Robustness Of Multi-Exit Architectures Against Adversarial Attacks

To remediate this, researchers have devised multi-exit architectures where deeper networks fix mistakes of shallow networks. In multi-exit architectures, the stack of processing layers are interleaved with early output layers or ‘exits’. However, the computational savings provided by this approach may not be robust against adversarial attacks. DeepSlothRecently, a group of researchers from the University of Maryland and Alexandru Ioan Cuza University studied the robustness of multi-exit architectures against adversarial attacks; and discovered DeepSloth. The standard defence against the adversarial samples was found to be inadequate against the slowdown.