Covert Sequential Hypothesis Testing

Authors

Image provided by Meng-Che Chang
Meng-Che
Chang
Georgia Institute of Technology
Profile
Matthieu
Bloch
Georgia Institute of Technology

Abstract

We consider the problem of covert sequential testing, in which a legitimate party attempts to run a sequential test while escaping detection from an adversary.Specifically, the legitimate party's decisions should meet prescribed risk constraints and, simultaneously, the adversary's observations induced by the test should remain indistinguishable from the observations obtained in the absence of a test. Our main result is the characterization of the risk exponentγθ, which captures the asymptotic exponential decrease of the risk with the square-root of the averaged stopping time in the limit of low risk.An example is provided to illustrate how the covertness constraint influences the design of the sequential test.

Paper Manuscript