An Efficient Intrusion Detection System Oriented to Industrial Control Network

Abstract

The present invention proposes an efficient intrusion detection system for industrial control networks, including a data packet capture system, a data packet analysis system and a detection system based on TensorFlow. The data packet capture system uses DPDK to capture data packets, and then the data packet analysis system will The captured data packets are analyzed layer by layer, and the extracted features are put into the shared memory; the detection system based on TensorFlow loads the offline training model, continuously takes out the features from the shared memory as input for detection, and finally sends the detection results to Web side.

Description

technical field [0001] The invention relates to an intrusion detection system, in particular to an efficient intrusion detection system oriented to an industrial control network. Background technique [0002] With the introduction of national strategies such as "Industry 4.0" in Germany and the "re-industrialization" trend in the United States, and the accelerated integration of new-generation information technologies and manufacturing technologies such as cloud computing, big data, artificial intelligence, and the Internet of Things, the industrial control system consists of From the original closed and independent to open, from stand-alone to interconnected, from automation to intelligence. While industrial enterprises have gained great momentum for development, a large number of security risks have also emerged. The "Stuxnet" virus attack on Iran's nuclear power plant, the continuous attack on Ukraine's power grid, and the large-scale power outage in Venezuela have sounde...

AI Technical Summary

Problems solved by technology

Compared with rule-based detection, anomaly-based detection methods are difficult to bypass, which improves the ability of intrusion detection to a certain extent. However, most of the current applications of machine learning in intrusion detection focus on algorithms, which lack convenient Load the detection engine for the machine learning model

Images