The effectiveness of classification algorithms on IPv6 IID construction

Document Type

Conference Proceeding




School of Computer and Security Science, ECU Security Research Institute


Originally published as:

Carpene, C., Johnstone, M. N., & Woodward, A. J. (2017). The effectiveness of classification algorithms on IPv6 IID construction. International Journal of Autonomous and Adaptive Communications Systems, 10(1), 15-22.

Original paper available here


This study assessed the effectiveness of classifying IPv6 interface identifier (IID) address construction using machine learning algorithms. It was observed that IID construction can be reliably determined through the usage of assisted machine learning algorithms such as the naïve Bayesian classifiers (NBC) or artificial neural networks (ANNs). It was also observed that the NBC classification, whilst more efficient, was less accurate than the use of ANN for classifying interface identifiers. Training times for an unoptimised ANN were seen to be far greater than NBC, which may be a considerable limitation to its effectiveness in real world applications (such as log or traffic analysis). Future research will continue to improve the classification training times for ANN situations, potentially involving general-purpose computing on graphics processing units (GPGPU) systems, as well as applying the techniques to real world applications such as IPv6 IDS sensors, honeypots or honeynets.