Abstract: We propose an adversarial attack for machine-learning-based network intrusion detection systems that selectively alters only the most influential features. Unlike conventional attacks such ...