Visible to the public A Hybrid Packet Classification Algorithm Based on Hash Table and Geometric Space Partition

TitleA Hybrid Packet Classification Algorithm Based on Hash Table and Geometric Space Partition
Publication TypeConference Paper
Year of Publication2019
AuthorsHuang, Jiamin, Lu, Yueming, Guo, Kun
Conference Name2019 IEEE Fourth International Conference on Data Science in Cyberspace (DSC)
Keywordsaerospace communication, aerospace computing, Classification algorithms, classification performance, classification speed, Communication networks, complex network conditions, compositionality, computer network security, Decision trees, existing classification schemes, geometric space partition, hash, hash algorithms, hash matching, hash table, HGSP algorithm, high-speed packet classification requirement, huge packets, hybrid packet classification algorithm, integrated space-ground network, ISGN, Memory management, Network security, packet classification, Partitioning algorithms, pattern classification, Pattern matching, pubcrawl, Resiliency, security, Time complexity, tradition network, traditional Hicuts algorithm, tree data structures
AbstractThe emergence of integrated space-ground network (ISGN), with more complex network conditions compared with tradition network, requires packet classification to achieve high performance. Packet classification plays an important role in the field of network security. Although several existing classification schemes have been proposed recently to improve classification performance, the performance of these schemes is unable to meet the high-speed packet classification requirement in ISGN. To tackle this problem, a hybrid packet classification algorithm based on hash table and geometric space partition (HGSP) is proposed in this paper. HGSP falls into two sections: geometric space partition and hash matching. To improve the classification speed under the same accuracy, a parallel structure of hash table is designed to match the huge packets for classifying. The experimental results demonstrate that the matching time of HGSP algorithm is reduced by 40%-70% compared with traditional Hicuts algorithm. Particularly, with the growth of ruleset, the advantage of HGSP algorithm will become more obvious.
Citation Keyhuang_hybrid_2019