Cyber Attack Detection Using Spatio Temporal Patterns

Abstract

Cyber-attacks are rapidly growing as the Internet evolves, Furthermore, the state of cyber security is risky. A brief definition of ML/DL methods is provided, as well as Deep Learning and Machine Learning approaches for intrusion detection analysis. Using temporal or thermal correlations, each of the phases was represented by papers that were indexed, read, and summarised. There are many commonly used datasets in ML and DL, and we'll discuss the challenges of using ML/DL for cyber security, as well as make some recommendations for future study. The KDD data set is divided into four categories in this paper: simple, information, host and traffic, with all data attributes being classified using the Modified Random Forest algorithm. As a result of this empirical study on the data set, the contribution of each of four categories of attributes on DR and FAR is seen, which can significantly improve the data set's suitability for achieving full DR with minimal FAR. The proposed model successfully achieved 91 percent accuracy of classification using only 12 selected features and 97 percent accuracy of classification using 36 features, while all 42 training features achieved 98 percent accuracy of classification.