CVE-2026-91926 Overview
CVE-2026-91926 is a memory leak vulnerability in gss-ntlmssp, an open-source GSSAPI mechanism implementing NTLM authentication used on Linux systems. The flaw resides in the NTLM target-info parser, which processes AV_PAIR entries within NTLM CHALLENGE messages. When a crafted CHALLENGE message contains duplicated string-valued AV_PAIR entries, the parser allocates memory for each value without freeing prior allocations of the same type. A malicious or man-in-the-middle server can send repeated authentication challenges to gradually exhaust client memory, resulting in denial of service. The weakness is tracked as [CWE-401] Missing Release of Memory After Effective Lifetime.
Critical Impact
A remote server can trigger gradual memory exhaustion on gss-ntlmssp clients during NTLM authentication, degrading availability of dependent services.
Affected Products
- gss-ntlmssp GSSAPI NTLM mechanism library
- Red Hat Enterprise Linux distributions bundling gss-ntlmssp (per Red Hat advisory)
- Applications and services using gss-ntlmssp for client-side NTLM authentication
Discovery Timeline
- 2026-09-15 - CVE-2026-91926 published to the National Vulnerability Database
- 2026-09-17 - Last updated in NVD database
Technical Details for CVE-2026-91926
Vulnerability Analysis
The NTLM authentication exchange uses a CHALLENGE_MESSAGE sent from server to client. That message includes a TargetInfo field composed of a sequence of AV_PAIR structures. Each AV_PAIR has a type identifier and a value, where several types carry string values such as the NetBIOS computer name, DNS domain name, and target name.
The gss-ntlmssp parser iterates through the AV_PAIR list and allocates memory for each string value it decodes. When two or more pairs share the same string-valued type, the parser overwrites the destination pointer with a new allocation without releasing the previous one. The earlier allocation becomes unreachable and leaks for the lifetime of the process.
Exploitation requires a client to initiate NTLM authentication against an attacker-controlled or intercepted server. Each authentication attempt can leak multiple allocations. Repeated attempts against long-running clients cause cumulative memory growth that eventually degrades or halts service.
Root Cause
The root cause is a missing free operation in the target-info parsing loop. The code assumes each AV_PAIR type appears at most once and therefore does not check whether the destination field has been assigned before overwriting it.
Attack Vector
The attack vector is network-based and unauthenticated. An adversary controlling or proxying an SMB, HTTP, LDAP, or other NTLM-negotiating server crafts a CHALLENGE message with duplicated string AV_PAIR entries, then forces repeated authentication attempts from the client.
No verified public exploit code is available for CVE-2026-91926. Refer to the Red Hat CVE-2026-91926 Advisory and Red Hat Bug Report #2533698 for technical details.
Detection Methods for CVE-2026-91926
Indicators of Compromise
- Sustained memory growth (RSS) in processes linked against libgssapi using the gss-ntlmssp mechanism, correlated with NTLM authentication activity.
- NTLM CHALLENGE messages containing duplicate AV_PAIR type identifiers observed on the wire.
- Repeated NTLM handshakes originating from a single client toward an unexpected or newly observed server.
Detection Strategies
- Inspect SMB, HTTP, and LDAP traffic for NTLMSSP CHALLENGE messages containing duplicate MsvAvNbComputerName, MsvAvDnsDomainName, or MsvAvTargetName entries.
- Baseline memory usage of NTLM client processes and alert on abnormal linear growth over hours or days.
- Correlate outbound NTLM authentication to untrusted external hosts using proxy and network telemetry.
Monitoring Recommendations
- Ingest process memory, NTLM authentication, and network flow telemetry into a centralized data lake for cross-source correlation.
- Alert when NTLM negotiations target servers outside the enterprise identity perimeter.
- Track service restart frequency for daemons that perform outbound NTLM, as unexplained restarts may indicate exhaustion.
How to Mitigate CVE-2026-91926
Immediate Actions Required
- Apply updated gss-ntlmssp packages from your Linux distribution once available; monitor the Red Hat CVE-2026-91926 Advisory for fixed versions.
- Inventory hosts and applications that link against gss-ntlmssp and prioritize patching for long-running services.
- Restrict outbound NTLM authentication to trusted internal servers using host and network firewalls.
Patch Information
Red Hat has published tracking under Bug Report #2533698. Consult your distribution's package repository for the fixed gss-ntlmssp build addressing the missing free in the target-info parser.
Workarounds
- Prefer Kerberos over NTLM where possible by adjusting GSSAPI mechanism preferences in application configuration.
- Terminate outbound NTLM sessions at an authenticated proxy that validates CHALLENGE message structure.
- Configure service supervisors to enforce memory limits and automatic restart for processes performing NTLM authentication until patches are deployed.
# Example: constrain a systemd-managed service to bound memory exposure
# /etc/systemd/system/<service>.service.d/override.conf
[Service]
MemoryMax=512M
Restart=on-failure
RestartSec=5s
Disclaimer: This content was generated using AI. While we strive for accuracy, please verify critical information with official sources.

