PhD Seminar Course on
Multimedia Security and Forensics
Cagliari, May 17-24, 2011
|This activity was made possible by the "Visiting Professors 2010" program of the University of Cagliari, sponsored by the Autonomous Region of Sardinia|
|Instructor:||Chang-Tsun Li - University of Warwick, UK|
Lecture 1 (2 hours): Tuesday, May 17, 10-12
Lecture 2 (2 hours): Thursday, May 19, 10-12,
Lecture 3 (2 hours): Friday, May 20, 10-12
Lecture 4 (2 hours): Tuesday, May 24, 10-12
Topic 1: Robust digital watermarking|
After an introduction of different types of digital watermarking schemes and their applications in multimedia security and forensics at the beginning of this topic, I will present robust watermarking and its application in multimedia copyright protection, including owner identification, transaction tracking / fingerprinting and copy control.
Topic 2: Fragile and semi-fragile digital watermarking
This topic deals with the use of fragile watermarking in multimedia authentication and integrity verification. Fragile and semi-fragile schemes are expected to be sensitive to manipulation, i.e., if manipulated, the watermark embedded in the host media should be destroyed such that the manipulated media will not pass the authentication.
Topic 3. Properties and security of digital watermarking
This topic is deliberately postponed until example schemes have been introduced in the previous two topics. Properties, such as embedding effectiveness, imperceptibility, data payload, need to be considered when designing watermarking systems will presented. Security requirements of various watermarking schemes will also be covered in this topic.
Topic 4. Steganography and Steganalysis
Steganography is the technique of hiding data in the plain media while maintaining the fidelity of the stego-media (i.e., the media with hidden data) in order to serve the purpose of covert communication. On the other hand, steganalysis is the competing technique for detecting the presence of hidden data in multimedia.
Topic 5. Source device identification based on device signatures extracted from images
Topic 1 – 4 is about using extrinsic information (i.e., watermark or secret data artificially embedded in the host media) for providing multimedia protection and authentication. However, those techniques are not applicable to unwatermarked multimedia. Topic 5 to 8 cover multimedia forensic techniques that rely on intrinsic information in the content (i.e., the information that is part of the original content). Topic 5 deals with the use of minute information (device signatures) left by the imaging devices in the images for identifying the source devices. Various types of device signatures will be introduced.
Topic 6. Source device identification based on enhanced sensor pattern noise
Topic 6 is also about source device identification. However, it targets specifically the identification techniques that rely on Sensor Pattern Noise (SPN) left in the images by the semiconductor sensor of the imaging devices. SPN is the unique noisy signal due to the imperfection during the manufacturing process of semiconductor wafers. An enhancer of SPN recently developed by the speaker will be presented.
Topic 7. Unsupervised pattern classification
Topic 7 is intended to pave the way for the presentation of topic 8. A generic unsupervised pattern classifier using random fields will be introduced. This generic classifier requires neither a training phase nor the user’s specification of the number of classes / clusters and the characteristics of the features / patterns (e.g., centroid of each class). Therefore, it can be adapted for various applications easily.
Topic 8. Unsupervised image clustering based on enhanced sensor pattern noise
There are circumstances where a forensic investigator has a large set of images taken by an unknown number of unknown cameras and wants to cluster those images into a number of groups, each including the images taken by the same camera, in order to narrow down the investigation. A blind image clustering method based on the enhanced sensor pattern noise will be introduced in this topic. The core of this image clustering technique is the generic classifier introduced in Topic 7.
Dep. of Electrical and Electronic Engineering
University of Cagliari, Italy