AI/ML in Cybersecurity of Critical Infrastructures
The performance of the supervised ML algorithms has been remarkable and useful for several areas. However, the biggest limitation of the supervised approaches is the availability of labeled data, that’s why unsupervised and self-supervised algorithms are gaining more popularity. In case of unsupervised ML, the algorithm learns the feature representation from the provided unlabeled data. This approach can be useful in cybersecurity, because unsupervised algorithms will focus on learning the normal communications of the network, and the small amount of anomalous data will be used for the algorithm testing and validating.
Goals:
- Develop an understanding of the differences between supervised and unsupervised ML
- Develop an understanding of the internet protocols
- Extracting cybersecurity features from network monitoring
- Develop an unsupervised ML for detecting network anomalies
- Create a visualization for displaying results.
Key Elements: Develop and validate an unsupervised machine learning algorithm for identification of abnormal network traffic. Extracting relevant features from network packets. Create a user interface for displaying results.
Skills: Machine Learning, Unsupervised learning, CyberSecurity, Tensorflow, Keras, python, Programming IDE
Projects:
- Open source and custom network security monitoring
- Feature extractor of cybersecurity features from network monitoring
- Developing a database for managing the data pipelines
- Developing of a machine learning anomaly detection system
- Create a dashboard for result visualization
Team Advisers: Milos Manic, Ph.D. (Professor, Computer Science), Elizabeth Baker, Ph.D. (Assoc Profession, Information Systems), Robert Dahlberg, Ph.D. (Assoc. Prof, Computer Science)
Students: Dr. Manic’s group (Harindra Sandun Mavikumbure, PhD student, Victor Cobilean, PhD student, anticipated: several CS undergrad and high school students), Prof. Baker’s group (anticipated several students)
Project Partner and Sponsor: US Army Cybersecurity Directorate, DOE, Idaho National Lab
Majors / Background: Highly motivated students from Computer Science, Electrical and Computer Engineering, Information Science, Homeland and National Security who want to learn more about cybersecurity. Students from the business and political disciplines are also welcome.
Contact: Prof. Milos Manic (mmanic@vcu.edu)