🚨 ACTIVE EXPLOITATION
- Langflow authenticated RCE flaw CVE-2026-105740 hit by 1,423 honeypot exploit attempts
Threat actors are actively exploiting a critical authenticated RCE vulnerability (CVE-2026-105740) in Langflow, with 1,423 attempts recorded against honeypot sensors.- Applies to Langflow, IBM's open-source platform for building AI applications and agentic workflows, versions prior to 1.9.0.
- CVE-2026-105740 (CVSS 9.9) lets any authenticated user, including non-admins, execute arbitrary OS commands by adding an MCP server with 'stdio' transport; the command field is passed to bash -c with no validation, allowlisting, or sandboxing.
- Attack flow: log in as a regular user, POST a malicious command to /api/v2/mcp/servers/{server_name}, then trigger execution by fetching /api/v2/mcp/servers?action_count=true; the config persists and re-executes on each server list load.
- Attackers authenticated via /api/v1/auto_login attempts and common default credentials, then used the bearer token against the MCP endpoint; captured payloads stole LANGFLOW_SECRET_KEY and LANGFLOW_SUPERUSER_PASSWORD to 146.70.211[.]112:18764 and probed /etc/passwd.
- 1,423 attempts since September 19, 2026 from 27 IPs across 15 countries; top source 192.255.229[.]92 also linked to brute-force and exploitation of Flowise CVE-2026-69255 and Langflow CVE-2026-55255, CVE-2026-9198, and CVE-2026-0768.
💥 BREACHES & INCIDENTS
UPDATE: CenterPoint Energy data breach impacts over 1 million Texans
CenterPoint Energy has disclosed a data breach affecting more than 1 million Texans after a hacker leaked stolen customer records.- CenterPoint Energy, a Houston-based utility serving ~7 million metered customers in Texas, Indiana, Minnesota, and Ohio, confirmed customer data was stolen from an external-facing system.
- Compromised data includes names, addresses, Social Security numbers, and driver's license numbers; the hacker claims 7.49 million records including phone numbers, account numbers, and billing details.
- The attacker, using alias '4d722e4d656f77', said they exfiltrated data by iterating through millions of IDs on CenterPoint's public API, which lacked rate limiting and WAF protection.
- CenterPoint learned of the breach from the hacker's cybercrime forum post; a 2.5 GB archive with ~3 million phone numbers was released on September 12, and the breach was disclosed in a September 14, 2026 SEC filing.
- 1,062,060 Texas residents were reported affected; lawsuits allege the breach occurred between August 17 and September 1, and utility operations were not impacted.
UK Legal Aid Agency: April 2025 hackers stole records dating back to 1978
The UK Legal Aid Agency says hackers who breached its systems in April 2025 stole records dating back nearly five decades, to 1978.- Affects legal aid applicants and providers in England and Wales; the LAA oversees ~2,000 providers and the Public Defender Service
- Attackers breached LAA online systems on 23 April 2025 and stole ~2.1 million sensitive records, including contact details, dates of birth, National Insurance numbers, criminal history, and financial data
- ~250,000 stolen records predate 2007, with the earliest verified record from 1978; most data covers January 2007 to 16 May 2025
- Provider data was also taken, including contracts, financial information, and LAA Online Portal credentials
- Scattered Spider has claimed the breach; three law firms (KP Law, Express Solicitors) are pursuing group action claims
- UK is investing over £200 million in modernising legal aid digital services; no evidence of data publication so far
🔓 CVEs & KEV
CVE-2026-19863 — Eclipse Foundation org.eclipse.jetty:jetty-ethereum — CVSS 9.3 — An unauthenticated client may access certain protected resources when using E...
CVE-2026-50619 — QNAP Systems Inc. Qsirch — CVSS 9.3 — Unauthenticated remote code execution vulnerabilityA path traversal vulnerabi...
CVE-2026-47786 — QNAP Systems Inc. HybridDesk Station — CVSS 9.3 — Pre-Authentication Command InjectionAn OS command injection vulnerability has...
CVE-2026-39997 — QNAP Systems Inc. HybridDesk Station — CVSS 8.9 — Unauthenticated Remote Input Injection in QNAP HD Station's remotepadA missin...
CVE-2026-18660 — Eclipse Foundation org.eclipse.jetty:jetty-server — CVSS 8.8 — Jetty proxy blacklist performs case-sensitive matching of case-insensitive DN...
CVE-2026-34008 — QNAP Systems Inc. Notes Station 3 — CVSS 8.5 — Privilege Escalation World-writable crontab path monitored and executed b...
CVE-2026-34009 — QNAP Systems Inc. Notes Station 3 — CVSS 8.4 — Container Escape Host user home directory mounted with write access into ...
CVE-2026-70111 — Microsoft Windows 10 Version 1607 — CVSS 8.1 — Microsoft Windows Server 0-Day Flaw Lets Attackers Crash Servers With a Single Packet
CVE-2026-18659 — Eclipse Foundation org.eclipse.jetty.ee11:jetty-ee11-servlet — CVSS 6.9 — Unauthenticated authorization bypass in ConstraintSecurityHandler - HTTP meth...
CVE-2026-18662 — Eclipse Foundation org.eclipse.jetty.ee11:jetty-ee11-servlet — CVSS 6.9 — In Eclipse Jetty, across various versions of
jetty-security, and `jetty-eeN...CVE-2026-47790 — QNAP Systems Inc. QuFTP Service — CVSS 6.3 — Privilege Escalation via QuFTP Log-based XSSA cross-site scripting (XSS) vuln...
🕵️ RESEARCH & DEEP DIVES
GhostAction campaign compromises ~1,900 GitHub accounts, targets 30,000+ repositories in October 2026 surge
The GhostAction credential-theft campaign surged in October 2026, compromising about 1,900 GitHub accounts to inject secret-stealing workflows into over 30,000 repositories.- Affects GitHub users and organizations using GitHub Actions CI/CD, including high-profile projects like kitao/pyxel (18,400 stars) and uber/athenadriver
- Attackers use stolen personal access tokens from infostealer logs to hijack maintainer accounts and push malicious workflows named security-audit.yml or github_actions_security.yml
- The workflows exfiltrate named GitHub Actions secrets plus cloud, AI, and SaaS credentials found in the working tree and full git history via curl POST over plain HTTP
- Payload evolved through three versions adding git history secret-scanning with 11 regexes covering AWS key IDs, GitHub/GitLab tokens, and AI secrets
- Data is exfiltrated to 193.32.204.199 (endpoints ?c=bling, ?c=blingnew, batch labels monami/mnsft/new/security); GitHub blocked workflow execution starting October 9, 2026
Zero-click spyware 'AdFection' delivered via targeted ads, Citizen Lab says
Citizen Lab researchers revealed the first technical demonstration of mercenary spyware delivered through zero-click targeted online advertising.- Citizen Lab and ad security firm Confiant revealed AdFection, the first documented zero-click spyware infections delivered via targeted online ads.
- Israeli firm Discover-Tech, posing as a legitimate ad company, facilitated mercenary spyware vendors including Candiru, Bindecy and Intelos via its Controlium platform, which de-anonymizes, fingerprints and infects devices when an ad renders.
- At least 45 Al Jazeera devices were infected with Candiru spyware linked with high confidence to Discover-Tech ads; ads also appeared on networks at the US Department of Defense, NATO, the IAEA and cybersecurity agencies.
- Between March and June 2026, Discover-Tech ads reached over 120,000 websites and over 70,000 mobile apps; Google and IAB Europe have since removed the firm from their vendor lists.
- A separate actor, DARK CARTOGRAPHER, used Discover-Tech infrastructure to exploit mobile carrier vulnerabilities and link advertising IDs to phone numbers; infrastructure overlaps with a US intelligence contractor.
Sources 📄 Source: citizenlab.ca · 📎 Coverage: cyberscoop.com · 👁 via CyberScoop, @campuscodi@mastodon.socialPoC released for exploited FortiClient EMS auth bypass CVE-2026-35616; EKZ infostealer attacks observed
CVE-2026-35616
A public PoC exists for the exploited FortiClient EMS authentication bypass CVE-2026-35616, which attackers have used to deploy the EKZ infostealer.- Applies to organizations running FortiClient EMS 7.4.5 through 7.4.6; versions 7.2, 8.0, FortiClient Cloud, and FortiSASE are not affected.
- CVE-2026-35616 is an improper access control flaw (CWE-284) allowing unauthenticated attackers to bypass EMS API protections and execute unauthorized code via crafted requests.
- Fortinet has confirmed exploitation in the wild; Arctic Wolf observed a threat cluster exploiting it to deploy the EKZ infostealer disguised as a Fortinet patch on EMS-managed endpoints.
- A Python PoC published on GitHub by Alaatk provides a command-line tool that takes a target host and optional port to test for unauthenticated