Welcome to InfoLab!

We are excited to welcome Leila, Abdulrahman, Mazen, and Zichen as new graduate students joining our lab. Looking forward to their contributions and collaborations!

Natural Language Processing (NLP) models even those built for transparency and interpretability remain highly vulnerable to adversarial manipulation. Our newly submitted paper, AdvChar, presents a stealthy black-box attack that preserves semantic meaning and interpretability while misleading classifiers. By applying minimal character-level modifications (only two characters on average), AdvChar can dramatically deteriorate model accuracy across seven NLP models and three interpretation paradigms revealing that interpretability tools can be exploited to disguise malicious inputs. Read more on arXiv. More...

Vision Transformers (ViTs), often paired with interpretation models, are widely viewed as robust and reliable for security-critical domains such as healthcare, autonomous driving, drones, and robotics. However, our latest paper challenges this assumption by introducing AdViT, a novel attack that successfully deceives both ViT classifiers and their interpreters. AdViT achieves a 100% attack success rate across diverse models, with up to 98% misclassification confidence in white-box and 76% in black-box settings while still producing convincing interpretations. These results reveal that even state-of-the-art transformer systems remain highly vulnerable to stealthy adversarial threats. Read more on arXiv. More...

Our paper was presented at the 15th International Conference on Electrical Engineering (ICEENG 2025). More...

Our paper was published in IEEE Transactions on Reliability. More...