Intrusion Detection for Industrial Control Systems Based on Open Set Artificial Neural Network

The security of industrial control systems (ICSs) has received a lot attention in recent years. ICSs were once closed networks. But with the development IT technologies, have become connected to Internet, increasing potential cyberattacks. Because are so tightly linked human lives, any harm them could disastrous implications. As technique providing protection, many intrusion detection system (IDS) studies been conducted. However, because complicated network environment and rising means attack, it is difficult cover all attack classes, most existing classification techniques hard deploy real since they cannot deal open set problem. We propose novel artificial neural based-methodology solve this Our suggested method can classify known classes while also detecting unknown classes. conduct research from two points view. On one hand, we use openmax layer instead traditional softmax layer. Openmax overcomes limitations softmax, allowing networks detect During training, on other new loss function termed center implemented improve ability. model learns better feature representations combined supervision loss. evaluate NF-BoT-IoT-v2 Gas Pipeline datasets. experiments show our proposed comparable state-of-the-art algorithm terms overall performance.