A Machine Learning Algorithm Based on Inverse Problems for Cyber Anomaly Detection
Ali Sever *
Department of Computer Information Systems, Pfeiffer University, Charlotte, NC, 28109, USA.
*Author to whom correspondence should be addressed.
Abstract
With the rapid rate of technological advance, digital communications have become an integral part of our lives in e-commerce, healthcare, education, and government. As the cyber world has expanded and become more complex, it has also generated severe threats to cyber security. Adversarial attacks such as anomalies and misuses are hard to detect with conventional methods as these cyber activities look very similar to genuine ones. There are many problems in anomaly and misuse detection of cybersecurity which can be considered as an inverse problem. In this paper, we have modeled anomaly detection system, Inverse Machine Learning Algorithm (IMLA), based on an inverse model approach with Riesz kernel and applying software system development concepts at each phase. For evaluation, the proposed approach IMLA have been compared with other state of the art supervised learning models. The experiments show the effectiveness of the proposed model IMLA.
Keywords: Anomaly detection, inverse problems, machine learning, cyber analytics, data mining