惯性聚合 高效追踪和阅读你感兴趣的博客、新闻、科技资讯
阅读原文 在惯性聚合中打开

推荐订阅源

L
LangChain Blog
酷 壳 – CoolShell
酷 壳 – CoolShell
雷峰网
雷峰网
量子位
V
V2EX
S
SegmentFault 最新的问题
月光博客
月光博客
博客园 - 【当耐特】
Hugging Face - Blog
Hugging Face - Blog
V
Visual Studio Blog
大猫的无限游戏
大猫的无限游戏
T
Tailwind CSS Blog
博客园_首页
博客园 - Franky
钛媒体:引领未来商业与生活新知
钛媒体:引领未来商业与生活新知
美团技术团队
Y
Y Combinator Blog
The Cloudflare Blog
C
Check Point Blog
Cyber Security Advisories - MS-ISAC
Cyber Security Advisories - MS-ISAC
腾讯CDC
B
Blog
Stack Overflow Blog
Stack Overflow Blog
P
Proofpoint News Feed

Cyber Security News

AIRecon: AI-Powered Penetration Testing Tool with Kali Linux Sandbox Critical Chrome Vulnerabilities Allow Attackers to Execute Arbitrary Code - Update Now! UNC3753 Uses Screen-Sharing Sessions and RMM Tools to Exfiltrate Sensitive Legal Data New OnionDrop Loader Campaign Uses gainmsg C2 to Deliver LegionLoader Payloads ClickFix Campaign Uses EtherHiding and GULoader to Infect Windows Users via Fake CAPTCHA Ghostwriter Hackers Abuse Gmail Admin-Themed Emails to Steal Credentials and 2FA Codes The Half-Life of Threat Intelligence: When Does an IOC Stop Being Useful? Critical Fortinet FortiSandbox Vulnerabilities Actively Exploited in Attacks Aembit Extends IAM for Agentic AI to Microsoft Copilot Studio India Temporarily Bans Telegram Messenger Over Medical Exam Fraud Microsoft 365 Device Code Phishing Campaign Bypasses Password Theft With Legitimate Login Flow AppViewX Launches Agent Identity Security to Govern Agents for the AI and Quantum Era Hackers Weaponize Microsoft Teams Relay to Hide Ransomware Traffic Developer laptops are the credential store attackers are picking through in 2026, GitGuardian announces Endpoint Protection Interlock and Rhysida Ransomware Operations Share Supper Backdoor and Malware Codebase Novo Nordisk Confirms Cyber Attack — Hackers Accessed Patient Medical Data and Internal AI Assets Russian and Chinese Influence Actors Use AI to Evade Bot Detection and Mimic Human Behavior Microsoft Teams Analyze the Wi-Fi Hotspot Data Connected to an Employee’s Device PRC-Nexus Hackers Exploit REDCap Servers to Spy on US Medical Research Institutions Infinite Campus Data Breach Exposes 137,000 Users Personal Details OptinMonster Plugin Hack Exposes 1.2 Million Wordpress Sites to Cyberattack Ransomware Ecosystem Consolidates Around LockBit Alumni, Qilin, Hyflock, and The Gentlemen Hackers Abuse Legitimate RMM Tools in The Quarry IRS and SSA Phishing Campaigns LiteSpeed cPanel Plugin 0-Day Vulnerability Actively Exploited in the Wild Cisco SD-WAN vManage Vulnerability Exploited in Zero-Day Attacks Nearly 14,000 SimpleHelp Servers Exposed Amid Critical Authentication Bypass Disclosure Microsoft Site Showing Warning Following Certificate Expiry DPAPISnoop Tool Extracts CREDHIST Hashes for Offline Windows Credential Recovery SHADOWBYT3$ Allegedly Claim Breach of Nintendo, Stealing Sensitive Data Anthropic Updated Privacy Policy to Include Identity Verification for Claude Users
GitHub Actions Checkout Update Blocks Workflows Triggered...
Abinaya · 2026-06-22 · via Cyber Security News

GitHub has rolled out a significant security enhancement to GitHub Actions by updating actions/checkout to block unsafe workflows that abuse the pull_request_target event.

The pull_request_target trigger is widely known as one of the most misused events because it runs with the base repository’s GITHUB_TOKEN, secrets, and default-branch cache access, even when the pull request comes from an untrusted fork.

When maintainers check out the head or merge commit of a forked pull request in this context, attacker-controlled code can run with full CI privileges, a pattern referred to as a “pwn request” that has driven multiple supply-chain compromises across the ecosystem.

To reduce this risk, GitHub has made actions/checkout v7 generally available with secure-by-default behavior designed to stop the most common pwn request patterns.

GitHub Checkout Update Disrupts Workflows

On July 16, 2026, GitHub will backport this enforcement logic into all currently supported major versions, so workflows pinned to floating tags like actions/checkout@v4 will automatically inherit the safer defaults without any manual change.

Pipelines that pin actions/checkout to a specific SHA, minor, or patch version will not be updated automatically and must be upgraded via Dependabot or established internal processes to benefit from the new protections.

Same-repository pull requests are unaffected, and the traditional pull_request event behavior remains unchanged, ensuring standard contribution workflows continue as usual.

The key change is that actions/checkout v7 now refuses to fetch fork pull request code in pull_request_target workflows and in workflow_run jobs where workflow_run.event is a pull_request* event, whenever the configuration clearly targets a fork PR.

The action fails if the repository input resolves to the fork’s repository, if ref matches refs/pull/<number>/head or refs/pull/<number>/merge, or if ref resolves to the fork PR’s head or merge commit SHA.

This update blocks insecure GitHub Actions patterns such as refs/pull/${{ github.event.pull_request.number }}/merge, ref: ${{ github.event.pull_request.head.sha }}, and repository: ${{ github.event.pull_request.head.repo.full_name }} in privileged workflows.

It prevents untrusted pull request code from reaching jobs that execute sensitive commands like run: ./scripts/deploy.sh, which may have access to secrets and production resources.

Under the new behavior, these workflows fail early instead of silently executing attacker-controlled code. GitHub stresses that the update is focused on the most prevalent misuse and does not claim to eliminate every variant of pwn requests.

Workflows remain vulnerable if they manually pull and execute untrusted code using run blocks that call git or the gh CLI to fetch arbitrary refs or repositories, since those operations bypass actions/checkout entirely.

Unsafe patterns triggered by other events, such as issue_comment, also remain out of scope for this change. However, GitHub may expand hardening to additional event types in the future.

The new logic specifically blocks fork PR head and merge checkouts but does not prevent checkouts of other untrusted repositories, meaning any design that executes code from untrusted sources in privileged events still requires careful review and threat modeling.

For legitimate use cases where fork PR code must run with elevated trust, for example, coverage generation using private artifact registries or authenticated checks on incoming changes, GitHub offers an explicit opt-out path.

After reviewing the official guidance on securely using pull_request_target, maintainers can add the allow-unsafe-pr-checkout input to the actions/checkout step to re-enable fork PR checkouts in these workflows.

The input name is intentionally loud and descriptive so that its presence is obvious during code review and static analysis, reinforcing that turning off the default protection is a conscious, high-impact security decision.

In practice, security-focused teams are encouraged to run untrusted forks under pull_request with restricted permissions and to reserve pull_request_target and any unsafe checkouts only for carefully audited pipelines that truly require access to secrets and sensitive resources.

Follow us on Google NewsLinkedIn, and X to Get More Instant Updates.

Abinaya

Abinayahttps://cybersecuritynews.com/

Abi is a Security Editor and fellow reporter with Cyber Security News. She is covering various cyber security incidents happening in the Cyber Space.