wfx-loot Operation: Attacker Probes 60,000 GitHub Repositories for Credential Theft

Attacker Hijacks 4,837 GitHub Accounts to Deploy CI Secret Stealer Across 1,806 Repositories

An attacker leveraged stolen credentials from 4,837 GitHub accounts to insert a sophisticated secret-stealing workflow into at least 1,806 repositories. The campaign began with test pull requests titled "docs: dependency audit notes" from branches named wfx-docs, targeting 61,764 repositories before deploying the stealer in successful cases. Aikido Security and SafeDep researchers documented the use of both workflow modifications and npm postinstall hooks, with the malware exfiltrating GitHub tokens, cloud credentials, and repository data to attacker-controlled servers and Telegram. The stealer also reached several open-source packages, including a malicious Packagist release and two Go modules, though package users were not always impacted. Overlaps with the earlier GhostAction campaign suggest credential reuse across multiple threat actors. Repository owners are advised to revoke tokens, rotate all accessible secrets, and monitor for specific indicators such as wfx- prefixed federated credentials in Azure and AWS.

ConnectWise Investigates Active ScreenConnect Attacks Breaching MSPs and Impacting Dozens of Organizations

ConnectWise is investigating an active malware campaign targeting managed service providers that uses compromised ScreenConnect environments to deploy rogue remote-access clients to customer devices. Huntress discovered the campaign on November 11, 2026, identifying six compromised MSPs and 286 affected endpoints across 47 organizations, with numbers continuing to rise. Attackers leverage ScreenConnect's Run Command functionality to install unauthorized clients, enabling remote access outside legitimate MSP control, and may also compromise technician workstations through reverse-pivot techniques during sessions with infected devices. ConnectWise has advised partners to avoid interactive sessions with suspected compromised devices, review account activity, revoke sessions, reset credentials, and verify multifactor authentication. Huntress provided indicators of compromise including attacker relay IP 185.99.255.28, C2 domains, rogue client IDs such as 2588ef5c15954b10, and files like Panel-Lateral.ps1. Earlier malicious activity linked to one rogue client ID was identified in a July 2026 batch script impersonating a Social Security Administration statement.

Threat Actors Compromise Qubole IAM Credentials to Breach Multiple Customer AWS Environments

A threat actor compromised two IAM users belonging to the Qubole big data platform and abused cross-account trust relationships to access customer AWS environments. The campaign ran from October 14 to 26 and impacted dozens of organizations that had previously configured IAM roles to trust Qubole’s AWS account. Attackers used the stolen credentials to assume roles, enumerate S3 buckets, retrieve secrets from AWS Secrets Manager and SSM Parameter Store, and exfiltrate data opportunistically. Researchers from Wiz, working with Sentra and Mandiant, traced the activity after observing suspicious behavior in customer environments. The attackers employed TruffleHog to validate discovered credentials including AWS keys, Microsoft Entra ID client secrets, and Okta client credentials. No public statement has been issued by Qubole, and the exact initial access vector remains unknown. Administrators are advised to review CloudTrail logs, rotate credentials, and remove unused cross-account roles.