01 00 00 OF AO Y, JtCALWL
0211 hi 43 41 4C 48 4C 04 OO 01 OO OO OB B8 43 41 50 48 CA1HI, , CAF'H
022C h: 4C 04 00 01 CO 00 OB B8 57 45 57 43 73 10 00 01 L W EWCs...
023i h: 0A 90 OC 20 06 20 OC 40 OC 20 OC 20 02 EO OC 20 3. . .a.
02 4( ih: 48 4b 48 43 73 02 00 01 00 00 00 OC 53 4 9 5A 57 HEHCs- SISK
h: 64 08 00 01 40 18 cc CC CC CC CC CD 53 49 5A 48 d. . .3.units ISH
q 2 f( hi 64 08 00 01 40 12 QC« 99 99 99 99 9A 43 41 50 57 d. . . 3.~~~~~cc APK
02 7 4C 04 00 01 00 00 OC 20 44 45 56 43 00 01 00 8C L DEVC. . .CE
02$i h: 44 56 49 44 46 04 00 01 49 4D 55 43 44 56 4E 4D DVIDF...IMUCDVKM
029( hi 63 01 00 10 49 4D 55 20 43 61 ec 69 62 72 61 74 C...IMU Calibrat
0?A 'h: 69 6F 6E 73 53 54 52 4D 00 01 00 60 41 43 4C 53 ionsSTRM... ACES
02 Bi hi 66 OC 00 01 41 ID 22 IE 41 1C 20 20 41 ID 40 22 f...A.".A. A.3"
hi 41 43 4C 42 £6 OC 00 01 3E 10 EO A8 BF C5 7A 14 ACDB1..=.
hi BE 14 42 42 47 59 52 42 66 OC OO 01 3B F2 AE 08 BBGYRBf . . .;6S.
02Ef hi 3B FA 48 30 BC 11 29 89 4F 52 4E b4 66 10 00 01 ;uH0H. ) V.ORNTf. . .
3 2 Fi 3F 02 56 8A BF OO 20 2E 3E F6 ED EE BF 01 F4 6F .>01i£.do
â– hi 54 4D 50 43 66 04 OO 01 41 F9 26 00 44 45 56 43 TMPCf . . . Alii . DEVC
hi 00 01 00 CO 44 56 49 44 46 04 00 01 42 41 43 4B ...ADVIDF...BACK
032« hi 44 56 4E 4D 63 01 00 09 42 61 €3 6B 20 4C 65 €E DVNMc...Back Den
hi 73 00 00 00 53 54 52 4D OO Cl OO 98 50 4F 4C 59 s. . .STRM...*PODY
Figure 5 Hex Analysis: Original. Note. In hex analysis, front lens still notates back and front. Test is done 0002 and
(Recording 1 GPBK0037.mp4)
20


During export, a selection of destinations is offered: editing, Facebook, YouTube, or Vimeo. Different resolutions may be selected as well: 5.2k, 4k, 3k, and 2k. Audio may also be exported in stereo or 360 audio, called "Ambix." When using GoPro studio to combine the front and back to make an actual stitched 360 video, the name is adjusted to VIDEO_0002 upon export. The file was rendered using export destination "Editing" in 360 Media Resolution 5.2k with the audio setting Stereo. This is the highest export resolution allowed out of the GoPro fusion software at this time. The hex analysis below is of the rendered 360 file from the GoPro Fusion through the fusion studio software. You can see the changes that occur between the original recording and recompressed recording.
GPBK0002.MP4
Startup
VIDEO_0002.mov X
* Edit As Hcxsr Run Script Run Template: V.p-t.bt NT t>
14 6 6 •'4 •*9 i 0 VI >4 20 20 20 !.■ 5 C 00 tcypqt
n ■’4 20 20 00 00 00 01 *Q *4 74 00 00 00 qt ....mdat....
90 .’t ct iC 01 .9 00 15 tb 89
04 F7 vc 00 :5 ?D F8 00 C A 00 00 OB 00 01 . • 1.. 1®
• ' oc 4? .’D 00 54 01 FF A5 CO 02 G. 1.7. .'/V. .
00 oc 00 03 00 :e 00 ca 00 OF 00 02 CO 10 CO 03
• l'i . ‘1 :o 00 IS 0A 00 FF BB 1 4 6 C 0 OA y * * * • f .«
FF AB 0A 00 FF B1 92 Cl FF BO 00 02 FF AF 08 OC y«. .yt*. y*. .y ..
FF AD 00 VI) F9 • 7 (.?. fit- FF. 17 i 55 49 •1 *1 v .Vu t p. .GrJID
10 00 00 4*T ED y AC f.2 03 F.F A3 4 9 AA EO 3D 4 F ...oiY-a.it:*a-o
•• DC 0A D*7 4 4 41 so 45 OA 63 32 : 39 Ob . ■ DATS 1 ■
2D 31 30 30 38 oc CO 54 4D 4 • 06 00 CO 63 -10-Of! . .Tl.MK. . .c
31 31 3A 35 3*: 1A 32 32 54 49 41 4 3 H 6 3 11:56:22T2.V. .
31 31 >A 35 36 3A 32 3? 3A 30 30 00 55 46 52 41) 11 :S«: J.’:00.DKHH
04 00 00 ’< 00 HK KK 4» 52 45 f r. •••: ip.VKr.r.E
F8 01 00 00 00 00 00 00 00 00 00 00 00 00 00 o
00 00 00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 po 00 00 00 00
00 00 00 oo 00 CO 00 00 00 00 00 00 00 00 00 00
! I’f.r 00 00 00 00 00 CO 00 00 00 00 00 00 00 00 00
00 00 : o CO 00 00 00 00 00
00 00 00 00 00 CO 00 :o 00 00 00 00 00 00 00 00
: A1 f 00 00 00 00 CO 00 00 00 CO 00
00 00 00 00 00 CO 00 CO 00 00 00 00 CO 00 CO 00
00 00 00 00 oc CO CO 00 00 00 oo 00 00 00
00 00 00 00 00 CO 00 CO 00 00 00 00 00 00 00 00
D1E0I; 00 CO 00 :o 00 00 00
C’lFOh 00 00 00 00 CO 00 CO 00 00 00 00 00 00 00 00
00 00 00 00 00 00 oc 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 oo 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 OC 00 00
00 Oo 00 CO 00 00 00 00 oo
00 00 00 00 00 00 00 00 00 00 00 00 CO o o o 00
00 00 00 00 CO 00 00 CO 00
00 00 00 00 00 00 00 CO 00 00 00 00 00 00 CO 00
00 00 00 00 CO 00 . 6 00 00 00 00
00 00 00 00 00 :o 00 00 00 00 00 00 CO 00 CO 00
00 00 00 :o CO 00 00 FF Cl 00 00 • •••••!<•••• yA. .
FF CO 00 00 FF BF 00 10 FF BF. 00 09 00 44 CO 01 yA. .. .y>4. ..D..
•1 1A Hit 1* 1A 00 IB 02 J
00 1C 01 40 00 IF oc 00 00 ID 00 00 00 20 CO 00 ...?
IK 00 2 1 22 00 01 23 0 0 10 t..
OK r oc 02 00 ' 4 p 32 30 32 32 32 32 32 32 By 20222222
Opened file E:\thesis\J60videos\GOPRO STUDIOWIDEO 0002.mov'.
Figure 6 Flex Analysis of the Files Rendered out of the GoPro Fusion Studio Software. File is VIDEO_002.mov. Much
less information is now stored in the header.
21


itmt
Cole*
fommrnl
Si
Boxful ftyp Oh Mb Fd BO
stnKt no *ck n>p lwe-12] Oh 8* n »0
MHO Brand 6h 4h DO
V«*«ft 20050300* o> 4h 00
CompiftiMr Brand <1* l CM. 41*
mdat 9078C7JC* ro OO
tbutl Ixiihnadrr hd: mdat »4h 10b fo B«
D©.[2] moov 9O70C75O* 2*70* Oo
vbuct twifiradrr hd' rttouv [we- 10084) •X»?8C?SO*i Bit Fd By
BmKP) •mini 9070C7S8h 0C* ro 00
BootfJ] Irak 0078C7C4h tBEEMi Fo Bo
Don{2) trak 9070C JAIH 75** r* Oo
Bo«[3] Irak 9078E&O9'i 3431* Fo Bo
D0m14) udt* 9070tt*«O* JMh Fo Oo
f ir Type Hox
►V<*4 OM« Don*
Mowe Don*
Header Don* Tiatk Bo*
Track Don Tiatk Bo*
User Data Do*
Figure 6 Cont. Hex Analysis of the Files Rendered out of the GoPro Fusion Studio Software
Below is an atom breakdown using AtomicParsley to demonstrate the loss of metadata between the captured raw video and the rendered 360 video.
Table 7
Atoms of the Original Recording
AtomicParsley comparison between the original files and stitched files
Atom ftyp @ 0 of size: 20, ends @ 20_________________________________________
Atom mdat @ 20 of size: 238870404, ends @ 238870424__________________________
Atom moov @ 238870424 of size: 57166, ends @ 238927590_______________________
Total size: 238927590 bytes; 114 atoms total. AtomicParsley v0.8_____________
Media data: 238870404 bytes; 57186 bytes all other atoms (0.024% atom overhead). Total free atoms: 162 bytes; 0.000% waste.
Table 8
Atoms of the 360 Recompressed Video
Atomic Parsley stitched file:
Atom ftyp @ 0 of size: 20, ends @ 20 Atom mdat @ 20 of size: 1, ends @ 21 Atom dat @ 21 of size: 365, ends @ 386
Total size: -1871121856 bytes; 15 atoms total. AtomicParsley v0.8
Media data: 0 bytes; 2423845440 bytes all other atoms (-129.540% atom overhead).
Total free atoms: 0 bytes; 0.000% waste.__________________________________
In Summary, the important observation is the reduction in the amount of atoms between the original recording and recompressed the video. Even though the file size increases, the amount of data stored within the metadata of the file is much less.
22


YouTube Up Down
After rendering video out, the video was uploaded to YouTube using YouTube-dl. After a wait of 2-3 hours to ensure that all formats were downloadable from YouTube, the file was downloaded using the -F function in YouTube-dl:
Mp4 = merged 138,140 (YouTube format selection on YouTube-dl)
MKV= YouTube-dl automated choice
The Mp4 file is now at a frame rate of 29.97 and has a resolution of 5120x2560 The MKV file is now at a frame rate of 29.97 and has a resolution of 3840x2160
GPBK0002.MW
Â¥ fdit Av hr* v'
Startup
VIDEO 0002. mov
V70EO_0002_yt-QQa>aL093WV.ml Run Scrip1! ^ Run Template
If*, V vi A3 00 30 23 42 SI C I . EOl < !> * . .
42 I b: .U 42 ¥/ HI • 42 FI Hi 00 42 R . . R ‘ . R*'. . . R, nr.
61 12 • f 1 5 « ? K l : 97 £1 34 4 s5 02 ie aI ; t .ii‘ r.. ?! , . .
P DC 67 . 00 00 04 IXT H E 11 40 9B 74 40 8‘*t o-a.M>tP
- 4 !j *• ■■¥. n ; i BB -R AR • 4 4:* Au T
66 53 AC D1 E5 40 BB ?c 53 AH 16 54 AE 6B c ; IS .TSkS
01 3C BB S3 Ah : 12 ; CJ . S3 A * •r. - "A?;- •
P2 Cl E5 4t> BB BE 53 A * H4 1C 5 1 RR if 5 s AC 64 ,. AK»3S«... f»kS
; 2 k • 01 50
00 00 ?? 00 00 ?? 00 00 00 00 00 00 00
00 0® 00 00 G 0 00 co 00 00 30 00 03 00 05 00
00 00 00 00 00 00 CO 00 50 00 00 00 00 00 00 CO
00 00 00 00 00 00 00 00 00 00 00 00 00
00 00 00 00 00 00 00 00 00 00 00 00 00 00
4 9 • • 1 BF 4 E.» .I'f.......
: - 5*4 2A 07 ill 93 4 0 i: 90 01 4C i 1 7 i u K*** - i *. B?M€.I.4V5
1 , K 39 1 > 2K i ? 3 i 41 >-n 4C r ; .*» r r '. : IfA.I ivl
3S 37 2F. 3D 33 2E 31 â–  3 73 A 4 90 OF 9E C( 20 57.6 3.1
7E AK AK . • B CD 03 71 4 4 V 4- X K»\. . 1
: :â–  40 00 00 CO :. 5 4 RE *B 01 00 00 00 00 A*? TKk
9V bf £•1 |U ?♦’ 31 7 A AK 01 vis!
00 4B D7 31 l: 1 73 cs . â–  :l "1 00 22 BS 9C !' 1 . F> .. . a*. . ";;or*|
«- n r r r PS • 5F • i ■ • 93 01 2“> ¥. A u A ••! r v v?*.*..iTf
84 31 fV 22 BA £ 3 C1 00 CC 00 OC 00 03 IP H 3 62 a. . y"5«i........w,
K RA 5 1 : : 00 • • R • ' RA .. ,. t... u* .
01 55 B1 D1 0*1 BB 81 ;i 55 B9 c » 01 AE 01 00
00 3A : Cl 0.! 7 3 CL 0 1 32 . • - 1
?? ?.!; *C 3‘J fS t K f; -. HS 4: 5F 41 ■ • » A HI " }:■»* * f r. 71 A AAC f .
02 k 1 C 1 11 H 8 1 32 B5 $3 4C .a Y.. .i '*
99 BO 00 00 00 62 t 4 B1 20 ». i A2 62 32 to .1 f:. ...bl cc...
5 4 r.\ » : 00 00 1 r.» ?r v4 F> -TA 1
VS 11 73 73 : 00 co 00 OC- o o 00 . €3 CO 01 CO
i .• :: I IA j*.
15 ;• • p 7 45 4E 43 41 44 45 52 44 *7 DB *1 ? 41 7 r EttEN-ODEFDa.Lav
£{ 35 J7 2K 38 33 2K 31 30 30 73 73 31 it . 1.1
00 00 At f j CO i 00 00 OO 4 r 1 CS .. . f A A
81 32 <7 CD 00 IE 45 A3 4 .. t MH
41 4K 4 4 4C 45 52 5F 4K 41 â– it 45 44 J? CC S3 tF ASCLEK NAKEDtOS â– 
• F. •• 49 F 1 * F • 4 eC » 1 1 it iH.lt. i. • r -
00 OC 00 3A C 3 CO Cl 00 CO 00 00 00 03 04 63 ...: cA A
fc. l ! € t r v |! Ai 44 ..if. "Kf. :
ss 52 •1 . 49 4F 4E 44 9 7 94 30 30 M 30 33 3A HOI* : :
34 32 ;.k 30 30 39 3C 30 30 30 3C 30 00 73 73 ... • . . .■
03 00 00 00 00 »A »: 1 CO 1 OO 00 00 OO • rA
4 -:3 ,'C. 11 02 •' 1 • •;i 22 .. ■*■..-■ -
4S A3 6$ 44 tjtj 52 41 • 49 43* 44 D7 94 3*3 30 E£ DUHATIONDi"0(j
'r» a:: 10 1A >4 <:• 2F. • v- i • ■f'l
00 OC IF 43 b*: 75 Cl 00 CO 00 OC 50 33 69 BF 64 . . .Clu F 3i;.„
Opened fil* E:\thes.i\360v.d«os\GOPRO YOUTUBE UP DOWIM\VlDEO 0002 yt QQei>aL093WY.mkv\ Poi
Figure 7 Hex Analysis of the Files Downloaded from YouTube in the MKV File Type. File is VIDEO_0002_yt-QQco8L093WY.mkv. Information has now been changed again in the header.
23


1 Nan** Vtlu* Start si» Color Cflnmwflt 1
BOje[0] Pl>P Oh 2 Oh pfl- Bg File Type Box
struct box header hdr Ptyp [«ie=24] Oh flh ?r- Bg
Wfljor brand isom 8h 4h Eg: 0s
Hntx Vvson 200^1 Ch 4h Eg- 8s
Compatible Brand[o] isom 10b 4h Fg: Bg
Compatible Brand(l) iS02 Hh 4h rg- Bg
COrt^atiWe Brand[2] AvCl I8h 4h Fg: Bg
Compatible Brand[3] mp4l ICh 4h Eg: Bg
bcncli] tree 20h Oh Eg: Bg Free Space Bo*
StAKt boxheader hdr free [mc=0] 20h Oh Eg- Bg
Bcx[2] most 2Bh 3428O0EH Fg: Bfl Media Data Box
struct bo*header hdr S ii £ M 2flh Oh fg- BO
BC*[3) mocw 3428CJ6H AMAh Fg: Bg Mode Box
struct box header hdr mcov [sirt=43842] 342BC36h Oh Fg; Bg
8o<0] rnvhd 347«C3£h OCh Eg: Bg Mpvn? Header Box
6W[1] trak 3428CAAh +*D€h Eg- Bg Tr*ck &wr
Box[2] trak 332D188J1 05960 Fg: BO Track Box
Boocfa] udta 34337101 62h Fg: Bg: User Data Box
Figure 7 Cont. Hex Analysis of the Files Downloaded from YouTube in the MKV File Type
01=6X0002. MP4
f tc •. As. Hex v
Startup
VIDEO 0002 mov
VIDEO 0002 yt QC?co6L093VW.ml VID£0_0002_vtH5OcoSL003WY.mp4 X
Sui Sens', v Run Template: MM.St v t>
i • * • /, ;; M. f rK /; • • * irT* T I r*F eP - • * it yp I ‘ ‘r>! ! ! ! * ‘r-
e* 73 iF 32 Cl 7C €3 31 7 C 34 31 00 00 C9 r c, 72 C 5 €5
•. 0C K CD i 4 tl '4 13 c .• A 11 04 DK 74 C4 K . B2 .1 Jt . . T §• . At A /
IK IK. 24 00 OS ! cs A DA 55 A "4 i{‘ : Ri FT.' • - rR ..(...A. mi A 9 . *vy i:
'IF 1A 70 AF A 4 A 3B ?F AK 1C 7 E 75 PB A , 1, , . • . 1,; T. • ;0,
5A A 2D CC KK I>9 CD c 7 • 99 F 3 53 92 03 Kl 5C 51 7B 19 93 >:• : ,. .ivjl. *
H j »R b<: 11 : UK « .* a 4t IK 11 * A A I •'B • 11 4 1 A A; *iK. n . * j ;H . r: f.K . .
aa i ■ l * r.F cr» Fr» B* 7E F. 3 54 r-A f*F '10 FF. OF Rr ED RD •C.a-31 tlH
tc - * BD f7 ■4D 40 11 A4 Ft : 9 7C 57 bl IK *: 3 55 14 Uvw*» X- Sr
'•A DC iFi B 3K Od BF A2 R 9F 2 3 ?C 4C 41 A 4 B ' : • BA 62 50k.?.£4 Vl rjAnVC°t-
i‘ i r r AK F CC 07 FT CF R7 FB F.D 12 IA cn C 9 7 5
94 or 05 33 ED FC 29 30 ' I D1 5t E5 FE FA AB B9 BA 7L< 9 A
Ac DS :-a . ; »B i A At 95 3K : B IB 2D 07 jB C9 D5 FI 5 BF •i~K. J j
F.D FA IB 1A 3? D5 D7 **F 97 A4 F.9 1K 8C K.‘. 4r* ♦ A AA SA 10... 3*^. J.U - *7
EP IV OF 9E no • 1. 4 A 90 i'5 5B 13* 70 70 F5 F '• 71 CB 4 A IF. BB • • 1 >;, 11 •
DC u AF IB 73 F3 • t 2 1 C3 C C 5F IF FC KA 3A B5 DD 07 7B 4D • ..• !H AMs lit IK
•JU *K -; D7 77 *1 FC 4A a: 4F r FB lx • Jt CA :I„-AR>. . . Us.
04 F.B 03 FA 5P C F 71 23 ® 7 IE BF 75 17 05 3A r S DC 7 r. 51 f'CV jli. scOyQ
01 AB 73 : a K2 It K k: • .: 1 • 37 3* A : 2A 7C •; C 5 3 C4 74 ItfS.i.ll. ..'VN' - -‘A*.
AD KK 10 CO 22 A FI AK 3D FA FIB 93 B r rK 9D 1 C j 50 FK p. A" r. •'*'*'•*4 r. . . AP
»r Of 7B no M B • PA r-A *F. 91 A1 r*7 1C *A Ft ♦ 9 F7 R4 s-f-r:-9.1 ' n .
At EE n •:3 E3 07 t 7 rD Fa aA AD 2A A3 CD 07 A2 13 33 27 eo 13 XyOS * 1E"C. •1
cc 9K 2A 91 KK A2 42 DO 95 C J 27 » K r • f.l 55 ?C * I*/' *;CBD *• * ..U -j
F:* F n? F» 31 *■ r K.B BB DC. r AK 14 r 47 A 1 B 7P D.Cftp -1 !?• •. VI. .VRfr
ei • ■ L'A 44 nx BF FA 7 A 4 3 D3 0 5 42 *: 7 F0 FF 2E *-v*“ ; • >* 7CJ. 0! • •}*-> .•/.
TO 79 Bl : - ZD CA Ot A 7 15 15 DS> Cl .*4 K9 DC • OK - DS fryyiOif.. n. . v.iAi-C' .. C
12 FI •» s; A * ■ 1 >0 FF. nr FA : r IE 1 k FI • F7 :..#=••
59 77 7F er 3; 37 FB 7D 90 44 r r » IB 97 f.O BA 4? ^A PA Yv. 1201 ,PM.? O-Bff
72 D3 • CC o« cl 9b V A-* F4 FK Kr A FI 33 • c :CAA :A>l: »- D'.S2S3.
hF CD 04 ?C 32 FC BB HA 4 1 45 CO 5K 57 B: CD 57 : : 15 01 . X 2U»‘A KA Ki.X..
*;D 58 •T' ’4 IP F ’ 7F. BP 1>‘ FF. BF » I 5r» 4-c ' A- .'r . ' lr . h ;.1 AVI., 31
to A4 (6 BC Cl FA C2 F 4 IB 92 D5 35 BA 3" 20 Kl 1A A9 5A 00 Y«MMl>0.‘C5e7
: B > B BB 5D 2 j .*1 •*. FA K ’ B* /SB D R •*4 ID 9K .‘/.»M»d-0g1.«. *
r'l if 8 9 r»D AO C* CB AR 07 • . r *, 5F F.l 4 C 07 hr m 91 !t FK ‘. 1 . * ?.« . W AI.. u ’ V V
99 E2 a: •;5 ■ TO 1A 4C 9D , 1 •A El D9 92 10 AASty t * •-; .11* t A#
HD tb 45 '* 65 40 CC 09 EC FA 04 BB ■ OK 01 97 FB 75 BD 25 4»: KueH 1V. l U. . -tlt:*a*
4B E 25 rR ! B‘. B! -. •. '.: :.r i:
4C BC 7® AF • F2 D7 30 20 :a ec F2 C4 D4 E2 3A CC 09 9F 47 it*
A DC BB 71% 92 Cb 41' 5F AA 1 • : - 18 CK 99 74 : . wF 4V.K *’m. i i;.v". .
4 KK 10 F2 iV Ft ’4 85 if: 93 33 oc D5 t 9 EA 35 52 D1 . 1. A: f.t u.n>l . a: nyHf)
70 BP I 7 7 A 4*'’ 74 ro 45 39 77 Cl ■ *s r r: rF. 71 7 B E> Sr* Sl.J.r " ..VFn '>!•*.
5 o £2 H 29 AK • A CC CD 14 BB BD AF c t BC CO KF 3K 47 35 Ya. >>■: . :.»H US.Al - .it.
tD BK 95 ::f 21 : a : ■ : ■ C 9 *.A » AA .* b: 7«A BC 4 A FI 4 A ' A. r; •/ ! MV *S4
ID c; r.F 9F CF OK FA 7» 77 BB \ \ *K. IB da 53 Cf AF F'i B7 ns -AI!Y!. nj ■» ii.;-XXiA'fl
B9 EA 2D 55 F 9 72 4 C2 4 * F it 79 FF ED .•!-0 tl JO : !*»'. S C.1 yj.'l
53 11 D2 5F IB Ft 3i BO 97 AS 21 B7 92 S? fi” 7 A B9 D3 53 SaO *0ftC*-a! * 11 :*6s
;’H A i •F :,r*. cc r - FF * A 93 r A t * C!> ' A IA -A c? A 1 F. > • iTtlny A>: .: : .1
9C 9A cT> 05 7E 55 2C cc 03 94 Bl A 8 F 82 19 00 Cl EE 9F . "A.. SI* . ,AlY
tD ’b 4Z> 35 Db AK BB KD 7 A !b BA At 11 OB BK 02 47 74 rlHi-V\»i iz l^ib. 4«. Jt
CC TD 7F . ’ KA DC u 1. rK CO 14 DS f-j: i:: CO • 15 BO iC D4 BF. : | /. A? 5 *• {)*
4A F3 OF AF Ff 02 1C E2 El : *F 1 A A1 E3 50 5T> 44 .'r.I y.i.-.AxYis; i
39 Of 42 to Db 47 K3 bl 15 57 E5 a: 03 1 • DK FF $6 FC C2 SC ?■ V lijC il •. X \ . q»V • lb •
Opened file E:\tlietu\360xideoi\GOPRO YOUTUBE UP DOWNWIDEO 0002 yt QQeoBL093WY.mp4.
Pos:
Figure 8 Hex Analysis of the Files Downloaded from YouTube in the MP4 File Type. File is VIDEO_0002_yt-QQco8L.093WY.mp4. We now have additional information stored with in the header but the information is still
differs from the original.
24


'emplate Results - Bas*Mec-3.
Name Value Start Size Color Comment
ftyptoj Oh 20h fs- Ba:
F4etype(4] isom Kl 4h ?9- Ba: http://ftyp4.uxn/
ubyte data[20] Ch 14h Ffl: Ba:
uuidUJ 20b 397*i Fa: Ba:
ubyte moov(2) 3B7h 2360h Fa: Ba:
mvtKl(0] 3BFh 6Ch Fa: Ba:
Vakil] 428h C77h Fa: Ba:
trak[2) 10A2H 1613H Fa: Ba:
udQ[3] 26B5h 62h Fa: Ba:
2?l7h lD8€9h Fa: Ba:
ubyte data! 121065] Z7l7h 106C9h Fa; Be:
mdat[4] 20000h 1849385h Fa: Ba:
Figure 8 Cont. Hex Analysis of the Files Downloaded from YouTube in the MP4 File Type
Significate changes in the header have occurred with the recompression through YouTube. Even between the MKV file and the MP4 file the amount of data stored within the header varies due to the recompression of the files. New data has overwritten the original data.
PRNU Results
A "print screen HQ" was taken from the GoPro VR Player and a PRNU analysis was run against it. Print screen HQ was used due to the inability to export a frame and this was the best option provided with the current features of the GoPro VR Player. Two cameras were tested against the GoPro after resizing using Paint Shop Pro to match the abstract size of the print screen file (2020 xl371). Currently, this is the best practice for exporting a frame from an original recording without having to use the Fusion editing software. A comparison was made between a Sony Falcon and a Canon PowerShot. The red histogram is a comparison of the evidence PRNU against the suspect camera and database. The database contains videos from various other cameras PRNU. The blue histogram shows the suspect camera intra-variability while red histogram shows the suspect and data base intervariability. Finally, the green histogram shows the position of evidence against the two, suspect and database. The results were as follows:
25


PDF
GPVRP-render-2018-10-11-12-02-57-116.png vs 2020x1371-GoPro360-A CC RGB=0.43476; logLR=2.1181
CCS
Figure 9 GoPro Rendered File Comparison to GoPro.
GPVRP-render-2018-10-11-12-02-57-116.png vs 2020x1371-SONY-Falcon-MCE CC RGB—0.00019284; logLR-0.023094
The hypothesis that the evidence image was taken with this camera has NO support.
Figure 10 Go Pro Comparison to Sony Falcon.
26


PDF
GPVRP-render-2018-10-11-12-02-57-116.png vs 2020x1371-Canon-PowerShot-A510-CD CC RGB=-0.00025427; logLR=0.066369
The hypothesis that the evidence image was taken with this camera has NO support.
Figure 11 Go Pro Comparison to Canon PowerShot.
Figure 12 Image Used for First PRNU Tests. This was extracted using the "Print Screen HQ" function within the
GoPro VR Player software.
27


Figure 13 A Screenshot from the 360 Point of View Taken from the GoPro Studio Software Before Rendering.
Figure 14 A Screenshot from the Fish Eye Point of View Taken from VLC Upon Playback.
Attempting to playback recording 1 in VLC results in a fish eye image. The image is warped and the color temperature is slightly different.
28


Figure 15 Image Captured from the 360Fly as the Fisheye Perspective.
29


Figure 16 Image Captured from the 360Fly as the 360 Perspective, Camera Not Moved.
30


Audio Results
Using ffmpeg, the audio was extracted from the test recordings, renderingGPBK0102.mp4, GPBK0103.mp4, VIDEO_0102.mov, VIDEO_0102.mov. The stereo file was split to two mono channels to be read by Matlab script:
FFmpeg -i stereo.wav-map_channel 0.0.0 left.wav-map_channel 0.0.1 right.wav
When the video was recorded in the classroom with a low-level recording, the microphone picked up ENF as seen at 52 Hz and 120hz in the spectral analysis. To verify this finding, a separate recording was made outside without around any lighting sources, and no ENF was found in the second spectral analysis.
time [sec]
Figure 17Waveform, Spectrogram and MDCT Map.
31


Power/Frequency [dB/Hz] Power/Frequency [dB/Hz] Power/Frequency [dB/Hz]
GPBKD121-pcm-L
Figure 18 LCF Analysis.
As indicated in the graph above, there is a roll off just below 50 hz. This was found in multiple recordings and no settings to adjust this.
32


GPBK0121-pcm-L
frequency [Hz]
frequency [Hz]
Figure 19 LTAS Analysis.
GPBK0121-pcm-L
frequency [Hz]
Figure 20 LTA Analysis Two.
33


CHAPTER IV
DISCUSSION
The fact that exporting a file via the GoPro VR player as an uncompressed image is not possible, which presents multiple issues because without opening the video in the player, the original image is warped and will not play back in a native player. A screenshot can be taken via the "print screen" or "print screen HQ." The result is a PNG file from the print screen HQ. The only way to render a video that has been stitched together out of the native player is to use the freeware Fusion Studio. As noted, most of the metadata is lost when using this editing software. After rendering the 360 files, the file size increased while the meta data decreased. PRNU (Photo Response Non-Uniformity), which describes the gain between power on a pixel versus the signal output. It is well defined on the GoPro Fusion, and there is no support for the two selected files. This is one measure of detecting manipulation when metadata and hex analysis fall short.
It can be concluded from these results that the inability to play 360 videos at this time with native players while they are warped using conventional tools such as FFmpeg or even QuickTime presents multiple problems for the forensic community. Having a stitched file results in the file being laundered and losing a lot of metadata. It is evident by the size of the atoms that the amount of information removed to get a 360 video in a playable format is quite significant. Without additional information provided from GoPro, geometry calculations are difficult. Information can be found in the file headers, but no information is provided from GoPro. Also notable is that data on both the back and front lenses are listed, with the meta data containing differing serial numbers between the two lenses.
34


Conclusion
The tested hypothesis was substantiated. A reduction in data from the recompression of the original recordings was found. Future research should be conducted to evaluate newer versions of the GoPro Fusion as well as the GoPro Fusion studio software. Further testing should also be done on different videos resolutions, with other makes of 360 videos, and with other upload mediums, such as Facebook or Twitter. A recording should be collected while the camera is connected to the power source in the wall or, if it is still captured, then via the mic. Additional research on whether ENF information is stored within the video or just in the audio is also recommended.
35


REFERENCES
[1] Chennamma, H., & Rangarajan, L. (2010). Image splicing detection using inherent lens radial distortion. International Journal of Computer Science, 7(6), 149-159.
[2] Johnson, M. K., & Farid, H. (n.d.). Exposing Digital Forgeries Through Chromatic Aberration.
[3] Conotter, Valentina & Boato, G & Farid, Hany. (2010). Detecting photo manipulation on signs and billboards. 1741- 1744. 10.1109/ICI P.2010.5652906.
[4] Whitecotton, C. M. (n.d.). YOUTUBE: RECOMPRESSION EFFECTS. Retrieved from http://www.ucdenver.edu/ academics/colleges/CAM/Centers/ncmf/Documents/Theses/2017%20Fall/Whitecotton_Thesis_Fall2017.pdf
[5] Goodison, S. et al (2015). Digital evidence and U.S. criminal justice system. The Rand Corporation, https:// www.rand.org/pubs/research_reports/RR890.html
[6] Melton, L. (2016).Data transmission _ Parallel v. Serial. Retrieved from https://www.quantil.com/author/laurel/
[7] Robinson, S. (2012). The storage and transfer challenges of big data. Sloan Management Review. Retrieved from https://sloanreview.mit.edu/article/the-storage-and-transfer-challenges-of-big-data/
[8] Tsun-Li C. (2014). FP7 IAPP project. Digital image and video forensics. Retrieved from https://warwick.ac. uk/fac/sci/dcs/people/chang-tsun_li/
[9] Natarajan, M (2003). Multimedia and data technology: The challenges and delivery. Bulletin of Information Technology, 23(4), 19-26.
[10] NIST on Hash Value, January 04, 2017. Retrieved from 2018 https://csrc.nist.gov/projects/hash-functions
[11] Hannan, M. et al (2018). A method for verifying and authenticating digital media. Applied Computing and Informatics, 14(2), 145-158.
[14] SWDGE. Retrieved from https://www.swgde.org/documents
[15] Daubert v. Merrell. Dow Pharmaceuticals 509 U.S. 579 (1993).
[16] Frye v. United States, 54 App. D.C. 46, 293 F. 1013 (1923).
[17] Federal Rules of Evidence, 702—Testimony by expert witnesses.
[18] Matney, L (2017). Review: Go Pro Fusion 360. Retrieved from https://techcrunch.com/2017/12/19/review-gopro-fusion-360-camera/
[19] Best practices in digital video evidence. In video forensic expert. Accessed 4/30/19 at www.videoforensic expert.com/best-practices-for-digital-video-evidence.
[20] Scientific working group on digital evidence. Best practices for image authentication. Version: 1.0 (July 11, 2018).
[21] Scientific working group on digital evidence. Minimum requirements for testing tools used in digital and multimedia forensics. Version 1.0 (November 20, 2018).
36


[22] Bowers, A. (2018). Collecting and producing technically sound, legally defensible litigation data. Accessed 4/30/19 at www.doelegal.com/2018/08/collecting - and - producing - litigation - data/.
[23] GoPro, INC. Fusion User Manual. Fusion User Manual, GoPro, 2017. 130-24092-000 REVC.
[24] 360Fly., INC. 360Fly Manual. 360Fly Manual, 360Fly, 2016.
[25] The Brand Amp. 360Fly Reviewers Guide. 360Fly Reviewers Guide, The Brand Amp, 2014.
37


Full Text

PAGE 1

A N ALYSIS OF 360 VIDEO FILES ACROSS MULTIPLE RESOLUTIONS AND CAMERAS b y N ICHOLAS J AMES P ELC B.S., University o f Color ado Denver, 2015 A t hesis s ubmitted t o t he Faculty of the G raduate S chool of the University of Colorado in partial fulfillment of the requirements for the degree of Master of Science Recording Arts Program 201 9

PAGE 2

ii © 201 9 N ICHOLAS J AMES P ELC ALL RIGHTS RESERVED

PAGE 3

iii This thesis for the Master of Science degree by Nicholas James Pelc h as been approved for the Recording Arts Program b y Catalin Grigoras, Chair Jeff M. Smith Jeff Merkel Date: May 18, 2019

PAGE 4

iv Pelc, Nicholas James (M.S., Recording Arts Program) Analysis of 360 Video Files Across Multiple Resolutions and Cameras Thesis directed by Associate Professor Catalin Grigoras ABSTRACT Forensic experts rely on the validity and reliability of da ta consistent with Daubert standards and Federal rules of evidence for identifying findings and formulating opinions about evidence in legal proceedings. Digital forensic experts have an acute awareness of possible distortions or manipulations of electronically recorded or transmitted data. A critical aspect of analyzing digi tal evidence is to inspect the reliability , absence of , degradation , or alteration of data used in forensic evaluation. New recording devices are constantly being developed . One such new tool is the 360 video available for Go Pro players and , often , for broadcast on YouTube. 360 video provides an opportunity to capture, record, and review multi dimensional digital information. For this study , several test patterns are analyzed using the 360 video process. The purpose of the research is to identify any po tential degradation, distortion, manipulation , or data loss when 360 evidence is transferred from an origin ating device to a secondary device. The form and content of this abstract are approved. I recommend its publication. Approved: Catalin Grigoras

PAGE 5

v This thesis is dedicated to my entire family. Thank you to my father who , when I was at an early age , showed me the world of forensic sciences and to my mom whose patience is unheard of .

PAGE 6

vi ACKNOWLE DG MENTS world of d igital forensics. The infor mation I gathered while in this program is incredible and extremely relevant. I am looking forward to continuing this educational journey and learning more. Jeff , thank you for allowing me to work in the lab and on the DARPA project. Hopefully , all the work that we have done and conti nue to do will help many people. Thank you , Leah , for helping keep us all on track , and E . A thank you is due to Jeff Merkel for helping with your knowledge of Ambisonics in this evolving world of audio. Finally , thank you to my mom , dad , and dogs.

PAGE 7

vii T ABLE OF CONTENTS CHAPTER I. INTRODUCTION ................................ ................................ ................................ ................................ ............ 1 Problem Statement ................................ ................................ ................................ .............................. 1 Literature Review ................................ ................................ ................................ ................................ . 1 II . METHODOLOGY ................................ ................................ ................................ ................................ ........... 3 Go Pro Fusion ................................ ................................ ................................ ................................ ....... 6 YouTube dl ................................ ................................ ................................ ................................ ............ 6 Matlab ................................ ................................ ................................ ................................ .................. 6 010 Editor ................................ ................................ ................................ ................................ ............. 7 MediaInfo ................................ ................................ ................................ ................................ ............. 7 Exif tool ................................ ................................ ................................ ................................ ................. 8 FFmpeg ................................ ................................ ................................ ................................ ................. 8 AtomicParsley ................................ ................................ ................................ ................................ ....... 8 Go Pro VR player ................................ ................................ ................................ ................................ ... 8 360Fly app ................................ ................................ ................................ ................................ ............ 8 GoPro app ................................ ................................ ................................ ................................ ............. 8 Fly360 Director ................................ ................................ ................................ ................................ ..... 8 VLC ................................ ................................ ................................ ................................ ........................ 8 Paint Shop Pro ................................ ................................ ................................ ................................ ...... 8 III . RESULTS ................................ ................................ ................................ ................................ .................... 12 YouTube Up Down ................................ ................................ ................................ .............................. 23 PRNU Results ................................ ................................ ................................ ................................ ...... 25 Audio Results ................................ ................................ ................................ ................................ ...... 31 IV . DISCUSSION ................................ ................................ ................................ ................................ .............. 34 Conclusion ................................ ................................ ................................ ................................ .......... 35 REFERENCES ................................ ................................ ................................ ................................ .................... 36

PAGE 8

viii LIST OF TABLES TABLE 1 . Metadata ................................ ................................ ................................ ................................ ...................... 12 2 . Video ................................ ................................ ................................ ................................ ............................. 13 3 . Audio ................................ ................................ ................................ ................................ ............................. 16 4 . Other 1 ................................ ................................ ................................ ................................ .......................... 18 5 . Other 2 ................................ ................................ ................................ ................................ .......................... 19 6 . Other 3 ................................ ................................ ................................ ................................ .......................... 19 7 . Atoms of the Original Recording ................................ ................................ ................................ ................... 22 8 . Atoms of the 360 Recompressed Video ................................ ................................ ................................ ........ 22

PAGE 9

ix LIST OF FIGURES F IGURE 1. Illustration of GoPro Fusion Different Resolutions ................................ ................................ ........................ 3 2. GoPro Fusion Diagram ................................ ................................ ................................ ................................ ... 4 3. Diagram of 360Fly ................................ ................................ ................................ ................................ .......... 5 4. MDCT E quation ................................ ................................ ................................ ................................ .............. 6 5. Hex A nalysis : Original ................................ ................................ ................................ ................................ .. 20 6. Hex Analysis of the F iles R endered out of the GoPro Fusion Studio S oftware ................................ ............ 2 1 7. Hex Analysis of the Fi les D ow nloaded from YouTube in the MKV F ile T ype .. . ............................................ 23 8. Hex Analysis of the Files Downloaded from YouTube in the MP4 File Type ................................ ............... 2 4 9. GoPro R endered F ile C omparison to GoPro. ................................ ................................ ............................... 26 10. GoPro C omparison to Sony Falcon. ................................ ................................ ................................ ............. 26 11. GoPro C omparison to Canon PowerShot. ................................ ................................ ................................ ... 27 12. Image U sed for F irst PRNU T ests ................................ ................................ ................................ ................. 27 13 . A S creenshot from the 360 P oint of V iew T aken from the GoPro Studio S oftware B efore R endering. ...... 28 14 . A S creenshot from the F ish E ye P oint of V iew T aken f rom VLC U pon P layback. ................................ .......... 28 15 . Image C aptured from the 360Fly as the Fisheye P erspective. ................................ ................................ ...... 29 16 . Image C aptured from the 360Fly as the 360 P erspective , Camera N ot M oved ................................ ........... 30 17 . Waveform, Spectrogram and MDCT Map. ................................ ................................ ................................ ... 31 18 . LCF A nalysis ................................ ................................ ................................ ................................ ................... 32 19 . LTAS A nalysis ................................ ................................ ................................ ................................ ................ 33 20 . LTA A nalysis T wo ................................ ................................ ................................ ................................ .......... 33

PAGE 10

1 CHAPTER I I NTRODUCTION Problem S tatement 360 v ideo has become an increasingly popular multidimensional recording device , yet the 360 technology has not been subject to a thorough analysis concerning its reliability with data transfer. For the technology to meet one of the basic Daubert standards, 360 v ide o require s testing and assessment of its consistency following transfer from the origin ating device to other media. Without such a n analysis , the 360 video technology might not be considered acceptable forensic evidence under Daubert standards , which requi re the demonstrated validity and reliability of assessment tools. Literature Review Over the past two decades, advances in technology have created significant opportunities for and challenges to the collection and analysis of digital evidence for both solving crimes and preparing cases for court. In a white paper prepared for the National Institute of Justice, Goodison, Davis, and Jackson (2015) provided an overview of digital evidence in the criminal justice system. They underscored major themes , including that , tion and chain of custody They also cite d the growing importance of video evidence in criminal investigations. Inherent to the growing importance of digital evidence is adherence to rules of evidence and case law. Testimony by expert witnesses (Federal Rules of Evidence, 702) requires that the presentation of data meet s standards of adequate reliability and validity to be considered as a basis for the findings and opinions provided by an expert. The courts have moved from a general acceptance standard , i n Frye vs . US , 293F. 1013 (1923) , to a scientific standard , as defined in Daubert vs . Merrell Dow Pharmaceuticals, 509 US 579 (1993). New technology, then , requires the development of research that supports these Federal standards and withstands scientific scrutiny. In the development of video evidence, data should both be recorded accurately and , in the chain of evidence , accurately transferred from one media device to another. Data transmission is the process of transferring data between two or more digital devices, and such transmission may involve either serial or parallel transfers (Melton, 2016). Robinson (2012) has i dentified some of the problems that can be encountered in data transmission . Others (e.g., Natarjan , 2003; Tsun Li, 2014 ; ) have addressed the manner by which these transfer

PAGE 11

2 problems can be addressed . Recent research has also presented methods for verifyin g and authenticating transmitted video data ( Whitecotton , 2017; Harran et al . , 2018).

PAGE 12

3 C HAPTER II M ETHODOL O GY Multiple c omparison p aradigms were used to test . These paradigms can be summarized as follows. Understanding this compar i son and analysis is facilitated by a detailed examin ation of the GoPro Fusion. The GoPro F usion has two lenses attached t o the body of the camera. There is a front facing lens and a back facing lens. Both lenses have an f stop of 2.0. It is essentially two cameras in one body. The GoPro records to two separate micro SD cards , which work in tandem. While recording on the Fusion, the operator can either record using a standard electrical ou tlet or the internal battery. The outlet on the camera is a USB C style jack. The portable battery life of the GoPro is around 1.5 hours maximum. The system can be navigated using the button the front of the unit o r by connecting to a cell phone v ia W i F i and using the GoPro app lication . Data can be written through the application, or to a mobile device, or the internal micro SD cards. Additional information can be stored to GoPro s cloud. The max imum resolution that can be recorded on the GoPro is 5.2K at 30 FPS ( frames per second ) , NTSC or 25 FPS, PAL. O r 3k at 60 FPS . Frames per second are the number of frames of video that are captured each second when recording . The term 5.2k refers to the horizontal lines and the vertical numbe r of pixels. At 5.2K , there are 4992 lines with a width of 2496 pixels. At 3k , there are 3000 horizont a l lines and a width of 1504 pixels. An illustration of the different resolutions is shown below . Figure 1 Illustration of GoPro Fusion Different Resolutions . The aspect ratio of the recordings is made at 16:9. Aspect ratio is the ratio of the width to height. In this case, the image is 16 inches wide and 9 inches high. The purpose of a Go Pro 360 camera is to capture not only a front view but a complete image of the entire surround of the camera. The field of view (FOV) refers to how

PAGE 13

4 much of t he subject is captured through the lens. The measurement is in degrees so that with the Go Pro camera , 360 degrees of the field of view. Therefore, the camera will capture the entire surrounding . The GoPro also has six (6) ax e s of stabilization. This increases camera stability in order to maintain smooth and steady shots. Protune is another feature on the GoPro Fusion as well as most of the products sold by the manufacturer . This feature allows the operator to be able to adjust both the International Standards Organization ( I SO) and the Exposure Value (EV) level. ISO is used to adjust the overall brightness of the image. With a lower ISO setting, the image will be darker , but there will be less noise. With a higher ISO setting, t he image will be brighter but will contain more noise. For the purpose of the test s used in this study, various ISO settings were used depending on the lighting situation. The settings available were 400, 1600 and 6400. EV, exposure value, ranges from 2.0 to +2.0. The default setting and what w as recorded w hich was the EV set to ( 0 ) . The higher the EV s et ting, the higher the brightness of the image. Pro tune is the ability to adjust the camera from auto settings to have more control by using variable IS O and EV settings . The GoPro Fusion , like many other digital cameras , has the capacity to adjust the date and time of the image acquisition. T his information is stored within the metadata. W hen the device is connect ed either to the GoPro app lication or to the proprietary software, Fusion Studio, the date and time will auto adjust to whatever the actual device setting . This adjustment is relevant if the device is recovered for time offset purposes. Location of the device using GPS information is recorded as well. This information is only ava ilable when the fusion is connected to the GoPro app lication . This information is then stored within the metadata of the file. The GoPro features four built in microphones in order to capture audio all around the camer a. The figure below depicts the various feature of the Go Pro 360 camera. Figure 2 GoPro Fusion Diagram .

PAGE 14

5 Another camera used in the tests in this study was the 360 Fly. This camera s or i ginal design purpose was f or aviation and military use. Th e camera s navigation is controlled by a single button that can be use d to pair to another device for total operating control. Once paired to another device, even more control over settings of the Fly can be maintained . Adjustm ents can be made for variables such as saturation, contrast, f speed and brightness. These are not necessarily detailed pro controls. They all feature a single slider , and they do not give numerical values of any specific degree of change is being made. Adjustment are not finite either. For the purposes of all test recordings performed in this study therefore , all settings were left at a neutral point. The Fly , like the GoPro, can record in many different resolutions. The maximum resolution that ca n record ed is at 2880 x 2880 with 30 frames per second. The aspect ratio is not 16:9 unless the camera is set to POV mode. Once that setting is made , further adjustments cannot occur including any changes to frame rates either. This camera features a sing le lens instead of having two lenses . Th e lens on the Fly is a single fisheye . A fisheye lens is one that shoots extremely wide images . The single lens can go to f 2.5. The field of view is again slightly different. Vertically , it has 240 degrees , and horizontally it has 360 degrees. Instead of recording two a SD card, the 360 Fly records internally with a total space capacity of 64gb. The battery life is slightly longer than the Go Pro, with the average time being two hours of battery life. The F l y 360 charges via a docking system that uses USB 2.0 technology. Like the Go Pro fusion, the F ly 360 stores GPS data within the metadata. This feature can be turned on or off within the settings on the Fly360 app lication . The features if the Fly are shown above. Figure 3 Diagram of 360Fly .

PAGE 15

6 Various d e finitions are ava ilable for image recording . For the purpose of this study , recordings were made at the highest possible resolution to capture and store the most maximum amount of data possible. This procedure w ould allow the capture and analysis of the m aximum change that occurs. A ll test recordings were at 5.2k on the GoPro at 30 frames per second. The Fly 360 recording s were at 2880 x 2880 at 30 frames per second. While it is po ssible to record with t he Fly at 6 0 frames per second, the resolution would drop to 1728 x 1728. Six recordings were collected and several types of data analysis were performed in order to compare any changes in the data acquisition and transfer . Files were then rendered out of the GoPro Fusion freeware, causing recompression in the files. The same process was done using the Fly360 test recordings. Th e recompressed files from the GoPro Fusion software were then recompressed through YouTube and an alyzed. The Go Pro Fusion software was used to create the 360 videos in multiple resolutions. Recording number one was used to create rendered file, VIDEO_0002.mov. YouTube dl was used to download the laundered videos from YouTube. Recording number one fr om the GoPro Fusion software, VIDEO_0002.mov, was used for the uploaded video to YouTube . The two YouTube files were created from this test as well, VIDEO_0002_yt QQco8L093WY.mkv and VIDEO_0002_yt QQco8L093WY.mp4 Matlab was used to analyze the PRNU, MDCT and ENF of the recordings. PRNU, (Photo Response Non Uniformity can be extracted and analyze d against the test databa se to see if the camera has support, limited support or no support to the images in the other camera in the database. MDCT was also analyzed. MDCT (Modified Discrete Cosine Transform) is designed to be performed on consecutive blocks of large data sets. It is not applied to an audio signal directly so that interference with the original data is minimized. The GoPro Fusion records audio in an AAC format. AAC is a ISO standard format that contains MDCT. Figure 4 MDCT E quation .

PAGE 16

7 LTAS (Long Term Average Spectrum) was performed on the audio data in order to record and demonstrate the mean, max, and min or the power. This helps us to define the overall recordings in the audio channel. ENF (electrical network frequency) was also evaluated in the audio recordings occurring in this study. ENF measures where noise from the main power supply is recorded into the actual recording. LCF, (low cut filter) was also measured. This variable provided information to determine if at a cer tain point, the audio recording has a low cut filter that removes data from the low frequency range. Recordings number one, GPBK0037.mp4, was used to analyze the PRNU. A Print Screen HQ was taken for analysis against the PRNU database. A Print Screen HQ w as used as that is the best possible way currently, to extract a single frame from a recording using the native player. This recording had the best exposure settings for a neutral test. Recording number three, GPBK0121.mp4, was used for the MDCT, LCF and L TAS analysis. This recording was longer in length with more background noise for more information for analysis. This simulates a more realistic environment. Recordings four, GPBK0068.mp4 , five, GPBK0129.mp4, and six, GPBK0130.mp4, were used for ENF analys is. Recording four was made in an outdoor location, five was made in the same location as one and two but for a greater duration, under fluorescent lights and run off battery power. Six was made with the same conditions as five but plugged into the wall fo r power instead of running off battery power. 010 Editor was used to perform a hex analysis of the original recordings, recompressed recordings through YouTube and recompressed recordings through the creation of the 360 videos with the GoPro Fusion software . Hex analysis (hexadecimal) allows examination of the bytes of the file . From there, an analysis of the raw information can be made without it being encoded . Analysis of both the header and footer of this information can be completed to observe if any changes have been made from recompression, or any other changes have occurred including for example if the file has decreased or increased in size, and what is missing or has been added . Test recording one was used for this ana lysis as well as the first render from GoPro Studio and the two downloaded files from YouTube. MediaInfo was used to assess the metadata and changes that occurred between the recompression of the videos in each stream. Metadata analysis was also used for this study to obtain a comparison between information is stored within the files . Test recording one, GPBK0037.mp4, the first render from the GoPro Studio , VIDEO_0002.mov, was used, the two downloaded files from YouTube were used VIDEO_0002_yt -

PAGE 17

8 QQco8L093WY .mkv and VIDEO_0002_yt QQco8L093WY.mp4 , test recording number one from the Fly360 , FLY08401.mp4 was used and the same recording but rendered out through the Fly360 director FLY08401(CONVERTED).mp4 was used in this test. Exif tool was used to verify the results from the MediaInfo information. All recordings were used in this test. FFmpeg was used to extract the audio stream from the video. All test recordings, one through six, had the audio stream extracted for analysis. AtomicParsley was used to demonstrate the loss in atoms upon recompression of the video. Recording one , GPBK0037.mp4 and the recording rendered from the GoPro Fusion software, VIDEO_0002.mov were used. The Go Pro VR player was used to playback the origina l recordings without having to render them and effectively alter the video as well as to create a print screen of the original file. The Print Screen HQ function was also utilized to extract the image for all of the PRNU tests. The GoPro app was used to s tart and stop all recordings made. The 360 Fly app was used to start and stop recordings and to adjust camera settings to the highest possible resolution . Fly360 Director was used for the creation of the 360 videos from the additional 360 camera. VLC was used in an attempt to play back the 360 video files without the GoPro native player. All test recordings were attempted to playback in VLC. Paint Shop Pro was uti lized to resize the images for the PRNU analysis. This was due to the unique size of the frame exported from the GoPro VR player. In addition to comparing the two cameras, an examination was completed of the files after being uploaded and downloaded to Y ou T ube. A popular way of sharing videos is by uploading them to the Y ou T ube site. It is important to see the change that occurs between the or i ginal video and the recompression through Y ou T ube. When recording to the GoPro Fusion , two files are created upon clicking record which captures a separate file per each lens. When these files are played back, they are warped and not playable as 360 videos yet. These videos must be imported into the GoPro Fusion software to create the 360 FOV videos. Two pieces of software are provided with the GoPro Fusion to aid with this data transfer . Upon rendering from the GoPro Fusion Studio ,

PAGE 18

9 several options are available including Video codecs, H.264, Cine F orm For the purpose of this study, CineForm was used because it is likely to produce the least amount of recompression . Video resolutions was a second option. From 5.2k (4992 x 2496), 4k (3840 x1920), 3k (2880x1440), 2K (2048 x 102 4_ are all ava ila ble options. 5.2K was used in this study becaus e it contains maximum data for the present study . Spatial Audio is the third option. Stereo was used for this stud y but 360 audio ( A mbix ) is another available option. Finally , DWarp ( Paral l ax Compensation) is the final option. This setting was left on for this stud y because DWarp remove s parallax lines in the raw stitched footage. Parallax is a displacement or difference in the apparent position of an object when viewed along two different lines of sight and is measured by the angle or semi angle of inclination between those two lines. Finally, the files were taken and uploaded to YouTube . This was to measure the recompression of the 360 files compared to the original data set to see what changes will occur upon download of the files. For the data a nalysis , best practices for measuring authenticity or recompression in transferred data were utilized for transfer from GoPro warped original files to a 360 playable format upload ed and download ed from YouTube . T hese best practices have been written as guidelines by the S cientific W orking G roup on Di gital E vidence (SWGDE). Image authentication is necessary to determine whether image data are accurate and valid representations of subjects and events. Importantly, these guidelines do not define specific analytic techniques or tools but a process to detect stag ing or manipulation of images in the manner of acquisition or transfer of images from one recording device to another. This process requires that the original image should be preserved , and any transferred image should maintain the integrity of the initial ly acquired data image. In this study, the transfer of data from the 360 G o P ro to another device for data preservation is tested . An illustration of this image transfer and integrity are described in the article from SWGDE on Best practices for digital v ideo evidence The d iscussion of this analysis includes the following topics : 1. I mplications for authenticity and reliability when making data transfers of 360 video ; 2. D irections for future research ; 3. Best practices for collecting video data from the recording device;

PAGE 19

10 The S cientific W orking G roup on D igital E Mi nimum (SWGDE, 2018). These guidelines differentiate the critical forensic tools for preservation, acquisition, hashing, and wiping digital data. Multimedia tools for imagery enhancement and analog video capture are also addressed . The purpose of digital testing evidence is to assess the confidence level which can be assigned to a tool or procedure to perform correctly and reduce the risk of errors. In this study, the acquisition and transfer of video images using the GoPro Fusion 360 as well as the Fly360 was tested regarding the accuracy and validity of data move d from an original recording device to a second device. This process of transfer of video images is described in an applied legal article on the collection and production of technically sound and defensible digital dat a in a litigation process by Bowers (2018). H e described several PDF pitfalls. Producing native files which hold all the metadata might be opened, corrupted, or inadvertently altered. Another problem can occur when multiple documents are produced in a sin gle PDF without the original metadata being shown in a corresponding load file. The receiving party might then have to separate documents which could open the door to challenges from opposing counsel or even court sanctions. B owers strongly urged the use o f document specialist s in the processing and management of native files. The specialist can identify common issues which are inaccurately collected and produced data. B owers also noted that the common legal practice of Bates N umbering c ould produce even obvious errors in identifying where one document ends and the next one begins. He summarized the top fo u r self collection issues (Bowers, p.5): 1. A single PDF was created, which contained multiple documents without any delineation between those documents. 2. Metadata had been altered by employees simply forwarding requested emails. 3. Emails had been printed and then scanned into PDF which remov ed native metadata. 4. The inability to easily identify the melody of relationships of files. E (Bowers, p.7 ) He U nder the newly amend ed FRE 902, metadata will not be self authenticating unless a qualified person has inspected the data, recorded the process used, and certified that an exact copy of the data was created 6) . Th is study was designed to test the hypothesis that there will be a significant change in the video data resulting in a loss of data accuracy due to recompression.

PAGE 20

11 As a starting point to eliminate misleading data, SD cards were formatted on recording devices before use. To avoi d camera motion , all recordings were made using a cell phone to start and stop recording. Version 01.70.00 of GoPro Fusion was used .

PAGE 21

12 C HAPTER III RESULTS The metadata and file structure of the non combined files were analyzed . 360 videos were captured to the G o P ro fusion via two micro sd cards , one for each side of the camera , which were labeled 1 100GBACK (BACK) and 2 100GFRT (FRONT) . One test recording was created for this research . One recording resulted in two files : GPBK0002 and GPFR0002. The table s below are a comparison of the metadata of the data collected through test recordings. Table 1 is of the metadata. You can see the format as well as file size changes between recompression. Table 2 is of the video content. Variances occur in multiple items in this table. From the aspect ratio to the bit rate. Table 3 is of the audio content and tab le. Again, multiple changes occur here. From the sample rate to even the duration. Four, five and six is additional metadata. In the original GoPro recording this information ngs, this information is not stored. Table 1 Metadata Informa tion Original Recording (Recording 1 , GPBK0037. mp4 ) Go Pro Studio (Video_0002 .mov ) YouTube Up Down MKV (VIDEO_0002_ yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_yt QQco8L093WY. mp4) Fly 360 Original Recording ( Fly360 Recording 1 , FLY08401. mp4 ) Fly 360 Rendered from the 360 Director software ( Fly360 Recording 1 Render , FLY08401(CONVERT ED.mp4 ) Format MPEG 4 MPEG 4 Matroska MPEG 4 MPEG 4 MPEG 4 Format Version n/a n/a Version 4 / Version 2 Format Profile N/A Quicktime n/a Base Media Base Media / Version 2 Base Media Codec ID mp41 (mp41) qt 2005.03 (qt ) n/a isom (isom/iso2/avc1 /mp41) mp42 (mp42/iso m) isom ( isom /iso2/avc1/mp 41) File size 228 MiB 2.26 GiB 15.5 Mb/S 52.2 Mib 54.4 Mib 24.4 MiB Duratio n 42 s 309 ms 42s 42ms 42 s 98ms 42 s 98ms 9 s 941ms 9 s 963 ms Overall bit rate mode Variable Variable Variable

PAGE 22

13 Table 1 Cont Metadata Overall bit rate 45.2 Mb/s 461 Mb/s 15.5 Mb/s 10.4 Mb/s 45.1 Mb/s 20.6 Mb/s Writing application n/a n/a Lavf57.83.100 Lavf57.83.100 Lavf57.25.100 Writing Library n/a n/a Lavf57.83.100 n/a Error Detection Type n/a n/a Per Level 1 n/a xyz +00.0000+000.0000/ Table 2 Video Information Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002 _yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt QQco8L093W Y.mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVER TED.mp4) ID 1 1 1 1 1 1 Format AVC CineForm VP9 AVC AVC AVC MultiView_ Count 2 Format/Info Advanced Video Codec Advanced Video Codec Advanced Video Codec Advanced Video Codec Format Profile High@L5.1 High@L5.1 Baseline@ L5.1 Baseline@L5.1 Format Settings CABAC/1 Ref Frames 2 Ref Frames 1 Ref Frames 1 Ref Frames Format settings, CABAC Yes No No No Format settings, RefFrameS 1 Frame 2 Frames 1 Frame 1 Frame Format settings, GOP M=1, N=15 M=1, N=30

PAGE 23

14 Table 2 Cont Video Codec ID Avc 1 CFHD V_VP9 Avc 1 Avc 1 Avc 1 Codec ID/Info Advanced Video Coding CineForm High Definition (HD) wavelet codec Advanced Video Coding Advanced Video Coding Advanced Video Coding Duration 42 s 309 ms 42 s 9ms 42s 9ms 42s 9ms 8 s 876 ms 9s 527ms Source Duration 9 s 491 ms Bit Rate Mode Variable Variable Bit Rate 45.0 Mb/s 460 Mb/s 10.3 Mb/s 46.4 Mb/s 21.3 Mb/s Width 2704 pixxels 5120 pixels 3840 pixxels 5120 piels 2880 pixels 3840 pixels Height 2624 pixxels 2560 pixels 2160 pixels 2560 pixels 2880 pixels 1920 pixels Display Aspect Ratio 1.030 2.000 16:9 2.000 1.000 2.000 Frame Rate Mode Constant Constant Constant Constant Variable Constant Frame Rate 29.970 (30000/1001) FPS 29.970 (29970/1000) FPS 29.970 (30000/1001) FPS 29.970 (30000/1001) FPS 28.027 FPS 28.027 FPS Minimum Frame Rate 1.542 FPS Maximum Frame Rate 37.943 FPS Color Space YUV YUV YUV YUV YUV Chroma subsampling 4:2:0 4:2:0 4:2:0 4:2:0 Bit Depth 8 Bits 8 Bits 8 Bits 8 Bits Scan Type Progressive Progressive Progressive Progressive Progessive Bits/(Pixel*Frame) 0.212 1.171 0.026 0.200 0.103 Stream Size 227 MiB (100%) 2.25 Gib (100%) 52.5 MiB (99%) 49.0 MiB (98%) 24.1 MiB (99%) Source Stream Size 52.4 Mib (98%) Title GoPro AVC Video Handle Language English English English English Default n/a n/a Yes Forced n/a n/a No Encoded Date UTC 2018 08 23 11:37:56 UTC 2018 10 08 18:56:14 UTC 2018 10 24 21:56:54 Tagged Date UTC 2018 08 23 11:37:56 UTC 2018 10 08 18:56:14 UTC 2018 10 24 21:56:54 Color Range Full n/a Limited Limited

PAGE 24

15 Table 2 Cont Video Color primaries BT.709 BT.709 BT.709 BT.709 Transfer characteristics BT.709 N/a Bt.709 BT.709 Matrix coefficients BT.709 BT.709 Bt.709 BT.709 Mdhd_Duration 8876 Writing library x264 core 148 r2597 e86f3a1 *Only found on Fly rendered : Encoding settings : cabac=0 / ref=1 / deblock=0:0:0 / analyse=0:0 / me=dia / subme=0 / psy=1 / psy_rd=1.00:0.00 / mixed_ref=0 / me_range=16 / chroma_me=1 / trellis=0 / 8x8dct=0 / cqm=0 / deadzone=21,11 / fast_pskip=1 / chroma_qp_offset=0 / threads=12 / lookahead_threads=2 / sliced_threads=0 / nr=0 / decimate=1 / interlaced=0 / bluray_compat=0 / constrained_intra=0 / bframes=0 / weightp=0 / keyint=250 / keyint_min=25 / scenecut=0 / intra_refresh=0 / rc=crf / mbtree=0 / crf=23.0 / qcomp=0.60 / qpmin=0 / qpmax=69 / qpstep=4 / ip_ratio=1.40 / aq=0

PAGE 25

16 Table 3 Audio Informati on Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002_ yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4) ID 2 2 2 2 2 2 Format AAC PCM AAC AAC AAC AAC Format/I nfo Advanced Audio Codec N/a Advanced Audio Codec Advanced Audio Codec Advanced Audio Codec Advanced Audio Codec Format Settings n/a Little/Signed n/a Format Profile LC n/a n/a LC LC LC Codec ID mp4a 40 2 sowt A_AAC 2 mp4a 40 2 mp4a 40 2 mp4a 40 2 Duration 42 s 304 ms 42s 42ms 42s 98ms 42 s 98 ms 9 s 941 ms 9 S 963 ms Source Duration n/a 42s 9ms n/a Bit Rate Mode Constant Constant Constant Variable Bit Rate 128 kb/s 1536 kb/s 126 kb/s 96.0 kb/s 126 kb/s Channel( s) 2 Channels Channel Positions Front: L R Sampling Rate 48.0 kHz 44.1 kHz 44.1 khz 48 kHz 48khz Bit Depth 16 bits Frame Rate 46.875 FPS (1024 SPF) 43.066 FPS (1024 SPF) 43.066 FPS (1024 SPF) 46.875 FPS (1024 (SPF) 46.875 FPS (1024 SPF) Compres sion Mode Lossy Lossy Lossy Lossy Lossy Stream Size 662 KiB(0%) 7.70 Mib (0%) 645 kib (1%) 117 kiB (0%) 153 Kib (1%) Title GoPro AAC 7.69 MiB (0%0 Languag e English English English Default Yes Yes Yes Alternat e group 1

PAGE 26

17 Table 3 Cont Audio Forced No Encoded Date UTC 2018 08 23 11:37:56 Tagged Date UTC 2018 08 23 11:37:56

PAGE 27

18 Table 4 Other 1 Informa tion Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002. mov) YouTube Up Down MKV (VIDEO_0002_ yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY084 01. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4) ID 3 n/a Type Time code n/a Format QuickTime TC n/a Duratio n 42 s 309 ms n/a Time code of first frame 12:01:17:2 1 n/a Time code , striped Yes n/a Title GoPro TCD n/a Languag e English n/a Encoded Date UTC 2018 08 23 11:37:56 n/a Tagged Date UTC 2018 08 23 11:37:56 n/a Bit Rate Mode CBR n/a

PAGE 28

19 Table 5 Other 2 Informa tion Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002. mov) YouTube Up Down MKV (VIDEO_0002_ yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt QQco8L093WY .mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4) Type meta Duratio n 42 s 42ms Bit Rate Mode CBR Table 6 Other 3 Informatio n Original Recording (Recording 1, GPBK0037. mp4) Go Pro Studio (Video_0002 .mov) YouTube Up Down MKV (VIDEO_0002_ yt QQco8L093W Y.mkv) YouTube UpDown MP4 (VIDEO_0002_ yt QQco8L093W Y.mp4) Fly 360 Original Recording (Fly360 Recording 1, FLY08401. mp4) Fly 360 Rendered from the 360 Director software (Fly360 Recording 1 Render, FLY08401(CONVERT ED.mp4) Type meta n/a mdhd_Dur ation 42309 n/a Bit rate mode VBR n/a Below, is the first hex analysis. The file was imported into 010 editor. You can see information such as the version number, name of the camera and the file type. This matches what we have found in the mediainfo data above.

PAGE 29

20 Figure 5 Hex A nalysis: Original. Note. In hex analysis, front lens still notates back and front. Test is done 0002 and (Recording 1 GPBK0037.mp4)

PAGE 30

21 During export, a selection of destinations is offered : editing , Facebook, YouTube , or Vimeo. Different resolutions may be selected as well : 5.2k, 4k, 3k, and 2k. Audio may also be exported in stereo or 360 audio , called When using GoP ro studio to combine the front and back to make an actual stitched 360 video , the name is a djus ted to VIDEO_0002 upon export . The file was rendered in 360 Media Resolution 5.2k with the audio setting Stereo. This is the highest export resolution allowed out of the Go P ro fusion software at this time. The hex analysis below is of the rendered 360 file from the GoPro Fusion through the fusion studio software. You can see the changes that occur between the original recording and recompressed recording. Figure 6 Hex Analysis of the Files Rendered out of the GoPro Fusion Studio S oftware. File is VIDEO_002.mov. Much less information is now stored in the header.

PAGE 31

22 Figure 6 Cont. Hex Analysis of the Files Rendered out of the GoPro Fusion Studio Software Below is an atom breakdown using AtomicParsley to demonstrate the loss of metadata between the captured raw video and the rendered 360 video . Table 7 Atoms of the Original Recording AtomicParsley comparison between the original files and stitched files Atom ftyp @ 0 of size: 20, ends @ 20 Atom mdat @ 20 of size: 238870404, ends @ 238870424 Atom moov @ 238870424 of size: 57166, ends @ 238927590 Total size: 238927590 bytes; 114 atoms total. AtomicParsley v0.8 Media data: 238870404 bytes; 57186 bytes all other a toms (0.024% atom overhead). Total free atoms: 162 bytes; 0.000% waste. Table 8 Atoms of the 360 Recompressed Video Atomic Parsley stitched file: Atom ftyp @ 0 of size: 20, ends @ 20 Atom mdat @ 20 of size: 1, ends @ 21 Atom dat @ 21 of size: 365, ends @ 386 Total size: 1871121856 bytes; 15 atoms total. AtomicParsley v0.8 Media data: 0 bytes; 2423845440 bytes all other atoms ( 129.540% atom overhead). Total free atoms: 0 bytes; 0.000% waste. In Summary, the important observation is the reduction in the amount of atoms between the original recording and recompressed the video . Even though the file size increases, the amount of data stored within the metadata of the file is much less.

PAGE 32

23 YouTube Up Down After rendering video out , the video was uploaded to YouTube using YouTube dl. After a wait of 2 3 hours to ensure that all formats were downloadable from YouTube, the file was downloaded using the F function in YouTube dl : Mp4 = merged 138,140 ( YouTube format selection on YouTube dl) MKV= YouTube dl automated choic e The Mp4 file is now at a frame rate of 29.97 and has a resolution of 5120x2560 The MKV file is now at a frame rate of 29.97 and has a resolution of 3840x2160 Figure 7 Hex Analysis of the Files Downloaded from YouTube in the MKV File Type . File is VIDEO_0002_yt QQco8L093WY.mkv. Information has now been changed again in the header.

PAGE 33

24 Figure 7 Cont. Hex Analysis of the F iles D ownloaded from YouTube in the MKV F ile T ype Figure 8 Hex Analysis of the Files Downloaded from YouTube in the MP4 File Type . File is VIDEO_0002_yt QQco8L093WY.mp4. We now have additional information stored with in the header but the information is still differs from the original.

PAGE 34

25 Figure 8 Cont. Hex Analysis of the Files Downloaded from YouTube in the MP4 F ile T ype Significate changes in the header have occurred with the recompression through YouTube. Even between the MKV file and the MP4 file the amount of data stored within the header varies due to the recompression of the files. New data has overwritten the original data. PRNU R esults was taken from the GoPro VR P layer and a PRNU analysis was run against it. Print screen HQ was used due to the inability to export a frame and this was the best option provided with the current features of the GoPro VR Player. Two cameras were tested against the GoPro after re si zing using P aint Shop P ro to match the abstract size of the print screen file (2020 x1371). Currently , this is the best practice for exporting a frame from an original recording without having to use the Fusion editing software. A comparison was made between a Sony Falcon and a C anon PowerShot . The red histogram is a comparison of the evidence PRNU against the suspect camera and database. The database contains videos from various other cameras PRNU. The blue histogram shows the suspect camera intra variability while red histogram shows the suspect and data base inter variability . Finally, the green histogram shows the position of evidence against the two , suspect and database. The results were as follows :

PAGE 35

26 Figure 9 GoPro Rendered File Comparison to GoPro. Figure 10 Go Pro Comparison to Sony Falcon.

PAGE 36

27 Figure 11 Go Pro Comparison to Canon PowerShot. Figure 12 Image U sed for F irst PRNU T GoPro VR Player software.

PAGE 37

28 Figure 13 A Screenshot from the 360 Point of View Taken from the GoPro Studio Software Before Rendering . Figure 14 A Screenshot from the Fish Eye Point of View Taken from VLC Upon Playback . Attempting to playback recording 1 in VLC results in a fish eye image. The image is warped and the color temperature is slightly different.

PAGE 38

29 Figure 15 Image C aptured from the 360Fly as the Fisheye P erspective.

PAGE 39

30 Figure 16 Image Captured from the 360Fly as the 360 Perspective , Camera N o t M oved.

PAGE 40

31 Audio R esults Using ffmpeg , the audio was extracted from the test recordings , rendering GPBK0102.mp4, GPBK0103.mp4, VIDEO_0102.mov, VIDEO_0102.mov. The stereo file was split to two mono channels to be read by M atlab script: FF mpeg i stereo.wav map_channel 0.0.0 left.wav map_channel 0.0.1 right.wav When the video was recorded in the classroom with a low level recording, the microphone picked up ENF as seen at 52 H z and 120hz in the spectral analysis. To verify this finding , a separate recording was made outside without around any lighting sources , and no ENF was found in the second spectral analysis. Figure 17 Waveform, Spectrogram and MDCT Map.

PAGE 41

32 Figure 18 LCF A nalysis . As indicated in the graph above, there is a roll off just below 50 hz . This was found in multiple recordings and no settings to adjust this.

PAGE 42

33 Figure 19 LTAS A nalysis . Figure 20 LTA Analysis T wo .

PAGE 43

34 C HAPTER IV D ISCUSSION T he fac t that exporting a file via the GoPro VR player as an uncompressed image is not possible , which pr e sents multiple issues because without opening the video in the player, the original image is warped and will not play back in a native player. A screenshot can be taken or print screen HQ . The result is a PNG file from the print screen HQ . The only way to render a video that has been stitched together out of the native player is to use the freeware Fusion Studio . As noted , most of the metadata is lost when using this editing software. After rendering the 360 file s , the file size increased while the meta data decreased. PRNU (Photo Response Non Uniformity ) , which describes the gain between power on a pixel versus the signal output. It i s well defined on the GoPro Fusion, and th ere is no support for the two selected files. T his is one measure of detecting manipulation when metadata and hex analysis fall short. It can be concluded from these results that the inability to play 360 videos at this time with native players while they are warped using conventional tools such as FF mpeg or even QuickT ime presents multiple problems for the forensic community. Having a stitched file results in the file being laundered and losing a lot of metadata. It is evident by the size of the atoms that the amount of information removed to get a 360 video in a playable format is quite significant . Without additional information provided from GoPro , geometry calculations are difficult. Information can be found in the file header s, but no information is provided from GoPro. Also notable is that data on both the back and front lenses are listed , with the meta data containing differing serial numbers be tween the two lenses.

PAGE 44

35 Conclusion The tested hypothesis was substantiated. A reduction in data from the recompression of the original recordings was found . Future research should be conducted to evaluate newer versions of the GoPro Fusion as well a s the GoPro Fusion studio software. Further testing should also be done on different videos resolutions , with other makes of 360 videos , and with other upload mediums , such as F acebook or Twitte r. A recording should be collected while the camera is connected to the power source in the wall or , if it is still captured , then via the mic. Additional research on whether ENF information is stored within the video or just in the audio is also recommended .

PAGE 45

36 REFERENCES [1] Chennamma, H., & Rangarajan, L. (2010). Image splicing detection using inherent lens radial distortion . International Journal of Computer Science , 7(6), 149 159. [2] Johnson, M. K., & Farid, H. (n.d.). Exposing Digital Forgeries Through Chromatic Aberrati on. [3] Conotter, Valentina & Boato, G & Farid, Hany. (2010). Detecting photo manipulation on signs and billboards. 1741 1744. 10.1109/ICIP.2010.5652906. [4] Whitecotton, C. M. (n.d.). YOUTUBE: RECOMPRESSION EFFECTS. Retrieved from http://www.ucdenver.ed u/ academics/colleges/CAM/Centers/ncmf/Documents/Theses/2017%20Fall/Whitecotton_Thesis_Fall2017.pdf [5] Goodison, S. et al (2015). Digital evidence and U.S. criminal justice system. The Rand Corporation. https :// www.rand.org/pubs/research_reports/RR890.html [6] Melton, L. (2016).Data transmission _ Parallel v. Serial . Retrieved from https://www.quantil.com/author/laurel/ [7] Robinson, S. (2012). The storage an d transfer challenges of big data. Sloan Management Review. Retrieved from https://sloanreview.mit.edu/article/the storage and transfer challenges of big data/ [8] Tsun Li C. (2014). FP7 IAPP project. Digital image and video forensics. Retri eved from https://warwick.ac. uk/fac/sci/dcs/people/chang tsun_li/ [9] Natarajan, M (2003). Multimedia and data technology: The challenges and delivery. Bulletin of Information Technology , 23 ( 4 ) , 19 26. [10] N IST on Hash Value, January 04, 2017 . Retrieved from 2018 https://csrc.nist.gov/projects/hash functions [11] Hannan, M. et al (2018). A method for verifying and authenticating digital media. Applied Computing and Informatics , 14 ( 2 ) , 145 158 . [1 4 ] SWDGE . Retrieved from https://www.swgde.org/documents [1 5 ] Daubert v. Merrell. Dow Pharmaceuticals 509 U.S. 579 (1993). [1 6 ] Frye v. United States, 54 App. D.C. 46, 293 F. 1013 (1923). [1 7 ] F ederal Rules of Evidence, 702 Testimony by expert witnesses. [1 8 ] Matney, L. (2017). Review: Go Pro Fusion 360. Retrieved from https://techcrunch.com/2017/12/19/review gopro fusion 360 camera/ [19] Best practices in digital video evidence. In video forensic expert. Accessed 4/30/19 at www. videoforensic expert.com/best practices for digital video evidence . [20] Scientific working group on digital evidence. Best practices for image authentication. Version: 1.0 (July 11, 2018). [21] Scientific working group on digital evidence. Minimum requirements for testing tools used in digital and multimedia forensics. Version 1.0 (November 20, 2018).

PAGE 46

37 [22] Bowers, A. (2018). Collecting and producing technically sound, legally defensible litigation data. Accessed 4/30/19 at www.doelegal.com/2018/08/c ollecting and producing litigation data/. [23] GoPro, INC. Fusion User Manual . Fusion User Manual , GoPro, 2017. 130 24092 000 REVC . [24] 360Fly., INC. 360Fly Manual . 360Fly Manual , 360Fly, 2016. [25] The Brand Amp. 360Fly Reviewers Guide . 360Fly Reviewers Guide , The Brand Amp, 2014.