A METHOD FOR IMPLEMENTING PRIVACY-PRESERVING SECURITY SURVEILLLANCE BY APPLYING CRYPTOGRAPHIC TECHNIQUES ON A REAL-TIME EMBEDDED DSP FRAMEWORK

1
A METHOD FOR IMPLEMENTING PRIVACY-PRESERVING
SECURITY SURVEILLLANCE BY APPLYING CRYPTOGRAPHIC
TECHNIQUES ON A REAL-TIME EMBEDDED DSP FRAMEWORK

• Presented By-
• Ankur Chattopadhyay
• CS591 PROJECT SPRING 2007

2
Background Of The Problem

• Todays surveillance mechanism leads to privacy
  invasion
• Legal issues in restricted places such as
  restrooms, private households
• Over the years, instances of unmonitored crime on
  the rise.

Eyes Of A Camera Always Watching You Everywhere
3
Examples Of Existing Surveillance Cameras
Existing Monitoring
4
The Present Issue To Deal With

• Current surveillance systems displace crime,
  rather than stop it
• - Employment of CCTV moving crime out of the
  camera boundaries
• Areas under surveillance become crime-free while
  unmonitored zones become targets for illegal
  activity
• - Criminal acts committed in a private location,
  such as a locker room or restroom

5
Some Interesting Facts

• Instances of crime rising in schools
• Most students avoid school restrooms out of fear
• Almost 2000 students physically attacked each
  hour of the school day
• The Unfortunate Truth
• - Existing surveillance technology lowering
  privacy for the average person
• - Simultaneously, pushing crime further out of
  the view of the cameras

6
The Challenge Faced

• So, the question posed to surveillance system
  designers
• How to apply technology in the right way and at
  the right place to enhance security while
  protecting fundamental privacy rights of
  individuals?
• The answer to this question lies in the
  technology of PrivacyCam
• - Future model of surveillance camera with
  privacy protection

7
PrivacyCam
Blackfin DSP Module
Omnivision CMOS Camera Module
Integrated System PrivacyCam
8
PrivacyCam The Technology

• Uses privacy through cryptographic obscuration
  (PICO) technique
• PICO on a tiny Blackfin DSP processor chip,
  integrated with a small Omnivision CMOS camera
  module
• Application runs on a real-time operating system
  (uCLinux under Linux) within the chip
• Application performs the necessary tasks for
  privacy enhancement
• - detection of the region to protect
• - encryption of that region

9
System Level Design Of The Technology
Implementation
Capture image from the camera
Detect the region to protect (using face
detection, skin detection, motion detection or
other methods)
Use encryption key, generate session key and
store the secured key, along with the protected
region information, as embedded within the image
Carry out encryption on the region to protect,
and pass on the encrypted data to the image
compression process pipeline
10
Detection Of Region Of Interest

• Background Subtraction Model
• Two separate image frames, background model and a
  captured one, compared against each other
• Compute per coefficient (pixel wise) difference
  for each 8 x 8 DCT block
• - Compare the obtained difference with a model
  threshold value. If the majority of the
  coefficient differences are greater than the
  threshold, we encrypt (encode) that block,
  otherwise we dont encrypt that block (for my
  research I have used JPEG image compression)

11
Overview of Secret Key Cryptography

• To transmit data securely over an insecure
  medium, two parties agree on a key in which to
  encrypt data.
• This key is usually exchanged through public-key
  cryptographic methods
• User A encrypts a block of data X with key W and
  sends this data to user B.
• By using the same key W, user B decrypts the
  ciphertext Y back into X

Insecure Medium
X
Z
Y
Y
Z-1
X
A
B
W
W
Fundamental Concept Due to algorithm Z, its
nearly impossible to recover data X from
ciphertext Y without key W. Guessing the key W
through exhaustive search is generally infeasible.
12
Outline Diagram Of Applied Algorithm
13
The AES Algorithm

• The Advanced Encryption Standard (AES)
• powerful standard cipher, that operates by
  performing a set of steps for a number of
  iterations called rounds
• AES is a symmetric block cipher, and its better
  efficiency and effectiveness in handling data
  blocks (bytes) makes it an automatic choice over
  the vulnerable DES (Data Encryption Standard)
• For my research I used the Rijndael Block Cipher

14
AES Outline Diagram
15
Rijndael Algorithm

• Rijndael Block Cipher Algorithm
• Developed by Joan Daemen and Vincent Rijmen
  (pronounced Rhine-doll)
• An extremely fast, state-of-the-art, highly
  secure symmetric algorithm
• Allows only 128, 192, and 256-bit key sizes
• Variable block length supported
• A block is the smallest data size the algorithm
  will encrypt

16
Some sample images from our PrivacyCam
application while monitoring a private household.
Each frame with a changing object is followed by
an encrypted version. Here the human face region
has been protected for hiding individual
identity, thereby enhancing privacy.
17
Uniqueness Of Our Technology

• Protects privacy - hides individual identity
• Encrypts the image-region to protect with AES
  (Advanced Encryption Standard) using an
  encryption key
• Enhances security against any possible crime
  scenario
• Upon legal authorization, recovery of the full
  original image possible
• Recovery process through decryption by accessing
  the encryption key

18
Technology Features/Advantages

• Unlike other existing technologies
• - PrivacyCam allows general surveillance to
  continue, without disrupting the privacy of an
  individual
• Compared to existing commercial privacy-enhancing
  applications like Emmitall
• Provides better system stability and free from
  the typical vulnerabilities of software
  implementation
• Forms a network based ethereal webcam sensor

19
Analysis Of Technology

• Low cost of embedded system hardware makes
  technology affordable
• Smallness in size of system components makes
  technology space-optimized
• Involves balanced DSP processors with minimum CPU
  overhead and very fast peripheral interfaces
• System provides embedded real-time application
  with performance in the order of microseconds
  

20
Target Application Areas

• As a general-purpose security camera
• in public places
• In restricted areas like restrooms, locker-rooms
  to name a few
• As a special vigilance camera
• in bathrooms of school buildings

Locker Room
Rest Room
21
Application Areas

• As a surveillance camera for monitoring
• - old home centers for elder care
• - trouble-prone zones of school/university

Elder Care Center
Examples of incidents of violence and trespassing
at campus
22
Recommendations

• Need to test PrivacyCam in more realistic
  conditions like restrooms and locker rooms
• Need to show more results of real-time
  performance in testing conditions
• Potential research work in improving the
  mechanism of detecting the privacy region
• Potential research work in trying out other
  public-key block cipher techniques
• Need to build on the technology by extending to
  audio audio-video surveillance

23
Conclusion

• Novel technology addressing the critical issue of
  privacy invasion in an efficient and
  cost-effective way in optimized space
• Strikes fine balance between privacy protection
  and security enhancement
• Meets all the ideal requirements of todays
  surveillance
• Growing and significant market
• Our technology challenges existing
  privacy-enhancing applications and surveillance
  systems

24
References

• Ankur Chattopadhyay, T.E. Boult, Privacy Cam a
  Privacy Preserving Camera Using uCLinux on the
  Blackfin DSP IEEE CVPR Embedded Computer Vision
  Workshop, 2007
• T.E. Boult, PICO Privacy Through Invertible
  Cryptographic Obscuration - IEEE Computer Vision
  for Interactive and Intelligent Environments,
  2005
• Michael Hennerich, Linux on the Blackfin DSP
  Architecture - Embedded Systems Conference
  Silicon Valley 2006
• Marc Van Droogenbroeck, Partial Encryption of
  Images for Real-time Applications - Fourth IEEE
  Signal Processing Symposium, April 2004

25
References

• J.M. Rodriguez, W. Puech and A.G. Borsb, A
  Selective Encryption for Heterogeneous Color JPEG
  Images Based on VLC and AES Stream Cipher - Third
  European Conference on Color in Graphics, Imaging
  and Vision, June, 2006
• W. Puech, P. Meuel, J.C. Bajard and M. Chaumont,
  Face Protection by Fast Selective Encryption in a
  Video - IET,Crime Security Conference June, 2006
• Andrew Senior, Sharath Pankanti, Arun Hampapur,
  Lisa Brown, Ying-Li Tian, Ahmet Ekin, Blinkering
  Surveillance Enabling Video Privacy through
  Computer Vision - IEEE Security Privacy, 2005