|
The secure transmission of information from a source to a destination is typically handled via encryption algorithms. In many instances, data that may or may not undergo encryption prior to transmission can be manipulated to encode messages. With successful encodings, seemingly innocuous channels, e.g., documents, data streams, audio, video, can operate covertly for secret message transmission in various applications, e.g., collusion in finance, electronic information/auction markets, transaction sequences, advertising applets, simulation. Such transmissions violate the principles of data integrity, security, and privacy, and can be a threat to the functioning of organizations that utilize the applications. The goal of this project is to study, design and implement algorithms for the generation and detection of hidden messages in dynamic contexts. To this end the research focuses on a unified framework that analytically and experimentally examines various forms of message encodings in synthesized data, exploiting inherent structure and nondeterminism in data to confound detectability. The first part of the work focuses on generation methodology, based on statistically synthesized data. Message encodings that are theoretically hard to detect become candidates for the second part of the work which is detection, based on pattern recognition techniques, statistics and heuristics. Core components of the work are both theoretical and experimental, with a focus on the limits of detectability. The topics of study include novel synthesizing methods and applications, probabilistic encoding algorithms, blind detection algorithms and benchmark generation.
The project is strongly motivated by fundamental research questions, with broad impacts on application areas such as those mentioned above, all of which are of importance to computer, commerce, and homeland security. An integral part of the work is the development of a benchmark generator which can be used as a testbed for new detection algorithms. By demonstrating how covert channels can operate in diverse applications and disciplines the project will foment cross-disciplinary research in security. It will involve Ph.D research students, support the development of new material in the secure computing curriculum, in graduate/undergraduate courses and research seminars. Results of the work, in the form of publications, reports, algorithms, software and experimental data, will be made available at: http://www.cs.purdue.edu/research/PaCS/spots.html.
|
CT-ISG: Dynamic Covert Channels: Generation and Detection of Hidden Messages