Projects

Large Language Models (LLMs) are increasingly used in software development through so-called “vibe coding,” where developers specify tasks in natural language and rely on the model to produce executable code. While this paradigm lowers barriers to entry and accelerates prototyping, it raises concerns about security. Prior studies show that a substantial fraction of AI-generated code contains exploitable vulnerabilities, and functional correctness does not guarantee safety. This paper investigates whether security-oriented prompting improves the security of LLM-generated code. We design ten representative Python tasks inspired by OWASP Top 10 and CWE categories, and evaluate outputs from an open-source 20B-parameter model using static analysis (Bandit) and lightweight runtime probes.

Digital attention fatigue is a pervasive challenge, yet most virtual reality (VR) interventions for restoration rely on passive nature exposure that lacks responsiveness to the user's internal state. Integrating Attention Restoration Theory (ART) with physiological computing, we propose a bioadaptive VR system that combines active mandala drawing within a 360° nature scene, using real-time heart rate variability (HRV) to modulate visual fog, ambient music, and haptic feedback. In a within-subject pilot study (N=11), we compared an AI-driven bioadaptive multimodal condition (AI) against an otherwise identical VR condition without bioadaptive multimodal feedback (NF), using behavioral (Oddball task), neural (EEG), autonomic (HRV), and subjective measures.