Security1 publisher2 min readPublished
Attackers are bypassing authentication on Cisco ISE with a crafted API request
Cisco says CVE-2026-76460 is under active exploitation. It scores 10.0, needs no credentials, and puts an unauthenticated attacker past the web management interface of an Identity Services Engine appliance.
The Watch · Security desk
What happened
- Cisco warned that CVE-2026-76460, a maximum-severity authentication bypass in Identity Services Engine scored 10.0 on CVSS, has come under active exploitation.
- Because the attacker is unauthenticated and remote, exploitation works without credentials or a local foothold.
- In the same week's reporting, Hacktron said it used Claude Opus 5 to chain an OpenAI SSO misconfiguration with a libheif bug in Discourse and reached employee ChatGPT accounts and internal repositories.
Compiled by The WatchSomething wrong?How this is made
Why it matters
- decision A 10.0 with confirmed exploitation leaves nothing for score-based triage to sort, so each ISE operator is deciding how far to pull the maintenance window forward.
- constraint Hardening applied to the ISE management UI does not cover this request path, so the usual admin-interface controls buy no time here.
- exposure Every network segment that can route to an ISE management API is now an administrative path into the appliance. Segmentation of the management plane is what now limits exposure.
- precedent Two identity-layer failures reported in the same week, one in an access-control appliance and one in an SSO deployment, point at attackers working the systems that grant access and using AI tooling to chain what they find.
"This vulnerability is due to insufficient authentication control on an API endpoint," Cisco said [3]. On the exploit path the wording is specific: "An attacker could exploit this vulnerability by sending a crafted request to an affected API endpoint. A successful exploit could allow the attacker to gain unauthorized access to the affected device by bypassing the web-based management interface" [4].
The attack never touches the login page. Restrictions layered on the admin UI sit on the far side of the request the attacker actually sends. The precondition is network reach to the API, and the attacker needs no account [2][4].
There is a second identity-layer failure in the same week's reporting. Hacktron said it used Anthropic's Claude Opus 5 to chain an SSO misconfiguration in OpenAI's identity infrastructure with a libheif remote code execution bug in the Discourse community forum, CVE-2026-32882, to reach OpenAI employees' ChatGPT accounts and then internal OpenAI repositories [6]. That was fixed 14 hours after responsible disclosure [7]. Upstream, the libheif flaw had been closed in version 1.22.0 in May 2026 [8], about four months before the September 2026 report [10]. The forum was running a libheif build whose fix had already shipped.
The recap's item on the Cisco advisory does not list affected ISE releases, fixed software versions, or an attributed actor [5]. Cisco called the flaw maximum-severity and said exploitation is already happening [1]. Until version numbers are in hand, the controllable variable is routing: which subnets, VPN pools and internet-facing addresses can currently open a session to an ISE management API [4].
What to watch
- Cisco publishing fixed ISE release numbers, a workaround, or indicators recovered from the exploited appliances.
- Whether CVE-2026-76460 lands on CISA's KEV catalog with a federal remediation date attached.
- Whether anyone attributes the ISE exploitation to a named cluster, and whether the activity goes past the bypass into post-access use of the appliance.