This paper introduces a new privacy filter adopted in the context of the DPT (Drone Protect Task) at MediaEval Benchmark 2015. Our proposed filter protects privacy by visually replacing sensitive RofI (Regions of Interest) by its shapes. A combination of steganography and scrambling is used in order to make this filter. Once the scrambling is applied on the pixels of the RofI, its MSB (Most Significant Bit) are hidden in the LSB (Least Significant Bit) of a cover image. Our filter fulfils four criteria defined by DPT: near-lossless reversibility, intelligibility, appropriateness and anonymization. We benchmarked the filter on the last three criteria and we get good results: 40 % for intelligibility and appropriateness, and 60 % for anonymization.
http://ceur-ws.org/Vol-1436/
http://www.multimediaeval.org
Graduate Outcomes Presentation Slides - English (v3).pptx
MediaEval 2015 - Privacy Protection Filter Using StegoScrambling in Video Surveillance - Poster
1. EURECOM – CAMPUS SOPHIATECH
450 route des Chappes F-06410 BIOT Sophia Antipolis
www.eurecom.fr
Efficient Privacy Protection in Video Surveillance by
StegoScrambling
Objectives:
• We proposed a new privacy filter, named StegoScrambling, for Video Surveillance
• Outperforms existing methods by fulfilling the four conflicting criteria
– Privacy, Intelligibility, Reversibility, Real Time
Concept:
• Combining:
– Cryptography to protect and recover original video only with a key
Key = Passphrase provided by user and then convert in number
– Scrambling to hide most of the information of regions of interest (RoIs) from an original video thank
to a XoR with numbers generated pseudo-random
𝑋𝑜𝑅𝐼𝑚𝑔 𝑖 = 𝑅𝑜𝐼 𝑖 𝑋𝑜𝑅 𝑅𝑎𝑛𝑑𝑜𝑚𝑁𝑢𝑚𝑏𝑒𝑟𝑠 𝑖 , ∀𝑖 𝑤𝑖𝑡ℎ 𝑖 𝑡ℎ𝑒 𝑏𝑖𝑡 𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛
– Steganography to hide the scrambling RoI in the Least Significant Bit (LSB) whereas the Most
Significant Bit (MSB) are dedicated to keep intelligibility (events of original video are recognized) with
a cover image (edges or shapes of the RoI).
𝑃𝑟𝑖𝑣𝑎𝑐𝑦𝐼𝑚𝑔 =
𝑖=0
5
𝑋𝑜𝑅𝐼𝑚𝑔 𝑖 ∗ 2𝑖
+
𝑖=6
7
𝐶𝑜𝑣𝑒𝑟𝐼𝑚𝑔 𝑖 ∗ 2𝑖
• Near lossless Reversible:
𝑅𝑒𝑐𝑜𝑣𝑒𝑟𝑒𝑑𝐼𝑚𝑔 =
𝑖=2
7
𝑃𝑟𝑖𝑣𝑎𝑐𝑦𝐼𝑚𝑔 𝑖 − 2 𝑋𝑜𝑅 𝑅𝑎𝑛𝑑𝑜𝑚𝑁𝑢𝑚𝑏𝑒𝑟𝑠 𝑖 ∗ 2𝑖
– Only the two LSB of the original image are lost which means at least pixel intensity of recovered
image decreases from 3 comparing to the original.
Natacha Ruchaud, Jean Luc Dugelay,
Multimedia Communication Department
Eurecom, Sophia Antipolis FRANCE
ruchaud@eurecom.fr
Passphrase
Seed
Random 6-Bit
RoI
6 MSB of the RoI Cover img in 2-Bit
6-Bit XOR img
XOR
2-Bit = 2 MSB
6-Bit =6 LSB
Privacy Img
Framework :
Existing Filter Privacy Intelligibility Reversibility Time
Pixelization/Blur Yes Yes No Fast
Masking/Warping Yes No No Fast
Morphing Yes No No Low
Scrambling Yes No Yes Low
Scrambling JPEG Yes No Yes Fast
StegoScrambling Yes Yes Yes Fast
Pixel example :
Original pixel b7 b6 b5 b4 b3 b2 b1 b0
X X
X X
1 1
0 0
Recovered pixel b7 b6 b5 b4 b3 b2 X X