Steganographic and Steganalytic Methods for Timing Channels with Side Information

University Of Illinois At Urbana-Champaign, Urbana IL

Investigators

Abstract

Steganographic and Steganalytic Methods for Timing Channels with Side Information This proposal addresses dual problems of steganography (design of coding techniques for covert communications) and steganalysis (design of statistical methods for detection of covert communications) applied to timing channels on asynchronous networks. Hidden messages are embedded by modulating the timings of a cover packet stream (e.g., SSH file transfer data; keystroke data for instant messaging or other interactive session) which plays the role of side information for the encoder. The encoder is also subject to causality and latency constraints. Hence the objective of this research is to construct and analyze novel methods for covert communications over timing channels with side information, as well as novel methods for detecting them. The project is organized around four research thrusts: 1) Develop mathematical formulations and research the information-theoretic limits of timing codes with side information. The problem formulation includes a statistical model for the input packet process (e.g., a renewal process, or a Markov process), steganographic constraints (to guarantee statistical undetectability of the embedded messages), causality constraints, and latency constraints. 2) Develop and analyze a family of practical codes that satisfy the above constraints. These codes are nonlinear and stochastic (they inject an optimized amount of ``noise'' via a queue) and we refer to them as queue-based steganographic codes. 3) Develop an efficient computational framework for optimizing information-theoretic functionals associated with queueing processes. 4) Construct a steganalysis learning machine (based on empirical data and various timing codes) which returns a binary decision about the presence of hidden data. Broad Impact and Intellectual Merit: This project is highly multidisciplinary and involves a synergy between research and educational activities in random processes, communications, coding, queueing theory, information theory, statistics, and learning theory. The project addresses fundamental questions with applications to information protection and cybersecurity. Other impact includes computational methods for queueing processs, steganographic methods for signal processing and other timing channels, etc. Many graduate students are interested in moving into such an attractive research area, which combines breadth and strong emphasis on fundamentals with practical relevance. We intend to train these students for leadership roles in information technology.