🛡️ CISO Intel — Wednesday, 29-04-2026

Due to API rate limiting, here is today’s briefing:

🔴 Critical Threats & Active Exploitation

GitHub.com / GitHub Enterprise Server — CVE-2026-3854 (CVSS 8.7) — A critical RCE vulnerability in GitHub’s internal Git infrastructure. An authenticated user can achieve remote code execution on backend servers with a single git push command by exploiting an injection flaw in user-supplied push options within internal service headers. This allowed cross-tenant access to millions of repositories on GitHub.com and full server compromise on GitHub Enterprise Server. GitHub has patched GitHub.com and released patches for GHES. Patch immediately.

VECT 2.0 Ransomware — This RaaS, active since December 2025, has a critical flaw in its encryption across Windows, Linux, and ESXi variants. Files larger than 131KB are permanently destroyed, not encrypted, due to discarded decryption nonces. Paying the ransom is pointless; recovery is impossible. This is a wiper masquerading as ransomware.

French Identity Document Agency (ANTS) — A massive data breach confirmed by France’s national identity document agency (ANTS) affecting 11.7 million accounts. Threat actor “breach3d” exploited an “elementary” Insecure Direct Object Reference (IDOR) flaw on the ANTS API. Stolen data includes full names, dates/places of birth, email addresses, home addresses, and phone numbers. Scanned identity documents were not affected.

🛡️ CVEs Worth Your Attention

⚡ New TTPs & Attack Research

UNC6692’s Microsoft Teams Social Engineering — Mandiant reports on UNC6692, a threat group using email flooding to overwhelm inboxes, then impersonating Microsoft IT support via external Microsoft Teams accounts. They trick victims into installing a malicious browser extension, SnowBelt, which provides persistent access and allows lateral movement without repeated authentication. This is T1566.002 (Spearphishing Link) combined with T1189 (Drive-by Compromise) and T1133 (External Remote Services) for persistence.

GopherWhisper APT’s Legitimate Service Abuse — A new China-linked APT, GopherWhisper, uses Go-based backdoors (LaxGopher, RatGopher, BoxOfFriends) that leverage legitimate services like Slack, Discord, file.io, and Microsoft Graph for C2 and data exfiltration. They also use custom loaders/injectors (JabGopher, FriendDelivery) for stealthy in-memory malware execution. This is T1102 (Web Service) and T1574 (Hijack Execution Flow) with T1055 (Process Injection).

LLM Prompt Injection Deep Dive — New research highlights prompt injection as a semantic attack, not a code execution one. It exploits LLMs treating all text as instructions, leading to data exfiltration, model hijacking, privilege escalation, and jailbreaking. Key takeaway: “prompt primacy” where later instructions override earlier ones. Defenses need explicit boundaries (structured formats), capability binding, and continuous validation.

🏗️ DevSecOps & Cloud Security

GitHub Actions Workflow Flaw & Supply Chain Attack — A critical flaw in a GitHub Actions workflow led to the compromise of the element-data open-source Python package (1M+ monthly downloads). Attackers used a malicious pull request to run a bash script, extracting tokens and signing keys to publish a malicious version (0.23.3) to PyPI and Docker. This version harvested cloud keys, API tokens, SSH keys, and .env secrets. This is a classic supply chain attack (T1195.002, T1588.006) leveraging CI/CD vulnerabilities.

Kubernetes NGINX Ingress Controller Vulnerabilities — Several security issues, including a critical one (CVE-2025-1974), discovered in the NGINX Ingress Controller (ingress-nginx). If you’re running multi-tenant workloads on GKE or GDC (VMware), pay attention. Privilege escalation vulnerabilities (CVE-2026-23209, CVE-2026-23074) also found in the Linux kernel affecting Container-Optimized OS nodes.

“Slopsquatting” via AI Coding Tools — A new supply chain attack vector where AI coding tools hallucinate non-existent package names. Attackers pre-register these hallucinated names, and developers copy-paste them into their projects, leading to malware installation. This bypasses traditional typosquatting, as the error originates from the AI, not human fat-fingering. This is a novel variant of T1195.002 (Supply Chain Compromise: Compromise Software Dependencies and Development Tools).

🔧 Patches & Vendor Releases

GitHub — Patches released for GitHub Enterprise Server to address CVE-2026-3854. GitHub.com was patched within 6 hours of discovery. Upgrade GHES immediately. 🟢 solid fix

Red Hat Enterprise Linux 9 — Update available for LibRaw addressing CVE-2026-24450 and CVE-2026-21413, both leading to arbitrary code execution. 🟢 solid fix

Mozilla Firefox/ESR — Security advisories published for Firefox versions prior to 150.0.1, Firefox ESR versions prior to 140.10.1 and 115.35.1. Fixes information disclosure and memory safety bugs. 🟢 solid fix

Spring gRPC — Fixes available for Spring gRPC versions 1.0.0 - 1.0.2 to prevent AuthenticationException messages from being reflected to unauthenticated clients (CVE-2026-40969). 🟢 solid fix

🧪 Threat Intel & Malware

M3RXDLS Ransomware Spike — M3RXDLS was the most active ransomware group in the last 24 hours, hitting five new victims across Switzerland and Australia. Brain Cipher and Medusa also claimed victims.

Ransomware Group Infighting (0APT vs. KryBit) — Two ransomware groups, 0APT and KryBit, engaged in a “turf war,” leaking each other’s operational data. KryBit hacked back at 0APT, defacing its leak site and exposing its full operational dataset, including access logs, PHP source code, and system files. This revealed 0APT’s 190+ claimed victims in January 2026 were fabricated. KryBit’s own leaked admin panel showed 2 admins, 5 affiliates, and 20 potential victims with ransom demands between $40k-$100k.

North Korean Hackers Target Web3 — North Korean hackers are reportedly using new attack chains targeting Web3. This is a consistent pattern for DPRK groups, focusing on cryptocurrency for funding.

🌐 Industry, Brand & Internet Security

Vimeo Data Breach — Video hosting platform Vimeo confirmed a data breach stemming from a compromised third-party vendor (Anodot analytics platform). Hackers accessed databases with technical data, video titles, metadata, and email addresses of some customers. No video content, login credentials, or payment card information was exposed. ShinyHunters group claims responsibility, threatening to leak data from Vimeo’s Snowflake and BigQuery instances if ransom isn’t paid by April 30.

Itron Cybersecurity Breach — Energy and water management firm Itron confirmed a cybersecurity breach where hackers accessed internal systems. No evidence of customer data impact or disruption to customer-hosted systems, but investigations are ongoing. This highlights critical infrastructure technology firms as ongoing targets.

Intuit’s Reverse Domain Name Hijacking Attempt — QuickBooks maker Intuit attempted to reverse hijack the domain QB.com via a UDRP panel after failing to purchase it for $2.5 million. The panel ruled against Intuit, citing a “Plan B reverse domain name hijacking” attempt. This underscores how large companies sometimes abuse legal processes when commercial acquisition fails.

Binance Brand Impersonation Phishing Site — binance-event.com identified as an active brand impersonation phishing site targeting Binance users. The domain was registered through a “bulletproof” Estonian registrar and uses an invalid SSL certificate. It’s flagged by multiple security vendors and categorized by Google Safe Browsing as SOCIAL_ENGINEERING.

Rise of Third-Party & Dormant Domain Abuse — A report, “Domain Security Report 2026,” highlights a significant rise in phishing and brand abuse fueled by third-party and dormant domains. 88% of homoglyph domains are third-party owned. Inactive domains with MX records are weaponized for brand impersonation and phishing. The report blames the lack of transparent WHOIS data (due to ICANN actions) for enabling malicious actors.

🤖 AI & LLM Security

LLM Red-Teaming Dashboard — A new open-source project, Infiltrate AI, was announced, designed as an LLM Red-Teaming Dashboard to automate prompt injection testing and security monitoring. It runs various injection attacks and provides real-time monitoring of model robustness. This is a defensive tool to address T0865 (Prompt Injection).

AI Security Misconceptions — Mandiant and NTT Data warn that AI adoption is reviving old security gaps. Key misconceptions include: model security equals system security, prompt injection is “just another input problem” (it’s not SQLi), AI outputs are harmless text, external data always improves reliability, and managed AI means the provider handles all security. The real risks lie in the seams between components, data flows, and how AI agents interact with broader systems.

📋 Compliance & Regulatory

HIPAA Security Rule at 21 Years — The HIPAA Security Rule is entering a “more prescriptive era” as it turns 21. Risk analysis, third-party oversight, incident response, and workforce training remain critical. This signals increased regulatory expectations for healthcare organizations.

NIST Guidance on Verifiable Digital Credentials — NIST published guidance on “From DMV to Wallet: Understanding Verifiable Digital Credential Issuance.” This covers the security controls and issuance practices for mobile driver’s licenses (mDLs) and other verifiable credentials, building on standards like OAuth 2.0 and OpenID Connect. This is about establishing trust in digital identity.

CISA Advisory on SOHO Router Botnets — CISA released an advisory on defending against botnets built from compromised consumer and SOHO routers. While not new controls, it emphasizes foundational security best practices like inventorying edge devices, baselining network behavior, identifying anomalous connections, using threat intel, and implementing MFA for remote access. The advisory links SOHO router supply chain risk to critical infrastructure.

💡 Marcus’s Take

We’re seeing a convergence of classic and novel attack vectors. The GitHub RCE is a stark reminder that even core development infrastructure has critical flaws, and supply chain attacks remain a primary threat, extending into AI’s growing ecosystem with “slopsquatting.” The VECT ransomware acting as a wiper is just cruel; it highlights the importance of offline backups and tested recovery over negotiation. Finally, the Mandiant report on UNC6692’s Teams impersonation and the ANTS IDOR breach scream about basic security hygiene: validate external communications, scrutinize third-party access, and fix the “really stupid” API flaws before they become front-page news. Don’t get distracted by the shiny new AI attacks; the fundamentals are still where most of us get hit.

CISO Intel Brief — Wednesday, 29-04-2026