Early Cloud Computing Intrusion Detection Using Time Series Data: A Multivariate Neural Model and Improved Zebra Optimization Algorithm Approach

Abstract

Utilizing the cloud is not growing as fast as it might because of security and privacy issues. False positives are still a problem with network intrusion detection systems (NIDS), even with their widespread usage. Moreover, the intrusion detection problem has not been treated as a time series problem, necessitating time series modelling, in many research. In this paper, we use time series data to suggest a unique method for early cloud computing intrusion detection. Our strategy uses a forecasting model built using the Multivariate Neural Model of the prophet and an Improved Zebra Optimization Algorithm (IZOA) to gauge its effectiveness. The problem of making false linkages between time series anomalies and assaults is particularly addressed by this method. Our findings show a notable decrease in the quantity of forecasters used within our forecast model—from 70 to 10—while demonstrating an improvement in performance metrics like median absolute percentage error (MDAPE), dynamic temporal warping (DTW), mean absolute percentage error (MAPE), mean square error (MSE), root mean square error (RMSE), and mean absolute error (MAE). In addition, our method has shown reductions in cross-validation, forecasting, as well as training durations of around 97%, 15%, and 85%, respectively.