Research On Behavior Accountability Of Agents Responsible For Privacy Negotiation In Social Networks

The rapid development of social networks has promoted the widespread occurrence of information sharing.While information sharing brings convenience to users,it also creates serious privacy leakage problems.In addition to traditional privacy protection technologies,emerging artificial intelligence technologies/agents are used for privacy protection to achieve the purpose of improving efficiency and improving security.Among them,the agent-based privacy negotiation can assist social network users to establish a privacy protection consensus before information sharing and has a certain pre-protection effect on privacy leakage.Accountability is an attribute that can be held accountable by the responsible party for actions or consequences.It is an important aspect of transparent and interpretable artificial intelligence applications.The accountability of the privacy negotiation process in social networks is of great significance for improving the transparency and interpretability of application platforms or systems.Therefore,this paper analyzes the characteristics of the privacy negotiation system and researches the accountability of agent behavior in the privacy negotiation system.The following main tasks are carried out:(1)Based on the privacy negotiation system proposed by Kekuluoglu D et al.,a request for accountability for agent behavior is proposed.We designed and implemented a privacy negotiation agent behavior accountability system.With the help of qualitative and quantitative accountability schemes,the accountability of the behavior of the privacy negotiation agent is realized.The system not only protects the private information in social networks but also improves the security and credibility of the agent.(2)This paper proposes a qualitative accountability scheme for agent behavior in privacy negotiation.First,according to the characteristics of the privacy negotiation system,several key indicators of the initiator agent and the negotiator agent in the negotiation process were extracted and set as the behavior indicators for the accountability of each agent's behavior;Secondly,it formalizes the requirements for qualitative accountability,and elaborates the calculation method of each behavior indicator threshold of the initiator agent and the negotiator agent;Finally,a qualitative accountability method is proposed,which can accurately determine whether the privacy negotiation agent has improper behavior and lock its specific location when there is improper behavior,and describe the algorithm implementation process.The experimental results show that the experimental results of qualitative accountability are consistent with the improper behavior set before the experiment,and the qualitative accountability program is accurate and effective.(3)This paper proposes a quantitative accountability scheme for the behavior of the privacy negotiation agent.First,it formalizes the requirements for quantitative accountability;Secondly,for different data types of privacy negotiation data sets,three types of quantitative accountability methods are proposed: simple quantification,weighted Mahalanobis distance,and improved Minhash,which can quantify the severity of misconduct by the privacy negotiation agent;Finally,the definition,improvement methods and algorithm realization process of the three quantitative accountability methods are explained in detail.In the simulation experiment,we first use the experimental results of simple quantitative methods to verify the qualitative accountability results,then compare the experimental results of three quantitative accountability methods,and finally analyze the experimental results of the improved Minhash method and make a horizontal comparison.The experimental results show that the quantitative accountability scheme is feasible and effective to realize the action accountability of the agent,and adding weight to improve the quantitative method also has higher accuracy.