SECURE LLM-ASSISTED TEST CASE GENERATION FOR ENTERPRISE SOFTWARE DEVELOPMENT PIPELINES

Abstract

Software testing plays a crucial role in ensuring the quality, security, reliability, and maintainability of enterprise software systems throughout the Software Development Life Cycle (SDLC). Modern enterprise software development increasingly adopts Agile methodologies, DevOps practices, Continuous Integration/Continuous Deployment (CI/CD), cloud-native architectures, microservices, and containerized applications, resulting in rapidly evolving software requirements and shorter release cycles. Conventional manual testing and script-based automated testing approaches often struggle to keep pace with these dynamic environments due to the significant effort required for test case creation, maintenance, and validation. Recent advances in Artificial Intelligence (AI), Natural Language Processing (NLP), and Large Language Models (LLMs) have introduced new opportunities for intelligent software test generation by automatically interpreting software requirements, source code, API documentation, and design specifications. However, enterprise adoption of LLM-assisted software testing remains limited because of concerns regarding data privacy, hallucinations, insecure code generation, intellectual property leakage, regulatory compliance, prompt injection attacks, and lack of explainability. This research proposes a Secure LLMAssisted Test Case Generation Framework for enterprise software development pipelines that integrates Large Language Models with secure software engineering practices, Retrieval-Augmented Generation (RAG), software requirement repositories, secure prompt engineering, policy enforcement, access control, and explainable artificial intelligence. The proposed framework retrieves relevant Software Requirement Specifications (SRS), user stories, API documentation, software architecture documents, historical test repositories, coding standards, and organizational security policies from enterprise knowledge bases before generating software test cases. Secure retrieval mechanisms combined with rolebased access control ensure that only authorized project information is supplied to the language model, thereby protecting confidential enterprise assets while improving contextual relevance. The framework employs transformer-based embedding models, semantic vector databases, secure Retrieval-Augmented Generation, prompt validation, output verification, automated security policy checking, and intelligent test generation to produce functional test cases, regression tests, integration tests, API validation scenarios, boundary value tests, security testing procedures, negative test cases, and compliance validation tests. Explainability mechanisms establish traceable relationships between software requirements, retrieved knowledge, generated test cases, and software components, enabling transparent software quality assurance and regulatory auditing. Continuous synchronization with enterprise repositories ensures that generated test suites remain consistent with evolving software requirements, coding standards, and organizational security policies. Experimental evaluation demonstrates that the proposed Secure LLM-Assisted Test Case Generation framework significantly improves software testing accuracy, requirement coverage, contextual relevance, security compliance, and developer productivity while reducing hallucinations, confidential information exposure, and manual testing effort. The integration of secure Retrieval-Augmented Generation with enterprise software development pipelines enables trustworthy AI-assisted software testing suitable for highly regulated industries including finance, healthcare, government, telecommunications, manufacturing, and cloud computing. The proposed research contributes to the development of secure, explainable, scalable, and intelligent software testing systems capable of supporting next-generation enterprise software engineering within modern DevSecOps and Industry 5.0 environments.