04-801-P4   Machine Learning for Computer Networks and Security (MLCNS)

Course description

The availability of large amounts of data and our increasing ability to mine such data to guide our decision-making processes have changed the way we do networking and security.

Heavy use of data has enabled us to discover properties/features of networks that would otherwise be difficult or even impossible to discover. Using these important discovered features, we can now better implement network functions (e.g., configuration, optimization, protection, prevention, healing, etc.) in an automated manner. For instance, thanks to the usage of data, Next Generation Networks (NGN) are expected to have a high degree of self-organization and will be composed of several sub-systems, each being a self-organized system by itself.

Unfortunately, the complex interactions of these multiple decision-making systems lead to emergent behaviors that are sometimes unintended and unanticipated. The use of data and machine learning algorithms to guide networking and security has also extended the attack surface: attacks can now target the data, the algorithm that mine the data, the decision-making process, etc., in addition to traditional targets such as the protocols, the applications, and the users.

In sum, this new data/ML for networking/security paradigm has led to two (dual) areas of interest in the community:

1. ML/Data Science for security and networking
2. The security of data-driven/machine-learned systems

Learning objectives

In this course, we will review the current trends of research in (1) and (2) and the techniques used to address these problems. The course will consider the theoretical aspects of these new paradigms and choose some targeted examples to apply/implement the learned techniques. The course will involve class presentations, reading materials, research, student presentations, and mini-projects (list not exhaustive).

Outcomes

At the end of the course, students will be familiarized with the challenges and opportunities pertaining to the applications of Data and ML to computer networking and security. For students pursuing a professional track, this course will help them be on top of new trends in networking and security. For students aspiring to pursue research, this course will offer the background into a promising Ph.D. research area.

Content details

Week 1

• Introduction
• Revision Networking & Security
• Data Collection (WITS, KDD, MAWI, IMPACT)

Week 2

• Revision Machine Learning

Week 3

• ML Applications in Networking (Traffic prediction, classification, & routing)
• ML Applications in Networking (Congestion Control & Network Management, QoE, QoS)

Week 4

• Next Generation Networks
• ICT Applications

Week 5

• ML Applications in Cybersecurity

Week 6

• The Security of Machine-Learned Systems

Week 7

• Wrapping
• Mini-Project Presentation

Faculty

Assane Gueye