This paper describes a prototype cross platform based on intelligent switching of Virtual Private Network (VPN) communications by means of artificial intelligence algorithms able to identify and classify attack risks in self-learning mode by analysing the traffic logs of the system. The platform is also suitable for disaster recovery, data migration and ensures virtualization of communications between nodes in case of risk detection. In order to test the models and evaluate the accuracy of the AI algorithms for risk detection and classification, a number of cyberattack scenario have been simulated. The proposed platform
implements Cassandra Big Data system interfacing with supernodes enabling data migration, security and disaster recovery. By comparing the performance of different AI algorithms, the results show that a XGBoost-based algorithm is the most efficient and accurate method for cyberattacks prevention, showing a remarkable ability of classifying and identifying characteristic patterns of the most representative traffic log variables. The research work has been carried out within the framework of a research industry project.
ORCID of Submitting Authorhttps://orcid.org/0000-0003-1744-783X
Submitting Author's InstitutionDyrecta Lab
Submitting Author's CountryItaly