Study of Evasion Attack using Feature Selection in Adversarial Environment

Not only Pattern recognition but also machine learning techniques have been increased in adversarial settings such as intrusion, spam, and malware detection, although its security against well-crafted attacks that aims to evade detection. Spam filtering is one of the most common application examples considered in adversarial machine learning. In this task, the goal is often to design feature selection against attacks. Here we use Random Forest Classifier to find evasion attacks. The ability of rapidly evolve to changing and complex situations has helped it become a fundamental tool for computer security. Evasion attacks assumes that the attacker can arbitrarily change every feature, but they constrain the degree of manipulation, e. g. , limiting the number of medications, or their total cost. Adversarial Feature Selection design phase are given in this paper