ImpNet is more than a backdoor attack

In a recent paper written by our summer intern Tim Clifford, we present a system-level backdoor attack and a complete ML backdooring taxonomy.

More details are discussed in the paper “ImpNet: Imperceptible and blackbox-undetectable backdoors in compiled neural networks” by Tim Clifford, Ilia Shumailov, Yiren Zhao, Ross Anderson, and Robert Mullins

• Taxonomy Website

• arXiv

• Media coverage: Light Blue Touchpaper, Schneier on Security, Pluralistic