Safety of GenAI Through the Lens of Security and Cryptography

Overview

Explore safety challenges of Generative AI through security and cryptography perspectives, with insights from Somesh Jha's theoretical approach to trustworthy AI.

Syllabus

• Introduction to Generative AI
Overview of Generative AI
Recent advancements and applications
Importance of safety in Generative AI
• Security Challenges in Generative AI
Identifying potential threats
Privacy concerns and data protection
Adversarial attacks on Generative AI models
• Cryptographic Techniques for AI Safety
Basics of cryptography relevant to AI
Encryption methods to secure AI models
Privacy-preserving machine learning techniques
• Somesh Jha's Theoretical Approach
Overview of Jha's contributions to AI safety
Insights from theoretical frameworks
Case studies and practical examples
• Trustworthy AI Systems
Principles of designing trustworthy AI
Verification and validation of AI models
Building resilient and robust Generative AI
• Ethical Considerations in AI Security
Balancing privacy and functionality
AI governance and regulatory frameworks
Societal impacts and ethical dilemmas
• Practical Approaches and Tools
Tools for securing Generative AI
Implementing cryptographic solutions
Hands-on exercises and projects
• Future Directions for AI Safety
Emerging trends in AI security and cryptography
Research opportunities and open challenges
Collaboration between academia and industry
• Conclusion and Review
Summary of key takeaways
Final discussions and Q&A session
Course feedback and evaluation

Subjects

Computer Science