Anomaly Detection in Online Data Streams Using Deep Belief Neural Networks
Lecture Notes in Networks and Systems
Internet technologies are now utilized in almost every domain to gather data streams and also to monitor important events in an organization. However, these data streams are affected by abnormal or unusual pattern widely known as anomalies, which is responsible for malicious attacks, hardware failure, software failure, and reading errors. Hence, a dynamic, effective anomaly detection model is established in this research to enhance the quality of data gathered by the networks. The deep belief neural network is employed to obtain promising results in anomaly detection. The classifier performance is improved by reducing data dimensionality through the process known as feature extraction. The experimental analysis demonstrates the effectiveness of the proposed methodology by comparing it with existing techniques such as RNN, BNN, CNN, and Autoencoders. The experimental outcome such as accuracy, precision, F1-score, and error of about 80.5714%, 94.5440%, and95.3% for the HDFS dataset reveals the effectiveness of the DBN model.
Open Access Status
This publication is not available as open access