COL867: Machine Learning For Networking
Special Topics Course, LH603, 2024
Course Description
Recent years have seen a growing trend of using machine learning (ML) across various domains, and computer networking is no exception. This course will provide an overview of the latest in using ML for networking. The first few weeks will focus on specific use cases of ML in networking such as in the domain of network security, application classification and performance prediction, and resource allocation. Later part of the course will delve into the research problems associated with building (task-agnostic) ML pipelines for computer networking. These include problems related to data collection and data representation as well as benchmarking, robustness, and explainability of ML models.
The objectives of the course are twofold: (1). provide an overview of key challenges and opportunities in integrating ML techniques into networking solutions; and (2). Gain hands-on experience by applying the concepts to real-world network datasets and understand the associated practical challenges.
Prerequisites
- Undergraduate Networking Course: COL334/672 or its equivalent
Grading
- Exams (35%)
- Assignments 30%
- Project (20%)
- Quizzes (15%)
Syllabus (Tentative)
We will first study ML solutions proposed for specific network tasks.
- Module 1: Traffic Classification
- Module 2: Resource Allocation
- Module 3: Application Performance Estimaton
- Module 4: Security
Next we will delve into the task-agnostic ML pipelines for networking.
- Module 5: Data Collection
- Module 6: Data Representation
- Module 7: Evaluation
- Module 8: Explainability
- Module 9: Synthetic Data Generation
Here is a list of papers that will be covered for each module.
Course Policy
- Honor Code: The course will follow the IIT Honor Code.
- Late Policy: Students are provided with a grace period of 96 hours for the entire semester. The grace period is counted in a granuality of hours. For instance, if you submit an assignment 1h30m past its deadline, 2 hours will be subtracted from the grace period. No late penalty is provided as long as you have a positive grace period balance. After the grace period expires, your credit for the assignment will be halved if it is submitted within a week after the deadline. No credit will be provided for later submissions. You still need to submit all the assignments to pass the course.
- Audit Policy: You need to score a B grade or above to get an audit pass. Also, you need to turn in assignments and projects for audit pass.
- Quiz: There will be surprise quizzes. Tentatively, 6 quizzes will be held and the best 5 will be counted towards final grading.
Resources
Background Readings and Videos
Datasets
- CIC Network Intrusion Detection Data (CICIDS)
- UC Irvine IDS/Botnet Dataset
- TLS Handshake Dataset
- NetML Challenge and Datasets (2020)
- MAWI Traffic Traces
Deep Learning
Visualization and Interpretation
Helpful Notebooks
Reading List
- Module 1: Traffic Classification
- Module 2: Resource Allocation
- Module 3: Application Performance Prediction
- Module 4: Security
- Module 5: Data Collection System
- Module 6: Data Representation
- Module 7: Robustness/Benchmarking
- Module 8: Explainability
- Trustee
- HotNETS??
- Module 9: Synthetic Data Generation
- Module 10: Miscellaneous