Conformal Prediction for Offensive Security

Abstract

Despite its introduction more than a quarter century ago, Conformal Prediction (CP) has seen surprisingly few applications to the cyber security world thus far. In particular, we observe that, while CP has been employed as a defensive measure in many recent works, its use for carrying out attacks (i.e., for offensive security) is hard to trace in the literature. We explore this gap, by presenting initial findings in two key areas of offensive security: Privacy-Preserving Machine Learning, and network traffic analysis.

Sources

• arxiv.orghttps://arxiv.org/abs/2609.05165
• arxiv.orghttps://arxiv.org/pdf/2609.05165v1