A New Approach to Classification of Malware Detection System

Since the appearance of the first virus detection system based on data mining proposed by Schultz in 2001, several studies have shown the effectiveness of data mining techniques in the fight against computer viruses. These systems, based on data mining, exploit available data on the previous attacks to achieve a smarter detection method. The majority of these systems is based on supervised learning. This limits their adaptations in dynamic environments (computer viruses environment) because they are not progressive after the training phase. In addition, the training of a model requires a large number of programs labeled as learning base. In this paper, we propose a virus detection system based on an evolutionary process of data mining where we will be able to optimize the number of examples for training while reducing the cost of labeling. A considerable improvement is made in the computer virus detection process through our system based on a new active and incremental learning architecture.

Active learning, Computer virus detection, data mining, Incremental learning