ABSTRACT
This chapter proposes a sparse relevance vector machine ensemble for adversarial learning. It presents the related work in adversarial learning. The chapter discusses the relevance vector machine model. It presents the gradient-based method for modeling adversarial attacks. The book presents experimental results on both artificial and real datasets. There are several theoretical conclusions regarding bounds on malicious noise rate and learning accuracy. Data mining tasks are made more complicated when adversaries attack by modifying malicious data to evade detection. The chapter presents a sparse relevance vector machine ensemble for adversarial learning. The algorithm sets individual kernel parameters to model adversarial attacks in the feature space by minimizing the log-likelihood of the positive instances in the training set.
