Secure and Explainable AIP System Integrated with Devops Principles Using HLGSRU and SCC

Abstract

Currently, to enhance the development, deployment, monitoring, and maintenance of Automated Invoice Processing (AIP) systems, Development and Operations (DevOps) are increasingly used. Nevertheless, traditional methodologies failed to ensure the security of audit log files from unauthorized persons or intruders, leading to data leakage. Thus, this paper proposes a DevOps-based secure and explainable AIP system using Hinge Loss Gated Smish Recurrent Contrastive Unit (HLGSRU) and Shimura Curve Cryptography (SCC). Primarily, admins create invoice entries, followed by text extraction. Further, by preprocessing the invoice NER data, the Named Entity Recognition (NER) model is trained, followed by layout preservation and visual feature extraction. Concurrently, Part of Speech (PoS) tagging and contextual and syntactic pattern identification are carried out using the Hidden Skellam Markov Model (HSMM) and Nesterov Accelerated Gradient-based Financial Bidirectional Encoder Representations from Transformers (NAG-finBERT), respectively, followed by feature extraction. Now, HLGSRU performs NER classification based on the extracted features and identified patterns, followed by a deep explanation. Lastly, AIP systems are developed based on NER by securely storing audit files using SCC through a CI/CD pipeline. Therefore, the proposed framework outperforms the other conventional methodologies with higher accuracy (98.75%).