AI-Enhanced Intelligent Internal Audit Automation Systems for Continuous Compliance Monitoring and Multi-Layer Risk Mitigation
الباحث الأول:
Laith Ali Muttar
الباحثين الآخرين:
Majid M. Manhosh، Hasan Abd Al-Hussein,Areej Abdalghfou,Abdulsalam Ali Hussein Alnoori,Abdulrazaq Shabeeb,Sinan Raheem Jasim,Hussain D. D,Wisam A Mohammedhasan
المجلة:
the 3rd International Conference on Cyber Resilience (ICCR-2025)
تاريخ النشر:
2 يوليو، 2025
مختصر البحث:
This research presents a high-performance AI framework for intelligent internal audit automation, integrating deep learning, semantic policy embeddings, and reinforcement-driven risk optimization. Leveraging fine-tuned transformer models and structu…
This research presents a high-performance AI framework for intelligent internal audit automation, integrating deep learning, semantic policy embeddings, and reinforcement-driven risk optimization. Leveraging fine-tuned transformer models and structured enterprise datasets, the system was evaluated across diverse audit domains including financial transactions, procurement logs, HR records, and vendor management data. The model consistently achieved detection accuracies exceeding 98.53%, with F1-scores reaching 0.94 and compliance alignment scores up to 0.94. Compared to traditional rule-based methods, the enhanced framework reduced control oversight errors by over 25% and significantly improved interpretability through SHAP-based explanations and anomaly heatmaps. The integration of contextual text embeddings, numerical audit features, and dynamic control evaluation enabled transaction-level compliance analysis and proactive risk detection. Unlike resource-intensive neural pipelines, the model maintained sub-2 second training cycles, ensuring deployment feasibility within ERP-integrated enterprise systems. The architecture supports cross-platform generalization and is extensible to various operational domains without requiring structural reengineering.
