In this paper,a model with two mutual learning neural networks named Tree Parity Machine(TPM) is firstly introduced,as well as its cryptographic property of weight synchronization with that of chaos cryptography is co...In this paper,a model with two mutual learning neural networks named Tree Parity Machine(TPM) is firstly introduced,as well as its cryptographic property of weight synchronization with that of chaos cryptography is comparatively discussed. A full empirical study on the stability and security of the TPM weight synchronization is conducted in detail. Then two improvement methods for the weight synchronization are proposed. Experiment results show that the improved TPM synchronization model can be efficiently against the third party attack. At last,a lightweight TPM-based key management scheme is proposed for TinySec on wireless sensor networks,which is full implemented on the Mica2 node and the performance test result is acceptable.展开更多
基金supported by the following funds:the Open Fund of the State Key Laboratory of Software Development Environment under Grant No.SKLSDE- 2009KF-2-01Beihang University, the National Basic Research Program of China (973 Program) under Grant No. 2005CB321901 and No.2010CB328106-3+1 种基金the Natural Science Foundation of China under Grant No.60773115the Open Fund of the Zhejiang Provincial Key Laboratory of Information Security
文摘In this paper,a model with two mutual learning neural networks named Tree Parity Machine(TPM) is firstly introduced,as well as its cryptographic property of weight synchronization with that of chaos cryptography is comparatively discussed. A full empirical study on the stability and security of the TPM weight synchronization is conducted in detail. Then two improvement methods for the weight synchronization are proposed. Experiment results show that the improved TPM synchronization model can be efficiently against the third party attack. At last,a lightweight TPM-based key management scheme is proposed for TinySec on wireless sensor networks,which is full implemented on the Mica2 node and the performance test result is acceptable.