Stolen GitHub accounts used to plant secret-stealing malware in projects
Security firm Aikido says hijacked accounts probed tens of thousands of projects from Nov. 10, and the malware ran in at least 1,835, taking CI secrets and cloud keys.
Key takeaways
- Aikido counts 4,837 compromised GitHub accounts that sent probe pull requests to 61,764 projects starting Nov. 10.
- Aikido says the credential-stealing malware ran in more than 1,800 projects and that the real number may be higher.
- StepSecurity says no GitHub Actions bug is being exploited; SecurityWeek reports the attack relies on valid stolen tokens.
- The malware targets CI secrets, cloud credentials and AI service keys, and makes AWS and Azure changes that survive secret rotation.
- Aikido advises treating every secret available to an affected project's workflows as compromised.
Attackers using stolen GitHub credentials have been planting credential-stealing malware in code projects since Nov. 10, according to security firm Aikido, which discovered the campaign and calls it "wfx-loot." Aikido counts 4,837 compromised accounts that sent probe pull requests to 61,764 projects. In at least 1,835 of them, it says, the malware ran and sent stolen credentials to the attacker. The targets are GitHub Actions secrets, cloud keys and tokens for AI services.
SecurityWeek reports that several security firms are tracking the activity. Another firm, SafeDep, found malicious commits in 8,093 public projects, though SecurityWeek notes that this does not necessarily mean the code ran there. Aikido's post gives 1,835 in its summary and 1,806 elsewhere, and says its count may understate the damage because the attacker deleted run logs.
How does the attack work?
According to Aikido, the attacker first submits a probe pull request named "docs: dependency audit notes" from a branch called wfx-docs. That request does nothing by itself; it only tests whether the stolen token has write access. Where it does, the attacker pushes a shell script of roughly 1,500 lines in one of three ways: inside an existing workflow, through an npm install hook, or in a new workflow on a short-lived branch.
Aikido says the attacker reverts the change within a minute and then deletes the workflow run and its logs. As Help Net Security puts it, a gap in the run numbering is often the only trace. Aikido says 92% of the probe pull requests were still open, which it takes as a sign that most owners have not noticed.
"The attacker does not exploit a bug in GitHub Actions," StepSecurity said, as quoted by SecurityWeek. The breach instead depends on valid tokens from developer and bot accounts. SecurityWeek reports that StepSecurity tied 173 of the probe requests to Fly.io's GitHub app tokens, but it could not tell how those tokens were obtained and saw no evidence of a breach of Fly.io's own infrastructure.
What does the malware steal?
Aikido's analysis says the script grabs GitHub Actions secrets and variables, workflow tokens, credential files for AWS, Azure, Google Cloud, Docker, Kubernetes, npm, PyPI and others, shell histories, and keys for AI providers and coding tools including OpenAI, Anthropic, Gemini, Hugging Face, Cursor and Copilot. It also reads the memory of runner processes.
It then tries to go deeper. According to Aikido, it trades GitHub OIDC tokens for cloud sessions and creates long-lived AWS access keys. It also opens an AWS role to any AWS account by changing its trust policy, and gives Azure applications new federated identity credentials. Help Net Security notes that these AWS and Azure changes outlast secret rotation, so teams must find and delete every one of them.
Did it reach software packages?
Aikido says automated publishing carried the malware into one release of a Packagist package, digitalelvis/neuronai-studio, and two tags of a Go module, onuragtas/docker-env (v1.0.278 and v1.0.280). SecurityWeek refers to two Go modules, while Aikido's list names one. For that module, the malicious code sits in a workflow file, and Aikido notes that Go builds do not execute workflow files. Aikido also found at least 50 hijacked users who appear to publish packages to registries such as npm, PyPI and crates, but it says the attacker appears to be after CI secrets rather than trying to spread through the package supply chain.
BleepingComputer reports that some compromised accounts belonged to Fortune 100 companies, to Nextcloud and WebdriverIO, and to government bodies in the US (the General Services Administration) and France. NextFish could not confirm that list in the other sources.
Aikido also saw overlap with what it calls the October GhostAction wave; BleepingComputer dates GhostAction to 2025. Accounts hit then were more likely to be hit again, and in more than 100 projects the new malware was placed inside GhostAction's own workflow files. Even so, Aikido believes different attackers are probably behind the two. None of the sources included a statement from GitHub.
What should affected teams do?
Aikido advises checking for a wfx-docs branch, the probe pull request, and commits titled "ci: refresh build cache," "chore: add ci helper" or "chore: add dependency audit workflow." If the malware ran, it says, every secret and cloud role that the project's workflows could use should be treated as compromised and rotated. SecurityWeek adds that organizations have been advised to revoke stolen tokens and look for unauthorized cloud changes.
Disclosure: NextFish's AI drafting desk uses Claude, a model made by Anthropic.
Questions readers are asking
What is wfx-loot?
Aikido's name for a campaign that uses stolen GitHub accounts to push credential-stealing malware into code projects. The name comes from the attacker's own labels for the malware and for stolen secrets.
Is this a GitHub vulnerability?
No. StepSecurity, quoted by SecurityWeek, says no GitHub Actions bug is being exploited; SecurityWeek reports the attack relies on valid tokens.
How can I tell if my project was hit?
Aikido says to look for a wfx-docs branch, a pull request titled "docs: dependency audit notes," the commit titles it lists, and gaps in workflow run numbers.
Which packages were affected?
Aikido names one Packagist package, digitalelvis/neuronai-studio, and two tags of the Go module onuragtas/docker-env. In the Go tags, the payload is only in a workflow file.
Sources
- Aikido Security: wfx-loot campaign pushes CI secret stealer to tens of thousands of repositories
- SecurityWeek: Attacker Targets 60,000 GitHub Repositories in Credential Theft Campaign
- Help Net Security: Check your GitHub repos: A secret stealer hit tens of thousands of them
- BleepingComputer: Hackers compromise 4,800 GitHub accounts to steal cloud credentials
- #github
- #supply-chain
- #credential-theft
- #malware
- #cloud-security
Get the day's top AI & tech stories on WhatsApp: Follow the NextFish channel
Get stories like this every morning.
Related
Cybersecurity4 min
ConnectWise urges ScreenConnect client updates as SCKit targets technicians
Microsoft, Huntress and Blackpoint have analysed the toolkit used against MSPs. It is built to jump to technicians, but researchers say they have not seen that work.
Cybersecurity3 min
Landmark Group says Shukran loyalty members' personal data was accessed
Names, contact details and dates of birth were involved, the Dubai retailer says, but not card details, passwords or points. It has not said how many members were affected.
Cybersecurity3 min
Operation Endgame takes down Gootloader, CastleRAT and StealC servers
The Justice Department says Gootloader infected at least 25,000 computers and was used against at least 44 US ransomware victims. Europol counts 235 servers hit in all.
Cybersecurity2 min
TSMC says leaked employee data matches records at a cloud provider
The chipmaker says its own systems show no sign of compromise after a seller offered a 6.9GB file of employee records on hacker forums for three bitcoin.