Deep learning has been widely used in cyber security disciplines. Deep learning-based intrusion detection systems currently play an important role in network security protection. However, these intrusion detection systems may be misclassified when an attacker introduces expert-designed, unrecognizable perturbations to the input samples for attack purposes. Adversarial machine learning has received much attention in the field of image recognition.However, there hasn’t been a lot of research done on intrusion detection. In this paper, we propose a novel attack method for generating adversarial samples. The attack method aims at spoofing intrusion detection systems with minimal modifications.