AI for Security

Course Objective

This course explores a range of topics at the intersection of security and artificial intelligence (AI). The primary goal is to develop students' ability to think critically about how AI techniques can be effectively applied to address real-world security problems.

Upon completion of this course, the students will:

• Understand several security problems, e.g., software vulnerability mining, vulnerability assessment, anomaly detection (K&U, LS, C)
• Understand AI techniques, e.g., neural networks, large language models (K&U, LS)
• Have knowledge of example(s) about how to apply AI to a domain-specific problem, e.g., model security events into sequential data, connect security incidents as graphs (AK&U, MJ, LS)
• Have developed critical thinking around AI and security (MJ, LS, C)

The acronyms above refer to the Dublin Descriptors covered by each learning objective, specifically: Knowledge and Understanding (K&U), Applied Knowledge and Understanding (AK&U), Making Judgment (MJ), Learning Skills (LS), and Communication (C).

Course Content

This course explores a range of rapidly evolving security topics that have seen significant advances through the application of artificial intelligence (AI) techniques. Topics may include, but are not limited to, AI-assisted software development, automated bug discovery, automated software repair, and vulnerability assessment.

The course is structured around concrete security problems. For each topic, we begin by introducing the core security challenge, then examine how it has been traditionally addressed. We subsequently explore how AI techniques have been applied to enhance or automate solutions to the problem. Finally, we discuss the open challenges and limitations that must be addressed before AI can be effectively applied to new or emerging security scenarios.

To understand state-of-the-art methods, the course focuses on the discussion of scientific papers published at top-tier venues. Some topics may be accompanied by practical implementations; therefore, programming skills are expected.

Additional Information Teaching Methods

The course is organized into weekly lectures accompanied by practical quizzes and one final report based on a selection of five research papers. A list of papers will be provided, from which each student will choose five. The report should present a critical analysis of a specific security issue addressed in the selected papers, comparing the strengths and limitations of the approaches used. Students are expected to demonstrate critical thinking by identifying open challenges and proposing possible future directions related to the problem.

Peer Review of Reports:

Students will participate in a peer-review process by evaluating reports submitted by their classmates. Each student will provide a detailed assessment based on predefined criteria, including the quality of analysis, clarity of writing, comparison of methods, and depth of critical thinking.

Self-Evaluation of Report:

In addition to peer review, students will complete a self-evaluation of their own report. This will involve providing a detailed self-assessment against the same grading criteria, encouraging reflection on their work and the learning process.

Method of Assessment

The overall grade is determined as follows:

• Quizzes (30%)

Three quizzes, each worth 10%.

• Report (20%)

Report content: 15%, assessed based on depth of analysis, clarity, comparison of methods, and critical insight.
Peer review quality: 5%, based on the thoroughness and constructiveness of the peer reviews provided.

• Final Exam (50%)

A comprehensive exam covering the full course content.

Pass or fail
Quality of the reviews. Additional points will be deducted from students who write poor (e.g., too generic, such as "it is good") or unfair reviews.

To pass the course, you must satisfy the following criteria:

• You must score at least 6.0 on the final exam, and
• You must score 6.0 in total (i.e., with all three components taken together).

Resit
There is a resit for the final exam (50%). You will have to score 6.0 to pass the resit.
It is not possible to resit the assignments and the peer reviews.

Please note that your grade will be registered as follows:

For the first exam date: if you did not submit any quizzes, report, and did not participate in the exam, the result will be registered as no show (NS). For the second exam date: if you did not participate in the resit exam, the result will be registered as no show (NS). In all other cases, you will receive either a numeric grade or incomplete (NVD).

If you did not meet the minimum grade requirements for the exam (6.0), but your weighted average is >= 6.0, your grade is NVD. Otherwise, your final grade is your weighted average, rounded to half points except for 6.0.

If you plan to drop the course, do not submit anything, as NVD is administratively considered a failing grade.

Entry Requirements

While there are no formal entry requirements, however, students are expected to have foundational knowledge in either cybersecurity or artificial intelligence to fully benefit from the material.

Literature

No set book. All materials (i.e., papers, slides) will be made available during the course.

Additional Information Target Audience

Master Computer Security
Master Computer Science
Master Artificial Intelligence

Additional Information

Please see the information on Canvas.

Explanation Canvas

The submission of the report and the peer review of the report will be done in Canvas/FeedbackFruit.

Because of the way the system works by automatically assigning reviewers, it is not possible to grant any deadline extension for the submission of the reports or the submission of the reviews.

Students should therefore carefully plan their activities and submit also a draft version of their reports well in advance of the deadline (multiple submissions are possible) rather than waiting for the last moment.

Recommended background knowledge

While there are no formal entry requirements, however, students are expected to have foundational knowledge in either cybersecurity or artificial intelligence to fully benefit from the material. Programming skills are also expected.