Dynamic covert channels in finance

Jorge R Ramos, Purdue University

Abstract

We present a technique for hiding information in stochastic settings through data-synthesizing schemes based on transform-expand-sample (T ES) processes. The technique is applicable whenever data generated by an application or process is sufficiently complex to exhibit random but structured behavior (such as in collective data transforms), and data trajectories have viable alternatives that are unverifiable or simply hard to verify. In such cases, a synthesizing procedure generates novel data that either actually replaces, or is generated instead of, application or process data. We begin by presenting methods for synthesizing data based on clone-generators, along with experiments. Next, we present an information-theoretic model that predicts under which circumstances we can achieve perfect security when information is hidden in such data. When information can be hidden at levels higher than typical levels of noise, message-neutralizing attacks will fail. If synthetic, stego and application/process data cannot be distinguished, secure stego transmissions can be launched. Finally, we study the security of our technique by doing a number of steganalysis experiments with the help of support vector machines. As opposed to other steganographic techniques, our scheme has an implicit key which makes it resistant to simple detection algorithms.

Degree

Ph.D.

Advisors

Rego, Purdue University.

Subject Area

Computer science

COinS