Publications.

In this paper, we design and propose a Confidence-Aware Empathic Conversation Framework to address the limitations of smaller-parameter Large Language Models (LLMs) such as FLAN-T5-Large in affective and mental health applications. We propose a novel architecture that gates LLM outputs … Read more

In this paper, we design and propose a Confidence-Aware Empathic Conversation Framework to address the limitations of smaller-parameter Large Language Models (LLMs) such as FLAN-T5-Large in affective and mental health applications. We propose a novel architecture that gates LLM outputs through a dual-metric scoring system combining keyword-based heuristic empathy checking and vector-space semantic relevance evaluation, unified as Stotal = H + (3 x R). We implement a Dynamic Sentiment-Aware Anchoring strategy that pre-fills assistant responses with emotionally appropriate anchors depending on detected user sentiment, restricting the decoder search space to a corresponding emotional subspace. We further employ a Confidence Gate with a strict threshold (tau = 2.5) that triggers a safe fallback response when no candidate meets the required quality bar. We evaluate our system through a multi-turn interaction session on academic and work-related stress scenarios using FLAN-T5-Large for generation and all-MiniLM-L6-v2 for semantic embedding evaluation. Our experimental results show that the proposed Hard-Anchoring strategy effectively prevents incoherent and emotionally misaligned outputs, with all generated responses remaining well above the confidence threshold. We achieve a mean system confidence of 4.74 across five turns, with the highest confidence score of 6.79 recorded when the system detected a user shift toward a positive emotional state and reinforced it appropriately. These results demonstrate that our framework substantially improves empathic coherence, response reliability, and emotional alignment in human-AI dialogue, offering a principled path toward safer AI deployment in mental health support contexts. Show less

Connected and software-defined vehicles now rely on edge infrastructure for access, payment, and identity-linked services. Entry and exit at a parking facility carry financial and privacy risk: a plate may be read incorrectly, a payment may arrive twice, the cloud … Read more

Connected and software-defined vehicles now rely on edge infrastructure for access, payment, and identity-linked services. Entry and exit at a parking facility carry financial and privacy risk: a plate may be read incorrectly, a payment may arrive twice, the cloud link may drop, and plate-payment records may stay linkable after the vehicle has left. We describe a cyber-resilient session protocol for software-defined vehicular edge infrastructure. The on-board unit exchanges entry, payment, exit and closure messages with a roadside edge service over IEEE 802.11p; a short-lived pseudonymous token holds the session together, payment and exit invariants are enforced locally, state is cached while the cloud is unreachable, and low-confidence ALPR events are routed to a review queue. The control logic was modelled in UPPAAL 5.0.0 and verified for duplicate-payment exclusion, penalty exclusion, exit-induced closure, synchronization consistency, deadlock freedom, eventual receipt completion, and receipt-deadline safety. An 11,000-second SUMO/OMNeT++/Veins run instantiated 200 vehicles across loss cohorts of 0/5/10/20%, with mean application-level response times of ~50 ms (token, payment, exit) and ~100 ms (receipts). The two evaluations give complementary formal-safety and wireless-execution evidence for privacy-preserving, cyber-resilient V2I session management. Show less