A Hybrid Privacy Preserving against Multi-attacks for Network Data Publication

Abstract

Network trace data provides valuable information which contributes to model network behavior, defends network attacks, and develops new network protocols. Therefore, releasing network trace data is highly demanded by researchers and organizations to promote development of network technologies. However, due to the sensitivity of the network track data, it is a potential risk for organizations to publish the original data which may expose their commercial confidentiality and customers’ privacy within their networks. Several methods have been proposed to prevent network track attacks, such as statistical fingerprinting and injection. Unfortunately, they are not sufficient to protect privacy because adversary can use more background knowledge to reach the intended attack, and this kind of attack is proved to be used. This paper proposed an attack model named Multi-Attacks by using more background knowledge. For this attack model, it extracts the inherent graphics structure between the source and destination IP addresses in the network trace data and proposes a solution, data swapping, to prevent the target host from being recognized, which is based on k-anonymity. Combined with other protection techniques, our method can effectively prevent this Multi-Attacks model while preserving the data utility and providing formal guarantees of confidentiality protection. And using data swapping method for privacy protection can provide a more perfect solution, reach a higher level of privacy protection and guarantee good data utility related to the anonymity-based approach. Lastly, our proposed algorithm is applied to different real datasets and demonstrate its effectiveness over several existing network trace data anonymization techniques.