An Unknown Protocol Clustering Analysis Method Based on Spectral Clustering

To study unknown protocols in networks, the primary task should be to unravel the unsupervised agglomeration of unknown protocols; To the current finish, associate degree algorithmic rule for unknown protocols combined with spectral agglomeration is projected, borrowing the idea of the longest common subsequence; For unknown protocols with totally different payload characteristics like camouflage or customization, the longest common subsequence is employed to live the similarity of samples, combined with spectral agglomeration, the protocol agglomeration drawback is regenerate into a graph partition drawback. The experimental results show that the algorithmic rule encompasses a smart agglomeration effect; and also the optimum agglomeration sequence threshold for the corresponding dataset is tested throughout the experiment, showing that totally different protocols square measure well-distinguished inside the edge.