Date of Original Version
Abstract or Table of Contents
Defending a server against Internet worms and defending a user’s email inbox against spam bear certain similarities. In both cases, a stream of samples arrives, and a classiﬁer must automatically determine whether each sample falls into a malicious target class (e.g., worm network trafﬁc, or spam email). A learner typically generates a classiﬁer automatically by analyzing two labeled training pools: one of innocuous samples, and one of samples that fall in the malicious target class. Learning techniques have previously found success in settings where the content of the labeled samples used in training is either random, or even constructed by a helpful teacher, who aims to speed learning of an accurate classiﬁer. In the case of learning classiﬁers for worms and spam, however, an adversary controls the content of the labeled samples to a great extent. In this paper, we describe practical attacks against learning, in which an adversary constructs labeled samples that, when used to train a learner, prevent or severely delay generation of an accurate classiﬁer. We show that even a delusive adversary, whose samples are all correctly labeled, can obstruct learning. We simulate and implement highly effective instances of these attacks against the Polygraph  automatic polymorphic worm signature generation algorithms.