FortiSandbox Exploited, FortiBleed Hits 86K Firewalls

Three critical FortiSandbox vulnerabilities (CVE-2026-39813, CVE-2026-39808, CVE-2026-25089) are under active exploitation, all rated CVSS 9.1 with no authentication required. Simultaneously, the FortiBleed campaign has compromised over 86,000 Fortinet firewalls across 194 countries through automated credential harvesting. Organizations running FortiSandbox must patch immediately and audit FortiGate credential exposure.

Three FortiSandbox Flaws Under Active Exploitation

Three critical vulnerabilities in Fortinet FortiSandbox — CVE-2026-39813, CVE-2026-39808, and CVE-2026-25089 — are being actively exploited in the wild, according to threat intelligence firm Defused Cyber (as reported by igor’s LAB). All three carry a CVSS 9.1 rating, require no authentication, and allow unauthenticated remote code execution or authentication bypass. Two were patched in April; the third was fixed only on June 9, 2026. Fortinet has not yet confirmed exploitation in its advisories, but independent researchers have observed attack traffic against internet-facing instances across multiple regions.

The Three Vulnerabilities, Decoded

Each flaw targets a different entry point on the FortiSandbox appliance, but shares a common thread: no credentials, no user interaction, no prior privileges needed.

CVE-2026-39813 is a path manipulation vulnerability in the JRPC programming interface. An attacker sends crafted HTTP requests to traverse the filesystem and bypass authentication entirely, escalating privileges on the compromised appliance. Fortinet’s advisory flags this as unauthenticated pre-auth access.

CVE-2026-39808 is a straight OS command injection in an API endpoint. The payload executes arbitrary commands via crafted HTTP requests — no login cookie, no CSRF token, nothing. If your FortiSandbox management interface touches the internet, an attacker can achieve remote code execution in two requests.

CVE-2026-25089, the newest of the trio, targets the web UI’s VNC session launch function. Fortinet’s product security team credited Adham El Karn with the discovery. Insufficient neutralization of JSON inputs enables command injection. This flaw affects FortiSandbox across all deployment models: on-premise, Cloud, and PaaS.

The “Vibecoded” Exploit

Defused Cyber made an unusual observation about CVE-2026-25089: the publicly observed exploit appears to have been generated by an AI model — what they described as “vibecoded”. The code is reportedly faulty and does not achieve reliable exploitation. No working public exploit for CVE-2026-25089 has been confirmed as of this writing.

This pattern of Fortinet appliance targeting is not new — we covered FortiClient EMS exploitation (CVE-2026-35616) in May and VPN zero-day campaigns just weeks ago. The vulnerability is real, the patch exists, and a competent attacker could develop a working version. The two April-patched flaws (CVE-2026-39813 and CVE-2026-39808) are the more immediate danger, as attackers have had over two months to weaponize them against unpatched appliances.

The broader signal is harder to ignore: AI-generated exploit code is now appearing in active campaigns. The barrier to entry for vulnerability exploitation is dropping, and defenders need to assume that flawed first attempts will iterate into functional weapons faster than traditional timelines suggest.

FortiBleed: 86,644 Firewalls Compromised in Parallel

The FortiSandbox exploitation lands against a backdrop of massive Fortinet compromise. SOCRadar disclosed “FortiBleed” on June 16, 2026 — a credential harvesting campaign that has produced a verified database of over 86,644 working credentials for FortiGate firewalls across 194 countries.

The operation, first flagged by security researcher Volodymyr “Bob” Diachenko, involves a Russian-speaking threat group running automated credential stuffing against internet-facing FortiGate SSL VPN appliances. The group processed 1.16 billion credential attempts against 320,777 FortiGate targets. Verified credentials were tested and confirmed through round-the-clock automated tooling. This mirrors the pattern we saw in recent cloud key exposure incidents, where leaked credentials fueled downstream compromise.

Hudson Rock’s analysis puts the number at 73,932 unique firewall URLs compromised, affecting 21,632 domains. The attackers intercept SSL-VPN authentication sessions, crack captured hashes on a 45-GPU cluster managed via Hashtopolis, and pivot into internal Active Directory environments for persistence and lateral movement.

Kevin Beaumont, who independently reviewed the dataset, confirmed the credentials are legitimate. In most cases, the FortiGate management interface was exposed to the internet — a configuration failure that compounds the damage from the initial credential compromise.

Why FortiSandbox Is a High-Value Target

FortiSandbox is not a standalone product in most deployments. It feeds threat verdicts to the broader Fortinet security fabric — FortiGate firewalls, FortiMail, and other products use its analysis to enforce blocking decisions and trigger automated incident response. Help Net Security notes that compromising FortiSandbox does not just give an attacker a single appliance — it corrupts the threat intelligence pipeline that downstream security controls rely on.

An attacker who compromises FortiSandbox can potentially suppress malicious verdicts, allowing malware through the very controls designed to stop it. They can also use the appliance as a pivot point into the broader network, since FortiSandbox typically holds elevated privileges within the security fabric.

What SecOps Teams Must Do Now

1. Patch immediately. All three FortiSandbox CVEs have fixes available. If you run FortiSandbox — on-premise, Cloud, or PaaS — apply the patches today. The April patches for CVE-2026-39813 and CVE-2026-39808 have been out for over two months. There is no excuse.
2. Segment FortiSandbox management. The management interface should never be internet-facing. Restrict access to internal management VLANs with network segmentation and IP allowlisting.
3. Audit FortiGate credential exposure. Against the FortiBleed campaign, check whether your FortiGate management interface or SSL VPN is internet-exposed. If it is, rotate credentials immediately — especially if you reused passwords across devices or failed to rotate after previous Fortinet advisories.
4. Review FortiSandbox verdicts. If your appliance was unpatched during the exploitation window, audit recent threat verdicts for anomalies. Look for files or URLs that received clean verdicts despite known indicators of compromise.
5. Monitor for post-exploitation. FortiBleed actors pivot into Active Directory. Check for unusual authentication patterns, new service accounts, and privileged group membership changes originating from Fortinet infrastructure.

References

• The Hacker News — Attackers Exploit Three Fortinet FortiSandbox Flaws
• igor’s LAB — FortiSandbox Under Attack: Three Critical Vulnerabilities Actively Exploited
• Help Net Security — Attackers Are Exploiting FortiSandbox Vulnerabilities
• SOCRadar — FortiBleed 2026: 86,644 Fortinet Firewalls Compromised
• CyberSecurityNews — FortiBleed: 70,000+ Fortinet Firewalls Compromised
• Arctic Wolf — Active FortiBleed Campaign Impacting Fortinet Devices