隐私保护下Shapley值算法研究

The Privacy Protection for Shapley Value

Shapley值法严格按照每个参与合作的参与者所做的边际贡献来分配收益.体现了某种程度上的公平性与合理性.但是该分配方案在确定每个参与者的利益过程中,需要共享各个参与者的成本和利润数据,极易引起参与者敏感信息的泄露.应用Paillier同态加密机制来对Shapley值法进行了深入研究,研究最终目的是确保在确定合作联盟收益分配方案的同时,每个参与者的隐私信息受到保护.

Most of Shapley Value is considered as prediction strictly on the payoff opportu- nities available to each coalition, conveyed by a single real number： gone are the actioris, the moves, and the individual payoffs. However, Privacy Protection for the calculating process of the distribution benefit of each coalition has also become inadequate and obsolete. We use the Paillier homomorphic encryption system to re-design the Shapley Value algorithm which is under the privacy protection, and to ensure that the sensitive information of participants will not be disclosed in determining the income distribution alliance programs at the same time.