High performance packet classification algorithm for network security systems using modified grid-of-tries.

In order to enhance the functionality of network applications including traffic engineering and intrusion detection, the packet classification problem has received extensive research during the past decade. Software-based packet classification algorithms are gaining significant attention due to their extremely high flexibility in satisfying various industrial requirements for security and network systems, which has coincided with the general improvement of hardware architectures and the rising popularity of multi-core multi-threaded processors over the past few years. These methods require extremely big tables internally to achieve fast classification, and the size of the tables could possibly grow along with the size of the rule set. They cannot be utilised with a big rule set as a result. To solve this issue, we present a novel software-based packet classification algorithm dubbed the grid-of-tries approach that combines the partition decision trees in a search table to support both high scalability and quick classification performance. Our proposed algorithm exhibits a very high categorization speed, regardless of the quantity of rules, with smaller tables and shorter table building time, while the majority of generic partitioning-based packet classification algorithms demonstrate acceptable scalability at the expense of reduced classification speed. In this study, we suggest the Grid-of-tries approach, a novel packet classification algorithm to bridge the theoretical and practical gap. In terms of classification speed, memory utilisation, and preprocessing time, our technique outperforms other well-known algorithms. The results of our tests demonstrate how the proposed method facilitates network systems to handle significant traffic in the most efficient way. [ABSTRACT FROM AUTHOR]