To Share or Not to Share: A Threshold Analysis of AoI under Server Compromise

Abstract

We consider a scenario where a source is sending sensitive information, under a general aim of information freshness, e.g., for real-time applications. This objective is quantified through age of information (AoI), for which we focus on the core formulas of queueing theory with preemption. While representing an idealization, this approach is chosen for its analytical tractability, which allows for the derivation of closed-form expressions that make the underlying trade-offs transparent. However, we consider that the source can use two servers, even in parallel, one of which is slower but trusted, whereas the other is faster but potentially compromised and insecure. Our analysis determines how much of potentially sensitive information must be sent to the insecure server under a multi-objective combination of keeping the average AoI low, while at the same time avoiding to disclose too much information. Our analytical findings, made possible by the tractability of this model, reveal a surprising structure of the solution, where a threshold effect is basically at play. Whenever the privacy component of the objective is stronger than freshness, one should completely avoid using the insecure server. Conversely, if average AoI minimization is more important, the insecure server can be used almost indiscriminately to exploit its higher processing rate.