Selected work

Representative Papers

Bridging the Gap: Adapting Evidence to Decision Frameworks to support the link between Software Engineering academia and industry

Feb 08, 2026

This study addresses the persistent gap between systematic literature reviews (SLRs) in software engineering and their practical uptake in industry, often referred to as the evidence-to-practice translation gap. To bridge this divide, the work introduces the Evidence to Decision (EtD) framework—originally developed in health sciences—into software engineering for the first time. By convening expert panels to conduct structured evaluations of SLR evidence against multidimensional criteria, the approach generates practitioner-oriented evidence briefs and actionable recommendations. This methodology strengthens the mechanism for translating research findings into real-world decisions, offering the first application of EtD in software engineering, identifying key dimensions essential for generating trustworthy recommendations, and highlighting major challenges that must be addressed for broader adoption of the framework.

0 citationsRead paper

Command&Control (C2) Traffic Detection Via Algorithm Generated Domain (Dga) Classification Using Deep Learning And Natural Language Processing

Nov 26, 2025

To address the limitation of static blacklists in detecting DGA-generated malicious domains, this paper proposes a novel detection method integrating NLP-based lexical features with LSTM deep learning. Unlike conventional entropy-based shallow analysis, our approach is the first to incorporate character-level n-grams and lexical diversity—linguistic features—into an LSTM sequence modeling framework, enabling joint semantic and structural discrimination of dynamically generated domains. Evaluated on a hybrid dataset of 100,000 domains, the model achieves 97.2% accuracy, substantially outperforming entropy-based baselines. It demonstrates particular strength in identifying sophisticated DGA families (e.g., Botnet, Crypto) and reduces false positives on benign domains to just 1.3%, effectively balancing high detection recall with low false alarm rates.

0 citationsRead paper

Fairness Testing in Retrieval-Augmented Generation: How Small Perturbations Reveal Bias in Small Language Models

Sep 30, 2025

This study uncovers a novel fairness risk in Retrieval-Augmented Generation (RAG) systems when deployed with small language models (SLMs): the external retrieval component can amplify biases triggered by sensitive demographic cues—such as race—exacerbating fairness violations beyond hallucination. We propose the first framework systematically applying metamorphic testing (MT) to assess RAG fairness, conducting sentiment analysis experiments on Llama and Mistral SLMs under controlled demographic perturbations. Results show that approximately one-third of metamorphic relations are violated by minor demographic perturbations; a hierarchical bias structure emerges with race as the top-level attribute; and all three evaluated SLMs exhibit significant fairness vulnerabilities. This work pioneers the integration of metamorphic testing with RAG fairness analysis, delivering a reproducible, empirically grounded methodology for bias detection in trustworthy RAG systems.

0 citationsRead paper

Generating Proto-Personas through Prompt Engineering: A Case Study on Efficiency, Effectiveness and Empathy

Jul 11, 2025

To address the time-intensive, cognitively demanding, and subjectivity-prone nature of proto-persona construction in early product discovery, this study proposes a generative AI–driven automation method grounded in prompt engineering. Employing a mixed-methods (qualitative and quantitative) design, we empirically validated the approach within authentic lean startup contexts. Results demonstrate significant improvements: a 62% average reduction in construction time, lowered cognitive load, and enhanced persona quality, reusability, and stakeholder acceptance—particularly in facilitating stakeholder alignment and MVP scope definition. Our key contribution lies in the first systematic investigation of human–AI collaboration mechanisms for proto-persona generation, empirically confirming generative AI’s capacity to stimulate cognitive empathy. However, limitations persist regarding domain specificity and deep emotional empathy. This work advances human-centered AI design by bridging generative capabilities with empathic user modeling in early-stage innovation.

0 citationsRead paper

Data Augmentation and Resolution Enhancement using GANs and Diffusion Models for Tree Segmentation

May 21, 2025

To address the challenges of low crown segmentation accuracy and scarce annotated data in urban remote sensing using low-resolution imagery, this paper proposes a domain-adaptive data augmentation framework integrating generative adversarial networks (GANs) and diffusion models. Specifically, we synergistically combine Real-ESRGAN—preserving structural fidelity during super-resolution—with Latent/Stable Diffusion—to generate semantically consistent synthetic training samples—establishing an end-to-end, weakly supervised augmentation paradigm that significantly reduces reliance on manual annotations. Coupled with a U-Net-based segmentation architecture and domain-adaptive training, our method achieves over 50% improvement in Intersection-over-Union (IoU) for tree crown segmentation across heterogeneous sensor platforms and flight altitudes. This substantially enhances model generalizability and robustness, offering a novel, cost-effective, and high-accuracy paradigm for dynamic urban forest monitoring.

0 citationsRead paper

Recent publications

Latest Papers

Bridging the Gap: Adapting Evidence to Decision Frameworks to support the link between Software Engineering academia and industry

Feb 08, 2026

This study addresses the persistent gap between systematic literature reviews (SLRs) in software engineering and their practical uptake in industry, often referred to as the evidence-to-practice translation gap. To bridge this divide, the work introduces the Evidence to Decision (EtD) framework—originally developed in health sciences—into software engineering for the first time. By convening expert panels to conduct structured evaluations of SLR evidence against multidimensional criteria, the approach generates practitioner-oriented evidence briefs and actionable recommendations. This methodology strengthens the mechanism for translating research findings into real-world decisions, offering the first application of EtD in software engineering, identifying key dimensions essential for generating trustworthy recommendations, and highlighting major challenges that must be addressed for broader adoption of the framework.

0 citationsRead paper

Command&Control (C2) Traffic Detection Via Algorithm Generated Domain (Dga) Classification Using Deep Learning And Natural Language Processing

Nov 26, 2025

To address the limitation of static blacklists in detecting DGA-generated malicious domains, this paper proposes a novel detection method integrating NLP-based lexical features with LSTM deep learning. Unlike conventional entropy-based shallow analysis, our approach is the first to incorporate character-level n-grams and lexical diversity—linguistic features—into an LSTM sequence modeling framework, enabling joint semantic and structural discrimination of dynamically generated domains. Evaluated on a hybrid dataset of 100,000 domains, the model achieves 97.2% accuracy, substantially outperforming entropy-based baselines. It demonstrates particular strength in identifying sophisticated DGA families (e.g., Botnet, Crypto) and reduces false positives on benign domains to just 1.3%, effectively balancing high detection recall with low false alarm rates.

0 citationsRead paper

Fairness Testing in Retrieval-Augmented Generation: How Small Perturbations Reveal Bias in Small Language Models

Sep 30, 2025

This study uncovers a novel fairness risk in Retrieval-Augmented Generation (RAG) systems when deployed with small language models (SLMs): the external retrieval component can amplify biases triggered by sensitive demographic cues—such as race—exacerbating fairness violations beyond hallucination. We propose the first framework systematically applying metamorphic testing (MT) to assess RAG fairness, conducting sentiment analysis experiments on Llama and Mistral SLMs under controlled demographic perturbations. Results show that approximately one-third of metamorphic relations are violated by minor demographic perturbations; a hierarchical bias structure emerges with race as the top-level attribute; and all three evaluated SLMs exhibit significant fairness vulnerabilities. This work pioneers the integration of metamorphic testing with RAG fairness analysis, delivering a reproducible, empirically grounded methodology for bias detection in trustworthy RAG systems.

0 citationsRead paper

Generating Proto-Personas through Prompt Engineering: A Case Study on Efficiency, Effectiveness and Empathy

Jul 11, 2025

To address the time-intensive, cognitively demanding, and subjectivity-prone nature of proto-persona construction in early product discovery, this study proposes a generative AI–driven automation method grounded in prompt engineering. Employing a mixed-methods (qualitative and quantitative) design, we empirically validated the approach within authentic lean startup contexts. Results demonstrate significant improvements: a 62% average reduction in construction time, lowered cognitive load, and enhanced persona quality, reusability, and stakeholder acceptance—particularly in facilitating stakeholder alignment and MVP scope definition. Our key contribution lies in the first systematic investigation of human–AI collaboration mechanisms for proto-persona generation, empirically confirming generative AI’s capacity to stimulate cognitive empathy. However, limitations persist regarding domain specificity and deep emotional empathy. This work advances human-centered AI design by bridging generative capabilities with empathic user modeling in early-stage innovation.

0 citationsRead paper

Data Augmentation and Resolution Enhancement using GANs and Diffusion Models for Tree Segmentation

May 21, 2025

To address the challenges of low crown segmentation accuracy and scarce annotated data in urban remote sensing using low-resolution imagery, this paper proposes a domain-adaptive data augmentation framework integrating generative adversarial networks (GANs) and diffusion models. Specifically, we synergistically combine Real-ESRGAN—preserving structural fidelity during super-resolution—with Latent/Stable Diffusion—to generate semantically consistent synthetic training samples—establishing an end-to-end, weakly supervised augmentation paradigm that significantly reduces reliance on manual annotations. Coupled with a U-Net-based segmentation architecture and domain-adaptive training, our method achieves over 50% improvement in Intersection-over-Union (IoU) for tree crown segmentation across heterogeneous sensor platforms and flight altitudes. This substantially enhances model generalizability and robustness, offering a novel, cost-effective, and high-accuracy paradigm for dynamic urban forest monitoring.

0 citationsRead paper