Application of Machine Learning Algorithms to KDD Intrusion Detection Dataset within Misuse Detection Context.

摘要: A small subset of machine learning algorithms, mostly inductive based, applied to the KDD 1999 Cup intrusion detection dataset resulted in dismal performance for user-to-root and remote-to-local attack categories as reported recent literature. The uncertainty explore if other algorithms can demonstrate better compared ones already employed constitutes motivation study herein. Specifically, exploration certain perform classes consequently, a multi-expert classifier design deliver desired measure is high interest. This paper evaluates comprehensive set pattern recognition on four found dataset. Results simulation implemented that effect indicated classification categories: specific algorithm specialized given category . Consequently, multi-classifier model, where associated with an which it most promising, was built. Empirical results obtained through indicate noticeable improvement achieved probing, denial service,