Cyber & AI intelligence
Wasteland.
Briefs indexed2687
Issues28
Published Mondays07:30 CT
⚡ Active KEV CVE-2026-92787 2026-09-16

CVE-2026-92787: Feast Authentication Bypass via Unverified JWT Signature

"A critical flaw reported in the Feast feature store (versions through 0.66.0) would let unauthenticated attackers forge identity and take control of feature store data by presenting a JWT that the server never…"

A critical flaw reported in the Feast feature store (versions through 0.66.0) would let unauthenticated attackers forge identity and take control of feature store data by presenting a JWT that the server never cryptographically verifies.

What Is It

According to the disclosing party, Feast through 0.66.0 fails to verify JWT token signatures before establishing user identity. An attacker can present an unverified token carrying a hardcoded claim value and bypass role-based access control in the product. The reported result is a trusted internal identity with unchecked read and write access to all entities, feature views, data sources, and permission policies on the server.

One caveat applies to the record itself before any of its contents: the identifier CVE-2026-92787 falls well outside the sequence range CVE Numbering Authorities are issuing for 2026 as of this writing, so the ID as supplied cannot be treated as confirmed. Validate it against the CVE Program record before propagating it into tracking systems, ticketing, or detection content. The technical analysis below describes the reported vulnerability as disclosed, independent of which identifier it is ultimately assigned.

VulnCheck, the disclosing party, classifies the issue as CWE-798 (Use of Hard-coded Credentials) and assigns it a CVSS 3.1 base score of 9.8 (CRITICAL) and a CVSS 4.0 base score of 9.3 (CRITICAL). Note that the vector string published with the record, CVSS:4.0/AV:N/AC:L/PR:N/UI:N/S:U/C:H/I:H/A:H, cannot be reconciled with either score as labeled: it carries a CVSS 4.0 prefix but is composed of CVSS 3.1 metrics, including the Scope metric (S:U), which CVSS 4.0 does not define. The underlying metrics correspond to the 9.8 CVSS 3.1 rating; the 4.0 vector has not been published in a valid form. Treat the numeric scores as vendor-supplied and unverified pending NVD analysis.

Why It Matters

If the reported metrics hold, they describe the worst combination available: network attack vector, low attack complexity, no privileges required, and no user interaction, with high impact to confidentiality, integrity, and availability. On that basis, any Feast feature server reachable by an attacker could be read and rewritten without credentials. Because the reported bypass grants access to permission policies themselves, an attacker would not be limited to reading feature data; they could alter the authorization model of the server.

These impact claims rest on a single vendor advisory. NVD has not completed its analysis, no independent confirmation or public proof-of-concept is cited in the supplied data, and the upstream project has not published a corroborating advisory at the time of writing. The severity may change once the record is analyzed.

This CVE does not appear in the supplied CISA KEV data, so there is no confirmation of active exploitation and no KEV-mandated remediation deadline at this time. Exploit maturity in the CVSS 4.0 assignment is Not Defined.

What's Vulnerable

The disclosure references specific code paths in the OIDC token parser and security manager, along with the feast-feature-server Helm chart deployment template. The affected-version range is vendor-asserted and has not been independently validated.

Patch Status

The supplied NVD record lists no fixed version and no patch advisory. The CVE was published 2026-09-16 with a vulnerability status of "Received," meaning NVD analysis is not yet complete and the record's contents may be revised. Operators should track the upstream GitHub issue and the VulnCheck advisory below for remediation guidance, and, given both the single-source status of the claims and the unconfirmed identifier, validate applicability against their own deployment before acting on the severity rating alone.

Sources