Аннотация
With the expansion of the Industrial Internet of Things (IIoT), the protection of Cyber-Physical System (CPS) is worth consideration. This research presents an integrated edge computing framework that aims at enhancing security by improving network traffic anomaly detection. It is done based on three advanced anomaly detection algorithms, namely Principal Component Analysis (PCA), isolation forest, and autoencoder. PCA is applied to the data to bring it in a compressed form where one is able to detect the outliers by having the most important features of the data extracted. The isolation forest algorithm, which was applied with the contamination factor of 0, is depicted below: 1, is used for the detection of the anomalies in the traffic, and, in fact, it was able to categorize several threats. However, the autoencoder model relies on reconstruction loss to decide if something is wrong with the data, with the 95th percentile as the standard for identifying activities that are odd. Edge computing increases the effectiveness of a framework for real-time anomaly detection and mitigation. The objective of this research is to evaluate PCA, isolation forest, and autoencoder methods and the outcome on the feasibility of the methods to enhance the security of IIoT systems.