Title:Achievable Rates of Attack Detection Strategies in Echo-Assisted Communication

Abstract:We consider an echo-assisted communication model wherein block-coded messages transmitted by a source reach the destination as multiple noisy copies. We address adversarial attacks on such models wherein a subset of the received copies at the destination are rendered unreliable by an adversary. Particularly, we study a non-persistent attack model with the adversary attacking 50% of the codewords at random. We show that this attack drives the destination to detect the attacked copies within every codeword before combining them to recover the information bits. Our main objective is to compute the achievable rates of practical attack-detection strategies as a function of their false-positive and miss-detection rates. However, due to intractability in obtaining closed-form expressions on mutual information, we present a new framework to approximate the achievable rates in terms of their false-positives under special conditions. We show that the approximate rates offered by our framework is lower bounded by that of conservative countermeasures, thereby giving rise to interesting questions on code-design criteria at the source. Finally, we showcase the approximate rates achieved by traditional as well as neural-network based attack-detection strategies, and study their applicability to detect attacks on block-coded messages of short block-lengths.