This paper proposes an ISVM (Improved Support Vector Machine) based cybersecurity attack detector used to improve defense against industrial control network attack and guarantee a safe and stable operation of chemical power systems, which abstracts the cybersecurity attack surveillance into pattern classification problem and solves it with the ISVM. This method uses the ISVM to train and learn a huge mass of historical data in the industrial control network in order to secure the attack detection rule with which the real-time supervision will be enabled on the industrial control network. In the end, a replication experiment is cited with real data in power secondary network from a 110kv intelligent substation to reveal that the method features less supervision time-consuming and higher supervision precision.
