Trust Computing Method To Resist Active Attack

At present, there are more and more applications running based on trust computing which cover various fields such as network communication, network shopping and even network finance. Trust computing has become an important part of all kinds of network applications. However, there are also many kinds of attacks on trust computing, as a result of which may make users experience worse, keep users from reasonable service, and even threaten the user's property security, thus it is necessary to study the security problem of the trust model.This paper first summarizes the general method of trust computing, analyzes the security problems that exist in the current trust models and common resistance strategy, and then divides active attacks into two classes: one is the auxiliary attack represented by white-washing which can help the malicious nodes continue their malicious behavior. The other is the specific attack including single attack, collusion attack and strategic attack.According to the causes for white-washing, this paper proposes an optimized trust model for white-washing. In this model, we take factors such as the time in system into account, and divide nodes in system into safe nodes and malicious nodes which can cause white-washing by making strict regulation principles. Then we put forward the guarantee mechanism based on TOPSIS(Technique for Order Preference by Similarity to an Ideal Solution), in which the nodes in dubious group can only recommend resource with the guarantee of safe nodes to ensure the accuracy of recommendation provided by the dubious nodes. Experimental results show that the algorithm increased the white-washing nodes' cost to continue their malicious behavior which made the behavior of white-washing meaningless.For specific malicious attack, this paper introduces a k-means clustering algorithm based safe trust recommendation mechanism. The mechanism considers the departure degree of evaluation behavior and similarity in cluster index to resist simple malicious attack and collusion attack and improves the updating method which can ensure the dynamic characteristics of the trusted computing requirements and resist the strategic attack. During trust computing, this model introduces the leader node to help simplify calculation. Finally, the experimental results demonstrate this model reduce the complexity of computing and effectively resist the three kinds of specific attacks.