Anomaly Detection via Improved Rain Optimization Algorithm and Stacked Autoencoder Hoeffding Tree

Abstract

Professionals of Cybersecurity suggest that cyber-attack cost of damage would widely increase. Huge web usage increases stress over the way of securely passing on electronic info. Diagnosing faults and anomalies in real-time industrial systems is a concern because of enough covering an industrial system’s complexity and difficulty. For security improvement, intrusion detection systems (IDSs) are applied for anomaly detection in network traffic. Now, IDS technology has concerns about performance based on false alarm notifications, unknown attack diagnoses, times, and accuracy detection. Machine learning (ML) methods have been increasingly applied in IDS for many years. However, such methods yet suffer from a lack of a labeled set of data, low accuracy, and heavy overhead. Reducing dimensionality plays an important role in IDS, as anomaly diagnosis from high dimensional network traffic attributes is a time-consuming process. Selection of Feature selection affects analysis speed. This paper concentrates on developing IDS effectiveness by applying the proposed Stacked Autoencoder Hoeffding Tree approach (SAE-HT) applying an Improved Rain Optimization Algorithm (IROA) for the selection of features. Experiments on the dataset NSL-KDD illustrate that our model multi-classification possesses great performance. In comparison to the other mechanisms of ML in the accuracy case, our model performs better than such mechanisms. The presented technique obtained an accuracy of 98.82 on a dataset of NSLKDD. Such outcomes of tests show that the presented strategy could properly and efficiently diagnose bad data.