NICOP - An adaptable hybrid knowledge base system for security intrusion detection by phishingbasedvector attacks incorporating multiple classification and prudence techniques.
Abstract
We have previously developed a successful knowledge base system using Induct RDR that classifies phishing-related intrusion attempts by examining attribute-based data associated with target (suspect and non-suspect) web sites. The system s purpose is to filter out phishing – it classifies data into two categories: phishing-related or not-phishing-related. The system is trained on cases associated with known phishing sites by using contextual attributevalue pairs related to network-based components. This training data is learnt and a model is produced using Induct Ripple Down Rules (Induct RDR), which is a knowledge acquisition (KA) technique used to populate and maintain a knowledge base. In this project, we will focus on two tasks: implementing the prudence ability for the current single RDR algorithm and developing this algorithm to provide multiple classification conclusions. The objective of the first task is giving the system the ability to detect anomalous data and suggest possible importance to expert. The current Induct RDR system will be modified in order to include outlier detection and self-examination functionalities, which flag all outlier cases and rule patterns in supervised test data. The aim of second task is giving the system the ability to provide multiple classification conclusions. Note that the most current AI learning algorithms are designed for a single classification. In order to achieve this aim, the multiple classification dataset will be constructed, and the induct RDR will be modified to provide multiple classification. The modified functionalities will be evaluated by using 10-fold cross validation, and simulated expert. The proposed research is to develop learning algorithms for smart service systems in a security domain. The cyber security is a major concern of IT systems in any organizations, including US armies and governments. The proposed algorithms will enhance the sustainability of knowledge base in smart systems and will be a solution for dynamically changing environment. Asian Office of Aerospace Research and Development in 2015-2016 originally supported this project. The discussion group includes Dr. Hiroshi Motoda, Brian J. Lutz, Lt Col, International Program Officer. In addition, the project is discussed with Dr. KO, Hiekeun (Higgin), Associate Director, Office of Naval Research Global, Tokyo. We are going to publish conference papers and journal papers as outcomes of this project. Also we are going to make software for testing the proposed algorithms and will deliver these as outcomes of this project.
Document Details
- Document Type
- DoD Grant Award
- Publication Date
- Sep 26, 2018
- Source ID
- N629091612219
Entities
People
- Byeong Ho Kang
Organizations
- Office of Naval Research
- United States Navy
- University of Tasmania