Application of Association Rule Mining in Preventing Cyberattacks

Abstract Designing a security solution should rely on having a good knowledge of the protected assets and better develop active responses rather than focus on reactive ones. We argue and prove that malicious activities such as vulnerabilities exploitation and (D)DoS on Web applications can be detected during their respective initial phases. While they may seem distinct, both attack scenarios are observable through abnormal access patterns. Following on this remark, we first analyze Web access logs using association rule mining techniques and identify these malicious traces. This new description of the historical data is then correlated with Web site structure information and mapped over trie data structures. The resulted trie is then used for every new incoming request and we thus identify whether the access pattern is legitimate or not. The results we obtained using this proactive approach show that the potential attacker is denied the required information for orchestrating successful assaults.