Adversarial attacks against supervised machine learning based network intrusion detection systems

• DOI: 10.1371/journal.pone.0275971
• Publication Date: 2022-10-14T17:42:18Z
• Authors: Ebtihaj Alshahrani; Daniyal Alghazzawi; Reem Alotaibi; Osama Rabie
• Keywords: Adversarial machine learning; Evasion (ethics); Attack model

Adversarial machine learning is a recent area of study that explores both adversarial attack strategy and detection systems attacks, which are inputs specially crafted to outwit the classification or disrupt training process systems. In this research, we performed two scenarios, used Generative Network (GAN) generate synthetic intrusion traffic test influence these attacks on accuracy learning-based Intrusion Detection Systems(IDSs). We conducted experiments including poisoning evasion different types models: Decision Tree Logistic Regression. The performance implemented scenarios was evaluated using CICIDS2017 dataset. Also, it based comparison IDS before after attacks. results show proposed reduced testing network models (NIDS). That illustrates our scenario negatively affected systems, whereas decision tree model more than logistic regression. Furthermore, disrupted NIDS, regression tree.