Towards a General Theory of Counterdeception
Abstract
Research in the fields of information hiding and digital forensics has introduced a new kind of problem, the deception problem, which is beyond the theoretical scope of mainstream cryptography. In this problem, one party attempts to detect malicious behavior, while the other party seeks to evade or fool a detection algorithm. This is at its core a classification or signal processing problem, except we are impeded not by noise, but by an intelligent adversary. These problems include many existing problems from virus detection to the detection of network attacks. The effort of an adversary to find malicious input that evades detection has yet to be quantified in the same way as brute-forcing an encryption primitive. This effort explored an adversarial version of detection and estimation theory, to uncover fundamental theoretical limits on an adversary s performance, or a detector s power in uncovering calculated deception, as well as limiting information leakage from its own outputs. Ultimately this effort seeks to answer the theoretical question of which party in a deception problem has the upper hand the detector or the deceiver subject to a set of initial conditions determined by the specific problem.
Document Details
- Document Type
- Technical Report
- Publication Date
- Feb 20, 2015
- Accession Number
- ADA620164
Entities
People
- Scott A. Craver
Organizations
- Binghamton University