“FortiBleed,” a mass credential-exposure campaign against internet-facing Fortinet FortiGate firewalls and SSL-VPN gateways, is escalating, and it’s a perimeter-and-identity story, not a single CVE.
Researchers — Arctic Wolf, Recorded Future, Bitsight, and Huntress among them — describe attackers harvesting configuration files from exposed FortiGate devices and cracking the stored credential hashes into working administrator logins. Reported counts of affected devices vary widely by source, roughly 30,000 to 86,000 across about 194 countries, so the headline numbers should be treated as researchers’ stated figures rather than a single confirmed total. The campaign surfaced publicly around June 13 after a researcher found an exposed actor server building a validated-credential database, and CISA issued a hardening alert on June 18 urging immediate session termination and credential resets.
Admin-level control of a network boundary is the worst case: rewriting firewall rules, intercepting VPN traffic, planting backdoor accounts, disabling logging, staging ransomware. This is the identity-and-exposure story in its rawest form — the breach isn’t a novel exploit, it’s unmanaged internet-reachable infrastructure plus reused or exposed credentials.
Sources: CISA, Arctic Wolf, Recorded Future, Bitsight, Huntress, SecurityWeek
Cyberattacks & Breaches
FortiBleed credential-exposure campaign continues to escalate. Attackers extracted configuration files from internet-facing FortiGate devices and cracked stored hashes into verified administrator credentials. Reported scale ranges from roughly 30,000 to 86,000 devices across about 194 countries depending on the researcher; CISA urged impacted customers to terminate sessions and reset all Fortinet VPN and admin passwords immediately.
So what: The through-line this week is exposure an organization already owns, not exotic zero-days. FortiBleed is unmanaged internet-facing perimeter gear plus harvestable credentials; the PTC bug below is a patch-gap; DirtyClone is a latent kernel flaw now made trivially exploitable. None of this is invisible to an organization with a live, tagged inventory of internet-reachable assets and identities — and all of it is catastrophic to one without it.
Vulnerabilities & Exposure
PTC Windchill/FlexPLM RCE added to CISA KEV amid active exploitation. CVE-2026-12569 (CVSS 9.3), an improper-input-validation remote code execution flaw in PTC Windchill PDMLink and FlexPLM, was added to the KEV catalog on June 26 citing active exploitation; attackers are deploying JSP web shells on compromised systems. PTC alerted customers June 17 and shipped patches across supported branches over the following days, then confirmed June 25 it was still seeing heightened threat activity. It’s the first-ever in-the-wild exploitation of a Windchill flaw, and this is enterprise product-data and lifecycle-management software sitting on manufacturing and engineering intellectual property.
So what: The through-line this week is exposure an organization already owns, not exotic zero-days. The PTC bug is a patch-gap failure — fix available June 17, KEV addition June 26 — and none of it is invisible to an organization with a live, tagged inventory of internet-reachable assets.
“DirtyClone” Linux kernel privilege-escalation flaw gets a public exploit walkthrough. CVE-2026-43503 (CVSS 8.8), a DirtyFrag-family local privilege-escalation flaw in the kernel’s __pskb_copy_fclone() path, lets a local user corrupt file-backed page-cache memory via a cloned packet and gain root. It was patched into mainline on May 21, and JFrog Security Research published the first working proof of concept for this variant on June 25. There is no confirmed in-the-wild exploitation yet — the news is weaponization-readiness, not active attack.
So what: DirtyClone is a latent kernel flaw now made trivially exploitable — none of this is invisible to an organization with a live, tagged inventory of internet-reachable assets and identities, and all of it is catastrophic to one without it.
AI Impact on Security
Agentic-SOC research surfaces a cascading-failure risk. Vendor and analyst commentary this week frames the maturing agentic SOC against a sharp counter-finding: in modeled agent networks, failures and poisoned context can propagate faster than traditional incident response can contain them. One widely cited research claim puts a single compromised agent at poisoning 87% of downstream decisions within four hours; treat that specific statistic as a single-source modeling claim, not an established field measurement.
So what: This is the same inversion that ran through the Cordyceps research earlier this week: agentic leverage cuts both ways. The defensible posture isn’t faith in the agents — it’s governance and visibility at the trust boundaries (what each agent may read, reach, and act on), plus cross-agent observability so a poisoned-context cascade is detectable before it spreads.
Regulatory & Policy
BOD 26-04’s first real test date passed. CISA’s Binding Operational Directive 26-04, issued June 10, replaced flat CVSS-score deadlines with a four-variable risk matrix — exposure, known-exploited status, exploit-automation potential, and technical impact — producing 3/14/60-day remediation tiers. The UniFi/Lantronix KEV batch carried a June 26 action date under it. The directive binds federal civilian agencies, but CISA is urging all organizations to adopt the risk-based model.
So what: The federal posture is shifting from patch-everything-by-CVSS to patch-what’s-exposed-and-exploitable-fast, which only works if an organization can continuously see exposure and exploitability across its estate. It’s further proof that a strong data foundation, not compliance paperwork alone, is what makes rapid, risk-based remediation possible.