Credential-Theft
Qubole's Trusted AWS Account Turned Against Its Customers: Dormant Cross-Account Roles Used to Loot Secrets and S3
From October 14 to 26, an attacker holding two compromised Qubole IAM users assumed customer cross-account roles, many dormant since spring, and pulled Secrets Manager, SSM, Lambda env vars and S3 data. Expedia has since confirmed a breach through an unnamed data-platform vendor.
Cursor Traces Backdoored Claude Code Extension to a Leaked Fireworks AI Key
Cursor's incident update says a key it issued to inference provider Fireworks AI exposed agent traces with credentials, which let an attacker plant trojanized Claude Code 2.1.312/2.1.313 builds with an SSH reverse-tunnel backdoor in Cursor's extension marketplace.
SCKit: Blank RMM+ Access Key Lets Attackers Push Rogue ScreenConnect Clients to Nearly 500 MSP-Managed Hosts
A third-party ScreenConnect extension, RMM+ 1.0.31-1.0.32, accepted an empty access key. Attackers used it to run SCKit through Run Command on 288 hosts in 90 minutes, then hit 205 more on November 13 from new infrastructure.
wfx-loot: 4,800 Hijacked GitHub Accounts Push a CI Secret Stealer That Deletes Its Own Workflow Runs
A campaign that started November 10 used thousands of compromised GitHub accounts to inject a GitHub Actions secret stealer into tens of thousands of repos. It reverts its commit and deletes the run log within about a minute, so the main trace left is a gap in run numbers.
No Exploit Required: October 2026 Was Four Stolen-Credential Breaches in a Trench Coat
LiteLLM, Gheist, UpdraftCentral, and Railway had no common CVE, but one common root cause: long-lived, over-scoped credentials sitting where attackers could reach them. A trend analysis, and the controls that would have changed each outcome.
MALFEX: Eight Malicious npm Packages Deliver Overlord RAT and a Node.js Stealer to Windows Developers
Checkmarx and CloudSEK detail MALFEX, a long-running npm campaign of eight malicious packages with 40,767 downloads that deliver the Overlord RAT, the movinlike stealer and a hidden downloader.
k0t Campaign Exposes 70,000+ Credentials from FortiGate, FortiWeb and GlobalProtect Edge Devices
Group-IB and NyxLab detail a suspected access-broker campaign against Fortinet and Palo Alto edge devices that leaked 70,000+ credential records, with unconfirmed ties to ransomware crews.
LiteLLM 1.107.0 Backdoored on PyPI and Docker: Python Startup-Hook Stealer Hits AI Gateway Deployments
A malicious LiteLLM 1.107.0 release on PyPI and Docker registries installed a startup-hook credential stealer and an hourly remote code loader. Anything that ran it should be treated as compromised.
Your CI Runner Holds the Keys: A Field Guide to Hardening GitHub Actions After GhostAction, Megalodon, and TanStack
Three different attacks in 2026 reached the same place: a GitHub Actions runner with secrets and an OIDC token in scope. Here's the attacker's checklist and a concrete, ordered hardening plan, with config snippets and detections, for infra teams that own CI.
GhostAction Returns: Stolen PATs Used to Push Malicious GitHub Workflows Across 30,000 Repositories
GitGuardian reports a resurgent GhostAction campaign: about 1,900 hijacked GitHub accounts pushed fake security workflows to harvest CI/CD secrets, with a spike on October 8-9 targeting over 30,000 repositories.
SubQuery's @subql/common Hijacked on npm: Credential Stealer and Reverse Shell in 5.8.3
A malicious @subql/[email protected] on npm steals cloud, SSH, GitHub and Kubernetes credentials, opens a reverse shell, and persists by injecting a rogue GitHub Actions workflow with stolen tokens.
MemTensor MemOS Packages Backdoored with sckit, a Go Credential-Stealing Worm, on npm and PyPI
Malicious releases of MemTensor's MemOS OpenClaw plugin (npm) and MemoryOS (PyPI) deliver sckit, a cross-platform Go implant that steals 13 credential types and carries worm templates for npm, PyPI and GitHub Actions.
Two Malicious CPAN Modules Smuggle a Python Backdoor Inside a Fake Certificate File
Crypt::SelfCertificate and IO::Socket::SSL::SelfCertificate, two CPAN distributions for generating self-signed certs, shipped versions that hide obfuscated Python code inside a sample cert.pem and execute it as a remote-fetch loader.
CVE-2026-0768: Unauthenticated Root RCE in Langflow's Validate Endpoint Under Mass Exploitation
CVE-2026-0768, an unauthenticated code-injection RCE in Langflow's custom component validator, is under active mass exploitation — VulnCheck honeypots logged 360 attacks since August 29 hunting for AWS and OpenAI keys.
Fire Ant Expands From VMware Hypervisors to Cisco IOS XR Routers and TACACS Servers
China-nexus actor Fire Ant has moved beyond VMware ESXi hosts to implant Cisco IOS XR routers and TACACS+ authentication servers, using purpose-built tooling to hijack GRE tunnels, hide commands from admins, and intercept credentials at the network's control plane.
Mini Shai-Hulud "Trinitite": TanStack Codegen Tool Poisoned via Hijacked GitHub OIDC Release Workflow
A new Mini Shai-Hulud wave dubbed Trinitite compromised @7nohe/openapi-react-query-codegen, a 150K-download-a-week TanStack code generator, by hijacking its GitHub Actions OIDC publish workflow rather than stealing a token.
Critical Veeam ONE Flaw Lets Unauthenticated Attackers Coerce SMB Auth From the Service Account
CVE-2026-65641 (CVSS 9.3) lets an unauthenticated network attacker force Veeam ONE's service account into an SMB authentication attempt, exposing Net-NTLM material for relay or offline cracking.
9,300+ Leaked AWS Keys Are Still Active — 768 Give Attackers Full Admin Control
A large-scale scan of public repos, Hugging Face datasets, Docker images, and CI logs found over 9,300 leaked AWS keys still authenticate — 768 with full corporate admin control.
14 Trojanized npm Packages Drop RedC2 4.0 Linux Backdoor With AI-Assisted C2
Fourteen npm packages disguised as calendar and streak utilities smuggle in RedC2 4.0, a commercial Linux implant whose LLM-driven operator console turns plain-English prompts into post-exploitation commands.
Phishing the Protocol: How 2026 Attackers Made MFA Irrelevant
Device code grants, app passwords, OAuth consent screens, and WhatsApp device linking all share one property: they're real login flows, not bugs. 2026's biggest identity attacks stopped stealing passwords and started collecting the tokens MFA can't protect.