research
Security of learning systems and software systems, and explainable fuzzy inference.
Backdoor Attacks in Large Language Model Reasoning
Investigating how the reasoning process of large language models can be compromised, combining chain-of-thought analysis with trigger-injection techniques.
Security Analysis of Federated Learning and Split Learning
Systematizing whether attacks and defenses established for federated learning transfer to split learning, and where the architectural split invalidates the assumptions each relies on (SoK manuscript in preparation for IEEE S&P 2027).
Improving Security of Systems Integrated in Ultrasound Devices
Developing a contract-based method to automatically scope security re-verification to only the modules an update could affect, instead of retesting the entire ultrasound software stack.
Interval Type-2 Fuzzy Sets: Behavior and Performance
Designed the Interval Creation Approach (ICA), a method for constructing interval type-2 fuzzy sets from type-1 fuzzy set, leading to a first-author paper at FUZZ-IEEE 2024, as a part of IEEE WCCI 2024.