Arbitrarily Varying Remote Sources

TL;DR

This paper investigates the rate distortion limits for lossy compression of a remotely generated, arbitrarily varying source observed through noisy channels, considering adversarial jamming and stochastic encoding strategies.

Contribution

It establishes bounds on the rate distortion function for AVRS with stochastic encoding under adversarial channel conditions, advancing understanding of secure source coding.

Findings

Derived upper bounds on the rate distortion function.

Derived lower bounds on the rate distortion function.

Analyzed the impact of adversarial jamming on coding strategies.

Abstract

We study a lossy source coding problem for an arbitrarily varying remote source (AVRS) which was proposed in a prior work. An AVRS transmits symbols, each generated in an independent and identically distributed manner, which are sought to be estimated at the decoder. These symbols are remotely generated, and the encoder and decoder observe noise corrupted versions received through a two-output noisy channel. This channel is an arbitrarily varying channel controlled by a jamming adversary. We assume that the adversary knows the coding scheme as well as the source data non-causally, and hence, can employ malicious jamming strategies correlated to them. Our interest lies in studying the rate distortion function for codes with a stochastic encoder, i.e, when the encoder can privately randomize while the decoder is deterministic. We provide upper and lower bounds on this rate distortion function.