Privacy-preserving approach for sharing and processing intrusion alert data

Amplified and disrupting cyber-attacks might lead to severe security incidents with drastic consequences such as large property damage, sensitive information breach, or even disruption of the national economy. While traditional intrusion detection and prevention system might successfully detect low or moderate levels of attack, the cooperation among different organizations is necessary to defend against multi-stage and large-scale cyber-attacks. Correlating intrusion alerts from a shared database of multiple sources provides security analysts with succinct and high-level patterns of cyber-attacks - a powerful tool to combat with sophisticate attacks. However, sharing intrusion alert data raises a significant privacy concern among data holders, since publishing this information means a risk of exposing other sensitive information such as intranet topology, network services, and the security infrastructure. This paper discusses possible cryptographic approaches to tackle this issue. Organizers can encrypt their intrusion alert data to protect data confidentiality and outsource them to a shared server to reduce the cost of storage and maintenance, while, at the same time, benefit from a larger source of information for alert correlation process. Two privacy preserving alert correlation techniques are proposed under semi-honest model. These methods are based on attribute similarity and prerequisite/consequence conditions of cyber-attacks.