Projects

NeSyDebates develops neuro-symbolic debate systems to safeguard Large Language Models and text-to-image generative AI against violations of Policies, Regulations, and Laws (PRLs). Using Computational Logic and Argumentation, the system detects, explains, and enforces PRL compliance in a transparent, contestable manner, validated in legal document processing and online safety settings.

Our research uncovers critical privacy risks in text-to-image diffusion models, showing that sensitive attributes such as authorship style and even dementia-related speech markers can leak into generated images. We demonstrate that adversaries can infer these attributes with high accuracy, even from images alone, highlighting threats of unauthorized profiling and discrimination. By developing new adversarial models, multimodal embedding analyses, and explainability-driven metrics, we expose how generative AI transforms and propagates sensitive information—underscoring the urgent need for privacy-preserving safeguards in next-generation AI systems. In response we spearhead the design and evaluation of foundational safeguards.

Study on the security and privacy of machine learning models and other applications in IoT and cyber-physical systems.

Study of adversarial capabilities and development of novel defense strategies for smartphone and IoT systems.

Study of effective and efficient authentication, authorization and access control mechanisms.