Experiment and improve reinforcement learning algorithms to enhance anomalous network behaviour detection
TeleMARS Pty Ltd
Network Intrusion Detection Systems (NIDS) represent one of the major techniques in preventing cyber-attacks that occur in network traffic. Over the past decade, a lot of research work has been conducted to explore the capabilities of artificial intelligence (AI) methods in developing NIDS solutions. Previous studies suggested that AI algorithms have promising potential in developing effective solutions to detect the increasing attacks.
The TeleMARS R&D team commits to advance AI-based methods, explore realistic approaches of deploying the research outcomes in real network environment, and support on-going research in wider community to achieve long term sustainable development.
The key objectives of this project are:
- Contribute to the development of NIDS;
- Contribute to research community in the subject of anomaly detection;
- Establish a practical collaboration framework to enable scientists and IT professionals from diverse backgrounds to work together to continuously contribute to NIDS research;
- Test and prove TeleMARS operations and technical frameworks, and the team capabilities; and
- Inspire and enable the participation of broader research community in cybersecurity domain supporting gender equality and inclusion.
This project commenced in September 2020 and was finalized in June 2021. The main activities included:
- Literature review and project design
- Data analysis and preparation
- Anomaly detection model development using Machine Learning methods including Reinforcement Learning methods
- Model experimentation
- Established evaluation pipelines to simulate real application environment
- Model capability evaluation applying different datasets
- Implementation of a collaboration framework supporting the research activities conducted by researchers and professionals with various backgrounds