DPI & DFI: A Malicious Behavior Detection Method Combining Deep Packet Inspection and Deep Flow Inspection

• Published in: Procedia engineering Vol. 174; pp. 1309 - 1314
• Main Authors: Guo, Yu-tong, Gao, Yang, Wang, Yan, Qin, Meng-yuan, Pu, Yu-jie, Wang, Zeng, Liu, Dan-dan, Chen, Xiang-jun, Gao, Tian-feng, Lv, Ting-ting, Fu, Zhong-chuan
• Format: Journal Article
• Language: English
• Published: Elsevier Ltd 2017
• Subjects: DFI; DOS; DPI; Probing; R2L; U2R; Probing; DOS; U2R; DPI; R2L; DFI
• ISSN: 1877-7058; 1877-7058
• DOI: 10.1016/j.proeng.2017.01.276

A malicious behavior detection approach which combines both the DPI (Deep Packet Inspection) and DFI (Deep Flow Inspection) is proposed, namely DPI & DFI. For the DPI & DFI method an outlier data mining method is employed. The fine-grained DPI is suitable for plaintext traffic, while DFI is a complementary for encrypted or emerging traffic. The collaborative detection approach includes three phases: DPI detection, DFI detection & comparison, and feedback. In present work, the C4.5 data-mining decision tree is adopted as classifier. The KDD Cup’99 benchmark is used and representative attack categories such as Probing, DOS, R2L (Remote to User) and U2R (User to Root) are evaluated. In-depth analysis demonstrates that the U2R and R2L attack categories lead to lower detection rate, and in particular the attack types contribute most are put forward. In future work, some other types of classifiers suitable to R2L and U2R attack categories should be investigated.