SEC-301: Security Challenges in Modern AI Systems

Course Syllabus

Course Description

The rise of the AI era has brought AI-enabled systems like ChatGPT and DALL-E into the spotlight, demonstrating how these systems can enhance our daily lives. However, alongside their benefits come significant challenges. As AI systems become increasingly integrated into our lives, the security risks associated with them grow more critical for researchers and practitioners alike.

In this competency, we will explore the multifaceted security challenges in modern AI systems. Topics will include AI threat modeling, AI-specific attacks, and AI security risk management. The first half of this competency will emphasize hands-on security threat analysis and modeling, using real-world scenarios to develop practical skills.

Recognizing that many AI advancements stem from academic research, the second half of this competency will delve into research works on AI safety. This will provide insights into the security challenges that academic research seeks to address and help you understand the broader landscape of AI security.

The learning approach will combine lectures and discussions. Through the lectures, you will gain the knowledge and skills necessary to analyze and address security challenges in AI systems.

General Information

Competency Code SEC-301
Competency Name Security Challenges in Modern AI Systems
Competency Credits 2
Competency Duration 4 Weeks (~7.5 Hours Per Week = 30 Hours in Total)
Instructor Dr. Charnon Pattiyanon <charnon@cmkl.ac.th>

Assessing Skills

  1. [SEC-301:00010] Analyze AI Security Risks - Successful students must be able to analyze potential security risks in modern AI systems.
  2. [SEC-301:00020] Analyze AI Security Threats using Analysis Techniques - Successful students must be able to apply analysis techniques to identify security threats and vulnerabilities in modern AI systems.
  3. [SEC-301:00030] Analyze AI-Specific Attacks Scenarios - Successful students must be able to analyze attack scenarios that can be exploited in modern AI systems.
  4. [SEC-301:00040] Understand AI Safety in Academic - Successful students must be able to demonstrate their understanding of current trends, methodologies, and result landscape of AI safety research.

Class Schedule and Topics

Week Lecture Topic
Week 1 Lecture 1: AI Security Risks
  • Security Risks in Modern AI Systems
  • Security Threats and Attacks Targeted to Modern AI System
  • Analysis Techniques to Identify Security Vulnerabilities in Modern AI Systems
[Lecture 1 (PDF)]
Assessment Announcement
Week 2 Lecture 2: Security Threat Modeling Techniques
  • Overview of Techniques and Tools to Identify Security Threats within AI Systems
  • Basic Security Model
[Lecture 2 (PDF)]
Week 3 Lecture 3: Techniques and Tools for Addressing Security Challenges
  • Transfer Learning
  • Federated Learning
  • Generative Adversarial Networks
[Lecture 3 (PDF)]
Week 4 Lecture 3: Techniques and Tools for Addressing Security Challenges
  • Homomorphic Encryption
  • Differential Privacy
[Lecture 3 (PDF)]

Assessment and Submission Guideline

This competency requires each student group to submit only one deliverables, which is:

  1. A Final Report: This is the final document summarizing the details of your assessment project. Please refer to the report template provided below.

To support students throughout this competency, the following documents are provided:

Grading Rubric

Final Report Grading Rubric

The final report accounts for 100% of your total score, equivalent to 300 points out of 300 points (based on six assessing skills). Each section in the report carries a different score weight. Please refer to the Final Report Template for the detailed score distribution of each section.