← Back to Blog

Security Briefing, May 7 2026: PAN-OS Zero-Day, vm2 Sandbox Escape, MuddyWater False Flag, and Microsoft AiTM Phishing

This morning’s pattern is ugly and consistent: attackers are winning by abusing the trusted layer around the system — the firewall portal, the sandbox wrapper, the collaboration workflow, and the sign-in flow users think they recognize.

---

---

1. PAN-OS Captive Portal is the highest-priority internet edge problem today

Palo Alto says CVE-2026-0300 is a critical buffer overflow in the User-ID Authentication Portal that allows unauthenticated root-level code execution on exposed PA-Series and VM-Series firewalls. Limited exploitation is already underway, and both vendor guidance and outside reporting point to one immediate truth: if Captive Portal is reachable from untrusted IP space, the risk is live now.

• Restrict Captive Portal to trusted internal networks or disable it until patches land.
• Inventory exposed PA and VM appliances now, especially internet-facing portals.
• Do not wait for KEV or mass scanning headlines before treating this as an incident-grade exposure.

---

2. vm2 proves again that “sandboxed” JavaScript is not a security boundary by itself

The vm2 bug tracked as CVE-2026-26956 lets attackers escape the sandbox and execute code on the host, with a public proof of concept already available. The issue hits Node.js 25 environments using specific runtime features, but the bigger lesson is broader: if your product runs customer-supplied JavaScript, the wrapper is part of the blast radius.

• Upgrade vm2 beyond 3.10.4 immediately where untrusted code execution exists.
• Review SaaS, automation, and coding platforms that rely on vm2 for isolation assumptions.
• Re-check host-level containment, not just library versioning, in multi-tenant execution paths.

---

3. MuddyWater is using ransomware theater to hide espionage work

Rapid7-linked reporting says the Iranian group MuddyWater used Microsoft Teams social engineering, screen sharing, credential capture, AnyDesk, DWAgent, and extortion messaging while never actually deploying file-encrypting ransomware. Chaos branding was the costume, but the underlying operation looked like access, persistence, and data theft.

• Re-examine “ransomware” incidents that lacked encryption but included remote management tooling and mailbox pressure.
• Treat Teams-driven help-desk style outreach as a credential and MFA manipulation vector.
• Hunt for persistence left behind after the extortion narrative steals defender attention.

---

4. Microsoft’s conduct-review phishing wave is built to steal tokens in real time

Microsoft observed more than 35,000 attempts across roughly 13,000 organizations using fake internal conduct notices, PDF lures, Cloudflare CAPTCHA gates, and adversary-in-the-middle pages that proxy Microsoft sign-in. That matters because AiTM phishing beats weak MFA by capturing session tokens, not just passwords.

• Warn users that compliance-themed PDFs and “review case materials” flows are active lures right now.
• Push phishing-resistant MFA where possible; basic MFA is not enough against live token interception.
• Hunt for suspicious sign-ins tied to cloud email senders, CAPTCHA-gated flows, and short-lived token abuse.

---

What security teams should do before lunch

1. Close or restrict exposed PAN-OS Captive Portal surfaces first.
2. Patch vm2 and review every path where customers or staff can run JavaScript on shared hosts.
3. Escalate Teams-based remote-support social engineering into the same playbook as email phishing and vishing.
4. Update phishing response to prioritize token revocation and session review, not only password reset.

---

---

Bottom line: the pattern today is trust inversion. Internet-facing portals, sandbox runtimes, collaboration tools, and familiar sign-in pages all become attack surfaces when teams assume the wrapper is the control.