Skip to Main Content
Netadhict , an effective tool for network traffic analysis, can classify current traffic into hierarchical clusters of a semantic kind using a revolutionary technique based on the similarity of their contents - (p, n)-grams with respect to n-bytes substrings at a p byte offset. The underlying algorithm, ADHICT, continues to refine its classified sets as long as certain sets keep receiving more than their "right share" of packets. Consequently, ADHICT has enormous potential to help us establish a best-effort bandwidth allocation basis and control congestion effectively. As a result, based on ADHICT, we propose to use merge and freeze operations in order to separate traffic into “equivalence classes” and balance them dynamically. To deal with congestion situations, we set up a maximum threshold: when packets dropped exceed the threshold, the same class set is limited by a firewall. We then allocate shares of bandwidth to each of these sets through an adaptive traffic shaping technique. Finally, to evaluate the effectiveness of the proposed mechanism, we perform two types of simulations, then classify each with our algorithm. The first simulation is conducted for normal traffic in a local network connected to the Internet, while the second simulation is done for abnormal traffic in which a large quantity of self-propagated worm packets cause congestion.