The detection of new threats has become a need for secured communication to provide complete data confidentiality. The network requires anomaly detection to shield from hurtful activities. There are various types of metaheuristic methods used for anomaly detection. In this paper, a new approach is proposed for network anomaly detection using multi-start metaheuristic method and enhancement in clustering algorithms. The main stages involved in the proposed approach are: preprocessing, clustering, training dataset selection and the performance evaluation based on training and testing dataset to detect anomalies. The performance of two clustering algorithms, i.e. K-means and expectation maximization (EM) is compared using detection accuracy, false positive rate, and detector generation time. The experimental results are based on NSL-KDD dataset. The results show that the EM clustering performs better than K-means clustering algorithm.
Anomaly detection is an important element in the domain of security. As a result, we undertook a literature review on ML algorithms that identify abnormalities. In this paper, we are presenting a review of the 101 research articles describing ML techniques for anomaly detection published between 2015 - 2022.The goal of this paper is to review research papers that have used machine learning to develop anomaly detection algorithmThe forms of anomaly detection examined in this study include system log anomaly detection, network anomaly detection, cloud-based anomaly detection, and anomaly detection in the medical profession. After assessing the selected research articles, we present more than 10 applications of anomaly detection. Also, we have shared a range of datasets used in anomaly detection research, in addition to revealing 30+ new ML models employed in anomaly detection. We have discovered 55 new datasets for anomaly detection. We've noticed that the majority of researchers utilize real-life datasets and an unsupervised learning technique to detect anomalies. Many ML methods may be applied in this subject, so we present a summary of all work done in the previous six years. Keywords Intrusion detection, Artificial intelligence, Anomaly detection, security, Machine learning.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.