Skip to main content


Exploitation as an inference problem


David Cock




In this position paper, we suggest that an adversary seeking to exploit a side channel should be viewed as an AI system performing inference under uncertainty. We demonstrate that this allows us to place a practical bound on the adversary's ability to exploit the channel. Specifically, by ensuring that the Bayesian network corresponding to the associated inference problem has large tree-width, the problem of inferring the secret given the observations is #P-hard.

BibTeX Entry

    publisher        = {ACM},
    doi              = {10.1145/2046684.2046702},
    author           = {Cock, David},
    month            = oct,
    slides           = {},
    year             = {2011},
    title            = {Exploitation as an Inference Problem},
    booktitle        = {4th Workshop on Artificial Intelligence and Security},
    pages            = {105--106},
    address          = {Chicago, IL, USA}


Served by Apache on Linux on seL4.