Privacy Preserving Association Rule Mining in Collaborative Intrusion Detection Systems with Fuzzy Data
Motahareh Dehghan Chachkamy, Babak Sadeghiyan · 2013
One area of research in information security is Intrusion Detection Systems (IDSs), which are installed on target systems and tracks the indication of attacks. Since, Intrusion Detection Systems for attack detection, need to information collection and analysis, concerns about the disclosure of individuals and systems and/or disclosure sensitive information of them exist. Therefore, despite the use of IDSs, we deal with a privacy violation. One of the issues of privacy in network intrusion detection systems (NIDSs) is that several organizations wish to collaborate together to prevent the penetration of their sites. To achieve this, they share normal and attack data of their IDSs. Since the data are sensitive, they don’t want to share explicit data. Now, how these organizations can operate data mining and/ or machine learning on aggregate data without violation on data confidentiality. Privacy concerns can prevent this approach - there may not be a central site with authority to see all the data. We present a privacy preserving algorithm to mine association rules from several organizations (IDSs). These organizations, partitioned horizontally. In this paper, we describe weighted Association Rule Mining from fuzzy and binary data, using secure sum method. This paper generally focuses on the association rule mining from KDD dataset and for instance, generates Neptune attack rules that will detect Neptune attack in network audit data using anomaly detection.