Privacy Preservation Framework Based on Social Links and Attribute Correlation

Social networks are important in our daily lives since people share photos, videos, and all types of media through them. But, sensitive data or information shared on these social networks is exploited for suspicious purposes. Social network providers have implemented privacy-protection techniques to protect data privacy, such as anonymizing identities and restricting profile access to friends. However, attackers reveal or infer hidden user attributes using social links or user behaviors. This paper analyzes social features that affect user privacy and discovers that the correlation between social attributes is directly propositional to attribute inference. Based on this, this paper proposes an efficient attribute correlation inference scheme (ACISACN) that uses random walks on the path with a restart. The proposed method examines how attribute correlation affects performance and uses the Kulczynski method to quantify these relationships as weights of edges in an enhanced social behaviour attribute correlation networks(SACN). The proposed technique is compared with state-of-art approaches ReAI and TAI, and the results show better results by revealing a stronger attribute correlation. The Kulczynski correlation measure and ratio of Information Gain are used to assess the performance.