神经密码协议模型的研究_陈铁明.pdfVIP

ISSN 1000- 1239PCN 11- 1777PTP Journal of Computer Research and De elopment 46( 8) : 1316- 1324, 2009 1, 2 3 2 2 陈铁明 Samuel H . Huang 刘 多 蔡家楣 1 ( 100 19 1) 2 ( 310014) 3 ( 45221) ( tmchen@ zjut. edu. cn) Research on Neural Cryptographic Protocols 1, 2 3 2 2 Chen T ieming , Samuel H . Huang , Liu Duo , and Cai Jiamei 1 ( State K ey L aboratory of Sof tw are Develop ment E nv ironment , Beihang Univ ersity , Beij ing 10019 1) 2 ( Colleg e of Co mp uter S ien e and Te hnolog y , Zhej iang Univ ers ity of Te hnol ogy , H angz ho u 3100 14) 3 ( I ntellig ent Sy s tem L aboratory , University of Cin innati, Cin innati, OH 4522 1) Abstract Research on discr etized-neura-l netw or k-based cryptographic protocols is a no el topic in infor mation security domain. A t the beginning of this paper, literature on such neural cryptogr aphy is sur ey ed, and then an interacting neural netw ork model, named tree parity machine ( TPM ) , is intr oduced w ith a sur ey of its theoretical research. Then, se eral important issues regarding T PM. s applications for cryptogr aphy are addressed and analyzed through engineering empirical study. Tw o key problem s, namely, the stability of w eight synchronization and the security of synchronization, ar e in estigated in detail. T w o new concept methods, the threshold in neuron acti ation function and hash-based synchronization check, are proposed to impr o e the TPM-based cryptogr aphy application scheme. ANOVA and statistical experiments results show that using thr esho