cve.bz

CVE-2026-78148 ggml-org llama.cpp: Improper Resource Shutdown or Release

6.9 mediumEPSS 0.7%

Medium (6.9) in ggml-org llama.cpp. EPSS 0.7% estimated probability of exploitation activity in the next 30 days.

On this page
A vulnerability was determined in ggml-org llama.cpp bec4772f6. This affects the function rpc_server::graph_compute of the file ggml/src/ggml-rpc/ggml-rpc.cpp of the component ggml-RPC Server. Executing a manipulation can lead to null pointer dereference. The attack may be launched remotely. The pull request to fix this issue awaits acceptance.

Severity

CVSS 4.0

6.9 medium
CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:L/SC:N/SI:N/SA:N

CVSS 3.1

5.3 medium
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
EPSS
0.7% (unchanged from 0.7% a week ago; 0.7% 30 days ago) estimated probability of exploitation activity in the next 30 days; 55th percentile. Scored 2026-11-17. A forecast for this CVE across the internet, not for your own systems.
CISA's assessment (SSVC)
Exploitation: none · Automatable: yes · Technical impact: partial, assessed 2026-08-24. From CISA's Vulnrichment in the CVE List: whether it is exploited, whether exploiting it can be automated, and how much control it gives.
CWEs
CWE-404 Improper Resource Shutdown or Release
CWE-476 NULL Pointer Dereference
OWASP Top 10
A10:2025 Mishandling of Exceptional Conditions
Published
Modified
NVD analysis
Deferred: the NVD does not plan to analyze it
Record from
the NVD, synced

Ask an AI about this CVE

Show the promptExplain CVE-2026-78148 for an engineer deciding whether to act. Use https://cve.bz/api/public/cves/CVE-2026-78148?via=ask-ai-copy and https://cve.bz/cve/CVE-2026-78148?via=ask-ai-copy as the source of truth: what is affected and which versions fix it, whether it is in CISA KEV, public exploits, and its EPSS (an internet-wide 30-day estimate of exploitation activity, not a forecast for my systems). If they give no fixed versions, use the CVE record https://cveawg.mitre.org/api/cve/CVE-2026-78148 or the vendor advisory, not third-party mirrors. Tell me how to check whether I am affected and how to verify a fix. Cite the sources and say where they are uncertain.

Opens a third-party assistant; its answer is its own, not cve.bz's. The prompt holds only this CVE's id and public links.

Affected

Versions

From the CNA's record in the CVE List.

ProductAffected
ggml-org llama.cppbec4772f6

Linux distributions

DistributionReleasePackageFixed inAdvisory
Debian14llama.cppno fix listedDEBIAN-CVE-2026-78148
Ubuntu26.04:LTS, 26.10llama.cppno fix listedUBUNTU-CVE-2026-78148

From the distributions' data via OSV.dev: the Debian Security Tracker; Ubuntu Security Notices © Canonical Ltd., CC BY-SA 4.0. Modified: reformatted and linked to CVE records. Credits. On the distributions' own pages: Debian · Ubuntu.

Red Hat

Red Hat rates this Moderate.

Red Hat's statement

Red Hat Product Security has determined that this vulnerability affects the ggml-RPC Server component of llama.cpp, which is an optional feature that requires explicit build-time enablement (GGML_RPC=ON) and operator action to start the ggml-rpc-server binary. The vulnerability does not affect the core llama.cpp library or typical usage patterns of Python bindings (llama-cpp-python) or Rust bindings (llama-cpp-2) that are bundled in Red Hat products.

Most Red Hat products that include llama.cpp dependencies (RHEL AI containers, goose package) bundle only the inference library bindings and do not build, ship, or expose the vulnerable RPC server component. These products are not affected by this vulnerability.

The impact is rated Moderate with CVSS score 5.3 (CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L) because exploitation requires the optional RPC server to be explicitly enabled and running, and results only in denial of service of that server process, not the broader system. The low availability impact (A:L) reflects that this affects an opt-in feature rather than a critical system component.

Red Hat's mitigation

For systems where the ggml-RPC Server is required and actively used:

1. Restrict network access to the RPC server port (default TCP 1234) using firewall rules to trusted clients only

2. Do not expose the ggml-rpc-server to untrusted networks or the public internet

3. Consider using SSH tunneling or VPN for remote RPC access

4. Monitor for and restart crashed RPC server processes until a patched version is available

For most Red Hat product deployments that use llama.cpp only through Python or Rust bindings, no mitigation is required as the vulnerable component is not included.

The upstream fix is available in GitHub pull request #25670 (https://github.com/ggml-org/llama.cpp/pull/25670). Update to a version containing this fix when available.

ProductPackageStateAdvisory
Red Hat Enterprise Linux 10gooseNot affected
Red Hat Enterprise Linux 9gooseNot affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-cuda-rhel9Not affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-gaudi-rhel9Not affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/bootc-rocm-rhel9Not affected
Red Hat Enterprise Linux AI (RHEL AI) 3rhelai3/disk-image-cuda-rhel9Not affected

Contains Red Hat Security Data © Red Hat, Inc., licensed under CC BY 4.0. Source: https://access.redhat.com/security/data. Modified: reformatted. Red Hat's page for CVE-2026-78148.

Attack patterns (CAPEC)

  • CAPEC-125 Flooding Likelihood: High · Severity: Medium
  • CAPEC-130 Excessive Allocation Likelihood: Medium · Severity: Medium
  • CAPEC-131 Resource Leak Exposure Likelihood: Medium · Severity: Medium
  • CAPEC-494 TCP Fragmentation
  • CAPEC-495 UDP Fragmentation
  • CAPEC-496 ICMP Fragmentation
  • CAPEC-666 BlueSmacking Likelihood: Medium · Severity: Medium

The attack patterns MITRE's CAPEC relates to this record's weaknesses, as MITRE names them. Credits.

References

Check your stackJSON