Security of Synthetic Data Generation
Research on security implications, auditing, and tracing of synthetic data generated by LLMs for downstream applications
Security of Synthetic Data Generation
Research on Large Language Models in Security of Synthetic Data Generation
Detecting AI-Generated Data in the Wild
New techniques to identify and audit synthetic data in downstream applications
Generating Private Synthetic Data via LLM APIs
Leveraging third-party LLMs without access to model weights
Unlocking Private Medical Text Generation
Using LLM Prompts to Create Synthetic Data While Preserving Patient Privacy
Stealing AI Art Prompts: A New Security Threat
How attackers can reverse-engineer valuable text-to-image prompts with minimal samples
Synthetic Clinical Data Generation for Privacy-Sensitive Applications
Using LLMs to create annotated training data for de-identification systems
Synthetic Data Privacy Risks
Detecting Information Leakage in LLM-Generated Text
Revolutionizing Cyberbullying Detection
Using LLM-Generated Data to Address Dataset Scarcity
Privacy-Preserving Synthetic Text
Training Better LLMs with Less Real Data
AI's Secret Languages
Preventing machines from developing communications beyond human understanding
Synthetic Tabular Data with Real-World Logic
Preserving Inter-column Relationships in Generated Data
Privacy-Preserving Synthetic Text Data
Generating high-quality text data without compromising privacy
Domain-Driven AI: The ELTEX Framework
Enhancing LLM specialization through explicit domain knowledge injection
Generating Paired Facial Images with LLMs
Leveraging AI to create thermal-visible facial image pairs for security applications
