Prerequisites: Familiarity with Probability, Machine Learning

Learning Objectives:

Machine Learning (ML) is increasingly used in sensitive and time-critical systems such as autonomous driving, cyber physical systems etc. to deliver higher performance and protect the confidentiality of the systems. Though ML based systems can be used to classify various malware attacks and develop intrusion detection systems, these systems are also susceptible to several adversarial attacks. This course covers a systematic approach on developing ML based cybersecurity methodologies. It will also cover adversarial attacks which intentionally forces ML systems to behave unexpectedly.

Learning Outcomes

  • Students will be able to develop ML models to classify malwares.
  • Able to implement simple intrusion detection systems using deep neural networks.
  • They will be able to demonstrate the vulnerabilities in ML systems and state methods to address adversarial attacks.

Course contents

1 Overview on Machine Learning with use cases from cybersecurity, classification of threats, attacks, vulnerabilities, malware, trojans etc. (6 lectures)

2 Classification of malware using supervised/unsupervised learning based on signatures and profiling. Decision Tree and context based malicious event detection (9 lectures)

3 Time Series Analysis and Ensemble modelling to detect deviation from normal behaviour, case studies in Reconnaissance detection (9 lectures)

4 Efficient Network Anomaly detection; familiarize with various stages of network attack and address using deep neural networks, develop intrusion detection systems (9 lectures)

5 Adversarial attacks on ML systems, model poisoning, black box attacks, white box attacks, state-of-art research paper reading on deep learning systems (9 lectures)

Text Books

  1. A. Hands-on Machine Learning for Cyber Security by Soma Halder, ISBN139781788992282

References

  1. Machine Learning and Security by David Freeman, Clarence Chio Publisher: O’Reilly Media, Inc. Release Date: February 2018 ISBN: 9781491979891
  2. Malware Data Science by Joshua Saxe with Hillary Sanders, ISBN-10: 1-59327-859-4 ISBN-13: 978-1-59327-859-5 Publisher: William Pollock

Metadata

Proposing Faculty: Dr.Vivek Chaturvedi Department: Computer Science and Engineering Programme: B.Tech Proposing date: Approved date: Proposal type: Offerings:

Past Offerings

  • Offered in Jan-May, 2022 by Vivek Chaturvedi
  • Offered in Jan-May, 2021 by Vivek