Host-to-network authentication is a weak link in the chain of modern network security systems. There is no widely accepted method of validating a computers identity on a network. Currently, the best commercial methods entail using the hardware address of an Ethernet network interface controller, but those can easily be changed. Spoofing the hardware address can lead to unauthorized entry and possible data theft on sensitive networks. In this work a possible solution for this problem is proposed and evaluated.

The proposed solution is to use device specific analog characteristics of the IEEE 802.3 normal link pulse (NLP) as a single factor authentication feature. Two sub-problems, inter-model and intra-model, are addressed in order to establish the effectiveness of the NLP as an authentication feature. The performance of a preliminary intra-model classification algorithm using this feature is evaluated and cross-referenced to a survey of Ethernet cable length changes, temperature change, and computer operational effects.