PhD Qualifying Exam – Tarfa Hamed
The School of Computer Science is pleased to announce that Tarfa Hamed will be presenting the PhD Qualifying Examination Research Proposal "Recursive Feature Addition is a Superior Feature Selection Technique for NIDS" on December 9, 2014, at 10:00 in Reynolds 219.
Recursive Feature Addition is a Superior Feature Selection Technique for NIDS
This research proposal focuses on designing a new feature selection method to improve the classification performance of Network Intrusion Detection Systems (NIDSs) and other problems. NIDSs have been suffering from limitations regarding their accuracy in detecting the incoming network intrusions and providing an effective protection to their users. These limitations usually come from the inaccurate selection of features used in the detection process, and that is usually reflected to the NIDS performance. In this proposal, a new machine learning-based feature selection method is designed and implemented to select the best subset of features for NIDS. The new method is considered as an embedded feature selection method and uses Support Vector Machine (SVM) as a core classifier. The proposed method ranks the input features in forward-selection manner according to certain ranking coefficient. Measuring the performance of the model will be based on several criteria: Accuracy, Detection Rate, False Alarm Rate, time to rank the whole features, and time to reach the optimal accuracy. To show that the new method is suitable for other problems (not just NIDS), several benchmark datasets from different areas have been chosen to apply the proposed method on. In addition, the method is compared against two wrapper methods, two filter methods and one embedded method to show its performance.
The proposal will be applied on the KDD cup99 dataset for NIDS which is the most popular dataset in NIDS studies and has been used as a benchmark by authors who work in this field.
Advisors: Stefan Kremer, Rozita Dara