K-Salp Swarm Anomaly Detection (K-SAD): A novel clustering and threshold-based approach for cybersecurity applications

Abstract

Anomaly detection is a critical task in various domains, particularly in cybersecurity, where ensuring data integrity and security is paramount. In this study, we propose a novel approach to anomaly detection utilizing both the K-medoid and Salp Swarm Algorithms. Our methodology involves clustering the data using K-medoid and determining thresholds with an improved Salp Swarm Algorithm, enabling the identification of outliers within datasets. We conducted experiments on real-world datasets to evaluate the effectiveness of our approach. Significantly, proposed method surpassed alternative methods in performance across 5 of the 10 datasets, thereby showcasing its superior efficacy. For example, It demonstrated superior performance compared to alternative methods, achieving an AUC value of 0.8651 on the Thyroid dataset. Additionally, our approach yielded outcomes falling within the average spectrum across 3 datasets. These observations underscore the effectiveness of our proposed method in factifying anomaly detection methods and factifying cybersecurity protocols.