Empirical research of the distribution function of synchronization time of neural networks in the key exchange protocol

Tehnologìčnij Audit ta Rezervi Virobnictva

Олена Романівна Малік¹ 

[1] National Technical University of Ukraine "Kyiv Polytechnic Institute, 37, Prospect Peremohy, Kyiv -56, Ukraine, 03056;

关键词: neural networks;    mutual learning;    key exchange protocol;   

DOI  :  10.15587/2312-8372.2014.26288

来源: DOAJ

【 摘 要 】

Neurocryptography is relatively new and little-studied area. Existing results show the prospects of this direction, but practical application requires a fairly complete analysis of existing protocols and systems. Analysis of the features of the key exchange protocol, which is built using the mutual learning of special-type neural networks - tree parity machines was conducted in the paper. Known existing protocol attack strategies were considered. A special part of the protocol is determining the synchronization moment of the neural networks of subscribers. The studies have revealed that the average number of the protocol iterations, required for synchronization significantly differs from the maximum value. To achieve the goal, the work of the protocol with a fairly large number of different neural networks was simulated. The synchronization time feature, found in the work shows the vulnerability of the studied protocol since it allows an intruder to carry out one of the known protocol attack strategies.

【 授权许可】

Unknown