Contribution of Non-Scrambled Chroma Information in
Privacy-Protected Face Images to Privacy Leakage
Hosik Sohn1, Dohyoung Lee2, Wesley De Neve1, Konstantinos N. Plataniotis2, and Yong Man Ro1
1Korea
Advanced Institute of Science and Technology (KAIST),
Image and Video Systems Lab.
2University of Toronto, Multimedia Lab
October 2011.
10th International Workshop on Digital-forensics and Watermarking (IWDW 2011)
-2-
Contents
1. Introduction
2. Layered Scrambling for Motion JPEG XR
3. Assessment of Chroma-induced Privacy Leakage
3.1 Objective Assessments
3.2 Subjective Assessments
4. Discussion and Conclusion
-3-
INTRODUCTION
-4-
1. Introduction
 Present-day video surveillance systems often come with high-speed network
connections, plenty of storage capacity, and high processing power
 The increasing ability of video surveillance systems to identify people has
recently raised several privacy concerns
 To mitigate these privacy concerns, scrambling can be leveraged to conceal
the identity of face images in video content originating from surveillance
cameras
Privacy protected surveillance videos
-5-
1. Introduction
 The past few years have witnessed the development of a wide range of
content-based tools for protecting privacy in video surveillance systems
 Dependent on the location where scrambling (or encryption) is applied, three
different approaches of scrambling can be distinguished
 Uncompressed domain scrambling
 Transform domain scrambling
 Compressed bit stream domain scrambling
 One of the main challenge is to concealment of privacy-sensitive regions by
making use of invertible transformation of visual information at a low
computational cost
-6-
1. Introduction
 Content-based tools for privacy protection need to find a proper balance
between the level of security offered and the amount of bit rate overhead
 In general, altering the visual information present in privacy-sensitive regions
typically breaks the effectiveness of coding tools
Security
level
Coding
efficiency
 To limit bit rate overhead, many content-based tools for privacy protection
only scramble luma information, leaving chroma information unprotected
-7-
1. Introduction
 In this paper, we investigate the contribution of non-scrambled chroma
information to privacy leakage
 To that end, we study and quantify the influence of the presence of nonscrambled chroma information on the effectiveness of automatic and human
FR
 Objective assessment: we apply automatic FR techniques to face images have
been privacy-protected in the luma domain
 Subjective assessment: we investigate whether agreement exists between the
judgments of 32 human observers and the output of automatic FR
-8-
1. Introduction
 FR vs. Perception-based security metrics for assessing privacy level
 Luminance Similarity Score (LSS), Edge Similarity Score (ESS), and Local
Feature-based Visual Security Metric (LFVSM)[1,2]
 However, these metrics are general in nature and are thus not able to take
advantage of domain-specific information (e.g., face information)
[1] Tong, L., Dai, F., Zhang, Y., Li, J. “Visual security evaluation for video encryption,” in: Proceedings of ACM International Conference on
Multimedia, 835–838 (2010)
[2] Mao Y., Wu M., "A joint signal processing and cryptographic approach to multimedia encryption," IEEE Transactions on Image Processing,
15(7), (2006), 2061-2075.
-9-
LAYERED SCRAMBLING FOR
MOTION JPEG XR
-10-
2. Layered scrambling for Motion JPEG XR
 The video surveillance system studied makes use of Motion JPEG XR to
encode surveillance video content
 Motion JPEG XR offers a low-complexity solution for the intra coding of high-resolution
video content, while at the same time offering quality and scalability provisions
 Layered scrambling for JPEG XR [3]
 Modified JPEG-XR encoder
Secret key
DC subband
LBT
LBT
Q
Scrambling
(RLS)
Pred.
LP subband
Q
Pred.
Adaptive
scan
Scrambling
(RP)
HP subband/Flexbits
Q
Pred.
Adaptive
scan
Scrambling
(RSI)
• Adaptive
entropy
coding
• Fixed
length
coding
[3] Sohn, H., De Neve, W., Ro, Y.M., “Privacy Protection in Video Surveillance Systems: Analysis of Subband-Adaptive Scrambling in JPEG
XR,” IEEE Transactions on Circuits and Systems for Video Technology, 21, 170–177 (2011)
-11-
2. Layered scrambling for Motion JPEG XR
 Overview of layered scrambling technique
- Random level shift (RLS) for DC subbands
DCcoeffe  DCcoeff  R(L),
- Random permutation (RP) for LP subbands
LPcoeff i e  LPcoeff j , where i  1,..., C , j  x1 ,..., xC ,
- Random sign inversion (RSI) for HP subbands
 HPcoeff, if r  1
HPcoeff e  
,
 HPcoeff, otherwise
 N denotes the number of MBs, L denotes the RLS parameter, K denotes the number
of non-zero LP coefficients in a MB, and M denotes the number of non-zero HP
coefficients in a MB.
-12-
ASSESSMENT OF CHROMA-INDUCED
PRIVACY LEAKAGE
-13-
3.1. Objective assessments
 Experimental setup
 FR technique used: PCA, FLDA, LBP
 Face images: 3070 frontal face images of 68 subjects from CMU PIE
(68 gallery, 340 training, and 2662 probe face images)
 Probe face images represent privacy-protected face images that appear in video
content originating from surveillance cameras.
 Performance evaluation: Cumulative Match Characteristic (CMC) curve
 Notations
Notation
Explanation
DC, LP, and HP
DC, LP, and HP subband
S3
DC+LP+HP
S2
DC+LP
S1
DC
Subscripts (Y, Co, Cg)
Luma and chroma channels (Y, Co, and Cg)
Prime (′)
Scrambled image data
-14-
3.1. Objective assessments
 Influence of distance measurement on FR effectiveness
 Distance metric: Euclidean, Mahalanobis, Cosine, and Chi-square distance
DE : Euclidean distance
DM : Mahalanobis distance
DC : Cosine distance
DH : Chi-square distance
 In the remainder of our experiments, we make use of the Euclidean distance metric for
PCA- and FLDA-based FR, and the Chi-square distance metric for LBP-based FR
-15-
3.1. Objective assessments
 Scrambled luma information
 Assumes that an adversary is not able to take advantage of the possible presence
of non-scrambled chroma information in the privacy-protected probe face
images
-16-
3.1. Objective assessments
 Scrambled luma and non-scrambled chroma information
 We investigate whether layered scrambling is still effective when the
scrambled luma channel and the non-scrambled chroma channels are
simultaneously used for the purpose of automatic FR
 Assuming that an adversary has access to the compressed bit stream
structure, and thus to the non-scrambled chroma information
 To take advantage of non-scrambled chroma information, we adopted
feature-level fusion
-17-
3.1. Objective assessments
 Scrambled luma and non-scrambled chroma information
-18-
3.1. Objective assessments
 Non-scrambled chroma information
-19-
3.2. Subjective assessments
 Experimental setup
 Number of observer: 32
 We presented three scrambled probe face images of different subjects to the
human observers for each experimental condition
 Assessment method
 Human observers were asked to select the gallery face image that is most similar to
the given probe face image
 human observers were also able to study the probe face images at different zoom
levels
Gallery face images use for subjective assessment
-20-
3.2. Subjective assessments
 Non-scrambled chroma information
-21-
3.2. Subjective assessments
 Scrambled luma and non-scrambled chroma information
-22-
DISCUSSION & CONCLUSION
-23-
4. Discussion
 For video surveillance applications requiring a high level of privacy protection, both
the luma and the chroma channels need to be scrambled at the cost of a higher bit
rate overhead
 Layered scrambling to both the luma (Y) and the chroma channels (Co and Cg)
-24-
4. Discussion
 Bit rate overhead
 Security (ideal case)
 Sub-sampling decreases the level of privacy protection, given the lesser amount
of data available for scrambling
 Total number of combinations required to break the protection of 10 MBs
reduces from 3.6×10722 (4:4:4) to 1.7×10360 (4:2:0)
-25-
5. Conclusions and future work
 This paper studied and quantified the influence of non-scrambled chroma information on the effectiveness of automatic and human FR
 Our results show that, when an adversary has access to the coded bit stream
structure, the presence of non-scrambled chroma information may significantly
contribute to privacy leakage
 For video surveillance applications requiring a high level of privacy protection, our
results indicate that both luma and chroma information needs to be scrambled at the
cost of an increase in bit rate overhead
 In order to compile a benchmark for privacy protection tools, future research will
focus on identifying additional worst case scenarios
-26-
Thank you for your attention!
-27-
APPENDIX
 Effectiveness of general-purpose visual security metrics
 Visual security metric used
 Luminance Similarity Score (LSS), Edge Similarity Score (ESS), and Local Featurebased Visual Security Metric (LFVSM)
 Lower the values computed by the visual security metrics, the higher the visual
security