Privacy and Security for Large Language Models - Baihan Lin

. .

Auteur(s) : Baihan Lin

Editeur : O'reilly Media

Langue : Anglais

Parution : 01/02/2026

Format : Moyen, de 350g à 1kg

Nombre de pages : 300.0

ISBN : 9781098160845

Résumé :

As the deployment of AI technologies surges, the need to safeguard privacy and security in the use of large language models (LLMs) is more crucial than ever. Professionals face the challenge of leveraging the immense power of LLMs for personalized applications while ensuring stringent data privacy and security. The stakes are high, as privacy breaches and data leaks can lead to significant reputational and financial repercussions.

This book serves as a much-needed guide to addressing these pressing concerns. Dr. Baihan Lin offers a comprehensive exploration of privacy-preserving and security techniques like differential privacy, federated learning, and homomorphic encryption, applied specifically to LLMs. With its hands-on code examples, real-world case studies, and robust fine-tuning methodologies in domain-specific applications, this book is a vital resource for developing secure, ethical, and personalized AI solutions in today's privacy-conscious landscape.

By reading this book, you'll:

• Discover privacy-preserving techniques for LLMs
• Learn secure fine-tuning methodologies for personalizing LLMs
• Understand secure deployment strategies and protection against attacks
• Explore ethical considerations like bias and transparency
• Gain insights from real-world case studies across healthcare, finance, and more

...

Biographie:
Dr. Baihan Lin is a leading computer scientist, neuroscientist, inventor, and professor specializing in speech and natural language processing (NLP). He holds faculty positions at Harvard University and the Icahn School of Medicine at Mount Sinai. Known for his expertise in trustworthy Neuro-AI and computational psychiatry, Dr. Lin has made significant contributions to these fields through his work at Columbia University, where he earned his PhD, and through his research at leading tech companies such as IBM, Google, Microsoft, Amazon, and BGI Genomics.

His research program focuses on developing intelligent speech and text-based systems to enhance human-AI and human-human interactions in healthcare. Notably, he developed the first-ever online and reinforcement learning (RL)-based speaker diarization system and RL-based interactive spoken language understanding (SLU) systems for children with speech and communication disorders.

Dr. Lin's work in deep learning, RL, and NLP has led to real-world applications, including AI companions for therapists and context-aware virtual realities. He has authored over 50 peer-reviewed publications and patents and has served on program committees and as a reviewer for over 15 top AI conferences and more than 20 journals. He has chaired tutorials and workshops at AAAI, INTERSPEECH, ICASSP, WACV, and IJCAI, focusing on RL, human-in-the-loop language technology, and most recently, the alignment, privacy, security, and governance of generative AI.

As a finalist for the Bell Labs Prize and XPRIZE, Dr. Lin's contributions in real-time algorithms advance the understanding of the human brain and mind, support disadvantaged individuals with mental health conditions, and drive the evolution of affective and empathetic AI in the era of large language models....

Sommaire:

As the deployment of AI technologies surges, the need to safeguard privacy and security in the use of large language models (LLMs) is more crucial than ever. Professionals face the challenge of leveraging the immense power of LLMs for personalized applications while ensuring stringent data privacy and security. The stakes are high, as privacy breaches and data leaks can lead to significant reputational and financial repercussions.

This book serves as a much-needed guide to addressing these pressing concerns. Dr. Baihan Lin offers a comprehensive exploration of privacy-preserving and security techniques like differential privacy, federated learning, and homomorphic encryption, applied specifically to LLMs. With its hands-on code examples, real-world case studies, and robust fine-tuning methodologies in domain-specific applications, this book is a vital resource for developing secure, ethical, and personalized AI solutions in today's privacy-conscious landscape.

By reading this book, you'll:

• Discover privacy-preserving techniques for LLMs
• Learn secure fine-tuning methodologies for personalizing LLMs
• Understand secure deployment strategies and protection against attacks
• Explore ethical considerations like bias and transparency
• Gain insights from real-world case studies across healthcare, finance, and more
• Examine the legal and cultural landscape of AI deployment