Integrasi CodeBERT untuk Peningkatan Keamanan dalam CI/CD Pipeline pada Proses Deployment Aplikasi ke Cloud

Abstract

The digital transformation and widespread adoption of cloud computing have driven the use of Continuous Integration/Continuous Deployment (CI/CD) methods in modern application development. However, conventional CI/CD practices often lack comprehensive security testing, focusing primarily on functional validation. This study proposes the integration of the CodeBERT artificial intelligence model into the CI/CD pipeline to enhance application security before cloud deployment. CodeBERT is utilized to automatically detect code vulnerabilities, including severity levels and vulnerability types, through a multi-task learning approach. A CVE-based dataset is used to fine-tune the model. The system is implemented using GitHub Actions and Firebase for development and deployment. Test results show that CodeBERT achieves an accuracy of 80.6% for severity classification and 76.0% for vulnerability type classification. This research demonstrates that AI integration—such as CodeBERT—into the CI/CD pipeline can proactively and efficiently strengthen security in cloud-based DevOps workflows.