Exploration of Machine Learning Algorithms for Development of Intelligent Intrusion Detection Systems

For anomaly-based intrusion detection systems, software implementation of several string-matching algorithms using deep/machine learning approaches are frequently available. The research paper aims to explore the use of machine learning algorithms in developing intelligent intrusion detection systems for preventing unauthorized access to computer networks. A complete intrusion detection system consists of packet decoding, capturing, pre-processing, and pattern or string matching. However, string matching is computationally the most intensive part. Consequently, several architectures/designs have been proposed to accelerate the performance of string matching. This paper provides an architectural evaluation of the existing deep/machine learning-based solutions of string-matching algorithms. First, the most recent state-of-the-art solutions are carefully identified. Then, the identified solutions are compared in terms of the used dataset, number of classes, and accuracy. The architectural evaluation of multiple classification algorithms in this paper allows users of the domain to select an appropriate deep/machine learning algorithm for string matching according to their needs.