Selected work

Representative Papers

A Security Analysis of Long-Horizon Agentic AI Systems: Threats, Evaluation, and Framework Development

Jun 12, 2026

Long-term autonomous AI systems currently lack systematic safety analyses and a unified evaluation framework. This work addresses this gap by conducting a comprehensive systematization of existing literature, developing a threat model, and constructing a taxonomy to propose the first security threat classification scheme and attack propagation analysis framework tailored specifically for such systems. By clarifying the landscape of extant threat types and their underlying propagation mechanisms, this study establishes a structured foundation and theoretical underpinning for future research and practical efforts aimed at enhancing the safety and robustness of long-term autonomous agents.

0 citationsRead paper

Brain Stroke Detection and Classification Using CT Imaging with Transformer Models and Explainable AI

Jul 13, 2025

Stroke is a leading cause of global mortality, necessitating rapid, accurate, and interpretable CT-based diagnosis in emergency settings. To address this, we propose a vision Transformer–based multi-class classification framework for automated differentiation among ischemic stroke, hemorrhagic stroke, and non-stroke cases. Our method innovatively integrates the MaxViT backbone with Grad-CAM++ for pixel-level interpretability, enhancing clinical trustworthiness without compromising performance. We further improve generalizability via comprehensive data augmentation and synthetic CT image generation. Experimental results on a clinical CT dataset demonstrate that the optimized MaxViT model achieves 98.00% accuracy and F1-score—significantly outperforming established baseline models (e.g., ResNet-50, ViT-B/16). This work delivers an end-to-end, high-accuracy, and clinically interpretable AI solution for acute stroke triage in emergency departments.

0 citationsRead paper

Smart Waste Management System for Makkah City using Artificial Intelligence and Internet of Things

May 25, 2025

During the Hajj pilgrimage, waste generation surges dramatically and exhibits highly dynamic spatial-temporal distribution, rendering conventional fixed-interval collection strategies ineffective and posing severe environmental and public health risks. To address this challenge, we propose TUHR—a scalable intelligent waste management system tailored for ultra-large-scale temporary religious events. TUHR introduces an adaptive waste management paradigm integrating ultrasonic fill-level sensors, MQ-series gas sensors (e.g., for H₂S detection), ESP32-based edge nodes, and lightweight AI models to enable real-time fill-level and hazardous gas monitoring, anomaly detection, and coordinated optimization of collection routes. The system supports dynamic scheduling and precise resource allocation, improving waste response time by 42% and reducing unnecessary patrol fuel consumption by 31.5%, thereby significantly enhancing sanitation resilience at holy sites. TUHR provides a generalizable, scalable methodology for intelligent solid waste management in high-density transient environments, directly supporting Saudi Arabia’s Vision 2030 goals for sustainable smart cities.

0 citationsRead paper

A Smart Healthcare System for Monkeypox Skin Lesion Detection and Tracking

May 25, 2025

The global monkeypox outbreak necessitates efficient, accessible intelligent diagnostic tools. This paper proposes ITMAINN, an end-to-end intelligent healthcare system featuring a novel three-module architecture integrating a lightweight MobileViT backbone: (1) mobile-based skin lesion image classification (supporting binary detection and six-class fine-grained differentiation), (2) personalized symptom trajectory tracking, and (3) government-level real-time epidemic situation awareness. Evaluated on public datasets, ITMAINN achieves 97.8% binary classification accuracy (F1 = 0.976) and 92.0% fine-grained classification accuracy. The optimized MobileViT model is deployed cross-platform in a mobile application, coupled with a real-time web dashboard. To our knowledge, this is the first work to deeply adapt MobileViT for monkeypox-specific dermatological lesion recognition—achieving high accuracy, low inference latency, and strong deployability. Deployed in real-world settings, ITMAINN enables early screening, geographically informed triage recommendations, and epidemiological trend analysis, significantly enhancing primary-care screening efficiency and public health response capability.

0 citationsRead paper

Financial Fraud Detection Using Explainable AI and Stacking Ensemble Methods

May 15, 2025

Addressing the regulatory and trust challenge in financial fraud detection—where model accuracy and interpretability are often mutually exclusive—this paper proposes a stacked ensemble framework for trustworthy AI. The method integrates high-performance base models (XGBoost, LightGBM, and CatBoost) and systematically unifies multiple interpretability techniques: SHAP (for global and local explanations), LIME (for instance-level explanations), partial dependence plots (PDPs) (for feature effects), and permutation feature importance (PFI), enabling multi-granularity, multi-dimensional interpretability. Evaluated on the IEEE-CIS real-world transaction dataset (>590K samples), the framework achieves 99% accuracy and an AUC-ROC of 0.99, substantially outperforming existing approaches. To our knowledge, this is the first work to achieve both high predictive accuracy and audit-grade interpretability in large-scale financial fraud detection, thereby satisfying stringent regulatory requirements—including GDPR and BCBS guidelines—on model transparency and decision traceability.

0 citationsRead paper

Recent publications

Latest Papers

A Security Analysis of Long-Horizon Agentic AI Systems: Threats, Evaluation, and Framework Development

Jun 12, 2026

Long-term autonomous AI systems currently lack systematic safety analyses and a unified evaluation framework. This work addresses this gap by conducting a comprehensive systematization of existing literature, developing a threat model, and constructing a taxonomy to propose the first security threat classification scheme and attack propagation analysis framework tailored specifically for such systems. By clarifying the landscape of extant threat types and their underlying propagation mechanisms, this study establishes a structured foundation and theoretical underpinning for future research and practical efforts aimed at enhancing the safety and robustness of long-term autonomous agents.

0 citationsRead paper

Brain Stroke Detection and Classification Using CT Imaging with Transformer Models and Explainable AI

Jul 13, 2025

Stroke is a leading cause of global mortality, necessitating rapid, accurate, and interpretable CT-based diagnosis in emergency settings. To address this, we propose a vision Transformer–based multi-class classification framework for automated differentiation among ischemic stroke, hemorrhagic stroke, and non-stroke cases. Our method innovatively integrates the MaxViT backbone with Grad-CAM++ for pixel-level interpretability, enhancing clinical trustworthiness without compromising performance. We further improve generalizability via comprehensive data augmentation and synthetic CT image generation. Experimental results on a clinical CT dataset demonstrate that the optimized MaxViT model achieves 98.00% accuracy and F1-score—significantly outperforming established baseline models (e.g., ResNet-50, ViT-B/16). This work delivers an end-to-end, high-accuracy, and clinically interpretable AI solution for acute stroke triage in emergency departments.

0 citationsRead paper

Smart Waste Management System for Makkah City using Artificial Intelligence and Internet of Things

May 25, 2025

During the Hajj pilgrimage, waste generation surges dramatically and exhibits highly dynamic spatial-temporal distribution, rendering conventional fixed-interval collection strategies ineffective and posing severe environmental and public health risks. To address this challenge, we propose TUHR—a scalable intelligent waste management system tailored for ultra-large-scale temporary religious events. TUHR introduces an adaptive waste management paradigm integrating ultrasonic fill-level sensors, MQ-series gas sensors (e.g., for H₂S detection), ESP32-based edge nodes, and lightweight AI models to enable real-time fill-level and hazardous gas monitoring, anomaly detection, and coordinated optimization of collection routes. The system supports dynamic scheduling and precise resource allocation, improving waste response time by 42% and reducing unnecessary patrol fuel consumption by 31.5%, thereby significantly enhancing sanitation resilience at holy sites. TUHR provides a generalizable, scalable methodology for intelligent solid waste management in high-density transient environments, directly supporting Saudi Arabia’s Vision 2030 goals for sustainable smart cities.

0 citationsRead paper

A Smart Healthcare System for Monkeypox Skin Lesion Detection and Tracking

May 25, 2025

The global monkeypox outbreak necessitates efficient, accessible intelligent diagnostic tools. This paper proposes ITMAINN, an end-to-end intelligent healthcare system featuring a novel three-module architecture integrating a lightweight MobileViT backbone: (1) mobile-based skin lesion image classification (supporting binary detection and six-class fine-grained differentiation), (2) personalized symptom trajectory tracking, and (3) government-level real-time epidemic situation awareness. Evaluated on public datasets, ITMAINN achieves 97.8% binary classification accuracy (F1 = 0.976) and 92.0% fine-grained classification accuracy. The optimized MobileViT model is deployed cross-platform in a mobile application, coupled with a real-time web dashboard. To our knowledge, this is the first work to deeply adapt MobileViT for monkeypox-specific dermatological lesion recognition—achieving high accuracy, low inference latency, and strong deployability. Deployed in real-world settings, ITMAINN enables early screening, geographically informed triage recommendations, and epidemiological trend analysis, significantly enhancing primary-care screening efficiency and public health response capability.

0 citationsRead paper

Financial Fraud Detection Using Explainable AI and Stacking Ensemble Methods

May 15, 2025

Addressing the regulatory and trust challenge in financial fraud detection—where model accuracy and interpretability are often mutually exclusive—this paper proposes a stacked ensemble framework for trustworthy AI. The method integrates high-performance base models (XGBoost, LightGBM, and CatBoost) and systematically unifies multiple interpretability techniques: SHAP (for global and local explanations), LIME (for instance-level explanations), partial dependence plots (PDPs) (for feature effects), and permutation feature importance (PFI), enabling multi-granularity, multi-dimensional interpretability. Evaluated on the IEEE-CIS real-world transaction dataset (>590K samples), the framework achieves 99% accuracy and an AUC-ROC of 0.99, substantially outperforming existing approaches. To our knowledge, this is the first work to achieve both high predictive accuracy and audit-grade interpretability in large-scale financial fraud detection, thereby satisfying stringent regulatory requirements—including GDPR and BCBS guidelines—on model transparency and decision traceability.

0 citationsRead paper