Laurens D'hooge

Seemingly robust ML-based network intrusion detection systems crumble when attacked with realistic, host-space perturbations, like a single-character change in a command string.