Adversarial Attacks on Machine Learning Models in Cyber- Physical Systems
Mahbub Rahman, Fei Hu · 2023
In this chapter, we investigate the vulnerability of several popular machine learning algorithms (like KNN, random forest, and SVMs) to a variety of attacks, namely evasion attack, membership attack, data poisoning attack. These attacks can occur during training and testing phase by manipulating the training data, labels, injecting foreign test sets, etc. These attacks can increase the misclassification rate and devastate the cyber-physical system (CPS).