Research

Current Research

Whose Specification Did You Test? The Hidden Interpretation Bias in LLM-Generated Test Suites

RASE Workshop @ ASE 2026 - Munich, Germany | Status: Under Review

This research examines the adequacy and reliability of REST API test suites generated by large language models. As LLMs are increasingly used to auto-generate tests directly from API specifications, this work investigates whether the resulting test suites faithfully validate the intended specification or instead encode the model's own interpretation of it - a hidden interpretation bias that can mask real defects. The study contributes an empirical analysis of LLM-generated REST API tests and outlines directions for making specification-driven test generation more reliable.

Research Areas: Software Engineering, Large Language Models, Automated Software Testing

Other Research

Depth-Derived Geometric Feature Sequences for Elderly Fall Detection Using a CNN-LSTM Network

BECITHCON 2026 - IUBAT, Dhaka, Bangladesh | Status: Under Review

Paper ID: 395 | Submitted: 16 June 2026 | Notification: 31 July 2026 | Camera-ready: 16 August 2026 | Conference Dates: 4-5 September 2026

This work proposes a CNN-LSTM network that learns geometric feature sequences derived from depth-camera data to detect falls in elderly individuals, contributing toward privacy-preserving, camera-based healthcare monitoring systems.

Research Areas: Biosignal Recording, Processing, Pattern Recognition and Machine Learning

Broader Research Interests

Software Engineering, Large Language Models, Automated API Testing, Machine Learning, Deep Learning, Computer Vision, CNN-LSTM, IoT Systems, Natural Language Processing, Healthcare AI, Data Science, Predictive Analytics, Privacy-Preserving AI, Federated Learning, Software Testing