USENIX Security2022Top-tier venue
Characterizing the Security of Github CI Workflows
Igibek Koishybayev, Aleksandr Nahapetyan, Raima Zachariah, Siddharth Muralee, Bradley Reaves, Alexandros Kapravelos, Aravind Machiry
Abstract
Continuous integration and deployment (CI/CD) has revolutionized software development and maintenance. Commercial CI/CD platforms provide services for specifying and running CI/CD actions. However, they present a security risk in their own right, given their privileged access to secrets, infrastructure, and ability to fetch and execute arbitrary code. In this paper, we study the security of the newly popular GitHub CI platform. We first identify four fundamental security properties that must hold for any CI/CD system: Admittance Control, Execution Control, Code Control, and Access to Secrets. We then examine if GitHub CI enforces these properties in comparison with the other five popular CI/CD platforms. We perform a comprehensive analysis of 447,238 workflows spanning 213,854 GitHub repositories. We made several disturbing observations. Our analysis shows that 99.8% of workflows are overprivileged and have read-write access (instead of readonly) to the repository. In addition, 23.7% of workflows are triggerable by a pull_request and use code from the underlying repository. An attacker can exploit these workflows and execute arbitrary code as part of the workflow. Due to the modular nature of workflows, we find that 99.7% of repositories in our dataset execute some externally developed plugin, called "Actions" 1 , for various purposes. We found that 97% of repositories execute at least one Action that does not originate with a verified creator, and 18% of repositories in our dataset execute at least one Action with missing security updates. These represent potential attack vectors that can be used to compromise the execution of workflows, consequently leading to supply chain attacks. This work highlights the systemic risks inherent in CI/CD platforms like GitHub CI; we also present our own Github action, GWChecker, which functions as an early warning system for bad practices that violate the identified security properties. 2 GitHub's CI/CD product is called GitHub Actions. However, to avoid any confusion with actions (the external modules that can be used in workflows), we use GitHub CI instead. 3 In the rest of the paper we use "workflows" to refer to GitHub Workflows
Ask about this paper
Your agent reads all of it.
Lune indexed this paper to the last equation, along with the top-tier papers that cite it. Ask a question and the answer quotes them.
Your agent calls
Luneget_paper_fulltext
Free to start. No credit card required.
Terminal
Install the CLIlune papers fulltext 044bf646-cc52-44b6-a0eb-4f6256b47b9dCited by top-tier papers12
- Resource Usage and Optimization Opportunities in Workflows of GitHub ActionsIslem Bouzenia, Michael PradelICSE 2024 · 15 citations
- More Haste, Less Speed: Cache Related Security Threats in Continuous Integration ServicesYacong Gu, Lingyun Ying, Huajun Chai, Yingyuan Pu et al.S&P 2024 · 4 citations
- Action Required: A Mixed-Methods Study of Security Practices in GitHub ActionsYusuke Kubo, Fumihiro Kanei, Mitsuaki Akiyama, Takuro Wakai et al.NDSS 2026 · 3 citations
- AssetHarvester: A Static Analysis Tool for Detecting Secret-Asset Pairs in Software ArtifactsSetu Kumar Basak, K. Virgil English, Ken Ogura, Vitesh Kambara et al.ICSE 2025 · 1 citation
- Leaky Apps: Large-scale Analysis of Secrets Distributed in Android and iOS AppsDavid Schmidt, Sebastian Schrittwieser, Edgar R. WeipplCCS 2025
Builds on2
- How Bad Can It Git? Characterizing Secret Leakage in Public GitHub RepositoriesMichael Meli, Matthew R. McNiece, Bradley ReavesNDSS 2019 · 130 citations
- Configuration smells in continuous delivery pipelines: a linter and a six-month study on GitLabCarmine Vassallo, Sebastian Proksch, Anna Jancso, Harald C. Gall et al.FSE 2020 · 43 citations
Related papers
- Toward Understanding the Security of Plugins in Continuous Integration ServicesXiaofan Li, Yacong Gu, Chu Qiao, Zhenkai Zhang et al.CCS 2024
- Continuous Intrusion: Characterizing the Security of Continuous Integration ServicesYacong Gu, Lingyun Ying, Huajun Chai, Chu Qiao et al.S&P 2023
- ARGUS: A Framework for Staged Static Taint Analysis of GitHub Workflows and ActionsSiddharth Muralee, Igibek Koishybayev, Aleksandr Nahapetyan, Greg Tystahl et al.USENIX Security 2023
- COSSETER: GitHub Actions Permission Reduction Using Demand-Driven Static AnalysisGreg Tystahl, Jonah Ghebremichael, Siddharth Muralee, Sourag Cherupattamoolayil et al.S&P 2026 · 3 citations
- Comment and Control: Hijacking Agentic Workflows via Context-Grounded EvolutionNeil Fendley, Zhengyu Liu, Aonan Guan, Jiacheng Zhong et al.CCS 2026
