Massive Supply Chain Attack Targets Tens of Thousands of GitHub Repositories

In a significant escalation of software supply chain threats, security researchers at Socket have uncovered a sophisticated and wide-reaching malicious campaign. The attack has successfully compromised tens of thousands of GitHub repositories, including those owned by major technology organizations. Among the high-profile entities affected is Uber, which researchers identified as running compromised packages within their environment. This incident marks a dangerous evolution in the tactics used by threat actors to exfiltrate sensitive data, bearing a striking resemblance to the "GhostAction" campaign that first surfaced in September 2025 and saw a resurgence in October 2026.

The scale of this latest breach highlights the inherent vulnerabilities in modern automated development environments. By exploiting the trust placed in GitHub Actions—the platform’s native automation tool for CI/CD (Continuous Integration and Continuous Deployment) pipelines—the attackers have managed to bypass standard security scrutiny, effectively weaponizing the very tools designed to streamline software delivery.

Malicious Workflows Spread via Compromised GitHub Accounts

The campaign was set into motion on October 8, 2026, when two specific GitHub accounts, identified by the usernames "henrywoo" and "kitao," began pushing a malicious GitHub Actions workflow to hundreds of repositories for which they held write permissions. GitHub Actions are essential to the modern developer workflow, acting as automated scripts that trigger specific tasks—such as code testing, software building, or application deployment—whenever certain events, like a code push or a pull request, occur within a repository.

The malicious workflow, masquerading under the innocuous filename ".github/workflows/security-audit.yml," was introduced via single-file commits. To avoid raising suspicion among repository maintainers or automated security monitors, the commits were labeled with deceptive messages such as "Add security audit workflow" and "Update security audit workflow." These titles were carefully chosen to appear as legitimate efforts to improve the security posture of the affected projects.

Despite the name, the workflow performs no genuine security auditing functions. Instead, it serves as a sophisticated exfiltration mechanism. Once triggered, the script is designed to harvest credentials from the affected repositories and transmit them to an external, attacker-controlled server located at the IP address 193.32.204[.]199. Communication with this server occurs over standard HTTP, allowing the data to be siphoned away from the internal environment. Because the workflow mimics the appearance of a standard security utility, it is frequently overlooked by developers and automated security scanners, allowing it to persist within the repository’s background processes indefinitely.

Socket researchers observed that the malicious file had been distributed across an staggering number of repositories. The infection was not limited to active or high-traffic projects; even repositories that had remained dormant for over a decade were targeted. This systemic approach suggests that the attackers utilized an automated script to enumerate and exploit every repository to which the compromised accounts had access, rather than manually selecting high-value targets. By casting such a wide net, the attackers maximized their chances of intercepting valuable credentials, regardless of the perceived importance of the individual project.

Expanding the Scope of the GhostAction Threat

This campaign represents a concerning evolution of the GhostAction threat vector. While earlier iterations of GhostAction primarily focused on stealing secrets stored specifically within GitHub Actions configuration—such as API keys, environment variables, and access tokens intended for automated workflows—the current variant has significantly broader capabilities.

Socket researchers have noted that this new iteration employs more aggressive data-scraping techniques. The malicious workflow does not merely look for stored configuration secrets; it actively parses the repository’s entire file structure and deep-dives into the comprehensive Git history. By utilizing regular-expression patterns, the script hunts for sensitive information that may have been committed to the source code, buried in configuration files, or hidden in older, long-forgotten versions of project files.

This is a critical distinction in the attack’s severity. Even if a developer has successfully scrubbed a sensitive credential from the current version of the code, the attacker’s script can retrieve that credential from the repository’s commit history. This "time-traveling" capability exposes credentials for a vast array of services, including cloud infrastructure providers and third-party AI service APIs. The potential for damage is immense, as stolen credentials could grant the attackers unauthorized access to production infrastructure, hosted customer applications, and proprietary third-party services that are increasingly critical to the functionality of modern software.

Persistence and Future Implications

As of October 9, 2026, the malicious workflow remained active on the default branches of several repositories, including high-profile examples like "uber/athenadriver" and "kitao/pyxel." Socket’s analysis confirmed that the workflow was not merely present but had successfully executed, proving that the credential-stealing phase of the attack was fully operational across these targets.

While the discovery of this campaign has sent ripples through the cybersecurity community, Socket researchers noted that, at the time of their report, they had not observed any evidence that the stolen credentials had been used to publish malicious package versions to major registries like PyPI or crates.io. However, this absence of immediate secondary activity should not be interpreted as a sign that the stolen information remains unused. Instead, it indicates that the attackers may be focusing on different goals, such as long-term persistence within target infrastructure, the exploitation of cloud environments, or the sale of credentials on underground marketplaces.

The situation underscores the persistent risk posed by supply chain attacks, where the compromise of a single contributor account or a trusted automation tool can have cascading effects across thousands of disparate projects. The automated, systematic nature of this attack, combined with its ability to scrape deep into repository history, makes it a formidable challenge for organizations to detect.

As the industry grapples with these findings, the incident serves as a stark reminder of the necessity for rigorous access control, the regular rotation of all credentials, and the implementation of automated secret scanning to prevent sensitive information from ever entering a repository. For organizations that rely heavily on GitHub Actions, the audit of existing workflows and the restriction of contributor permissions have become more vital than ever. The resilience of the software supply chain now depends on the ability of maintainers and security teams to proactively identify and neutralize these "ghost" processes before they can effectively weaponize the codebases they were meant to protect.

The ongoing presence of the malicious workflow in major repositories highlights the urgency of remediation efforts. As researchers continue to monitor the situation, the focus remains on identifying the full extent of the exposure and ensuring that organizations take the necessary steps to invalidate any credentials that may have been harvested by this sophisticated campaign. With the threat landscape constantly shifting, the GhostAction campaign remains a potent example of how automated workflows, when left unchecked, can be turned against their owners.