Application of Machine Learning Techniques for Software Anomaly Detection

A rising variety of platforms and software programs have leveraged repository-stored datasets and remote access in recent years. As a result, datasets are more vulnerable to malicious attacks. As a result, network security has grown in importance as a research topic. The usage of intrusion detection systems is a well-known strategy for safeguarding computer networks. This paper proposes an anomaly detection method that blends rule-based and machine-learning-based methods. In order to construct the appropriate rules, a genetic algorithm is utilized. Principal component analysis is used to extract the relevant features aimed to improve the performance. The suggested method is validated experimentally using the KDD Cup 1999 dataset, which meets the requirement of using appropriate data. The proposed method is applied to detect and analyze four types of attacks in a well-known benchmark dataset: Neptune, Ipsweep, Pod, and Teardrop, utilizing Support Vector Machine, Decision Tree, and Naive Bayes algorithms. After testing the characteristics specified in the training phase, the data is classified into attack categories and normal behavior during the machine learning phase.