Sceawere
Vulnerability Detail
CVE-2026-78148UPDATED Verified Sceawere Triage Sources: NVD / CISA KEV
llama.cpp RPC Server Null Pointer Dereference
Vulnerability Metadata
- Severity
- Medium
- Score / CVSS
- 5.3
- Creation Date
- 3h ago
- Vendor
- ggml-org
- Product
- llama.cpp
- Attack Type
- NULL Pointer Dereference
- Vector String
- CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:L
- Attack Complexity
- LOW
Narrative and Response
Description
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.
Executive Summary
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
Technical Details
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
Mitigations
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
References
Lorem ipsum dolor sit amet, consectetur adipiscing elit. Sed do eiusmod tempor incididunt ut labore et dolore magna aliqua. Ut enim ad minim veniam, quis nostrud exercitation ullamco laboris nisi ut aliquip ex ea commodo consequat.
Duis aute irure dolor in reprehenderit in voluptate velit esse cillum dolore eu fugiat nulla pariatur. Excepteur sint occaecat cupidatat non proident, sunt in culpa qui officia deserunt mollit anim id est laborum.
Additional Metadata
{
"score": "5.3",
"pubDate": "2026-08-24T00:16:46.293Z",
"pubdate": "2026-08-24T00:16:46.293Z",
"executiveSummary": "A vulnerability classified as a null pointer dereference has been identified in the ggml-org llama.cpp component ggml-RPC Server, specifically within commit bec4772f6. This security flaw impacts the rpc_server::graph_compute function located in ggml/src/ggml-rpc/ggml-rpc.cpp.\nThe vulnerability allows remote attackers to trigger a denial of service condition through the execution of a crafted manipulation. Successful exploitation results in a null pointer dereference, leading to application crashes and disruption of the RPC server functionality.\nThe attack vector is remote, enabling unauthorized external actors to interact with the vulnerable ggml-RPC Server instance without requiring prior authentication or elevated privileges.\nRisk implications include operational downtime and service instability for distributed machine learning workloads utilizing the vulnerable RPC architecture. Remediation requires applying the upcoming patch associated with the pending pull request.",
"technicalDetails": "The vulnerability resides in the ggml-org llama.cpp component, specifically within the rpc_server::graph_compute function implemented in the ggml/src/ggml-rpc/ggml-rpc.cpp source file.\nThe root cause stems from insufficient validation and sanitization of incoming RPC request data or graph structures processed by the ggml-RPC Server.\nWhen a remote client transmits a maliciously crafted computation graph or malformed RPC payload, the server attempts to parse and execute graph operations within the rpc_server::graph_compute routine.\nDue to the absence of proper null pointer checks or boundary validations on internal data structures, the execution flow attempts to dereference an uninitialized or null memory address.\nThe attack flow begins when an attacker establishes network connectivity with the exposed ggml-RPC Server endpoint. The attacker then crafts and transmits a specialized network payload designed to manipulate the graph computation sequence.\nUpon receiving the payload, the vulnerable function processes the input data, encounters the invalid reference state, and triggers an unhandled exception or direct segmentation fault.\nThe network exposure is direct, as the ggml-RPC Server listens for remote connections to distribute computational tasks across nodes.\nAuthentication and privilege requirements are minimal or absent, allowing remote unauthenticated actors to initiate the attack sequence.\nThe post-exploitation impact is strictly limited to a denial of service via application termination, as memory corruption leading to remote code execution is not indicated by the primitive vulnerability type.\nAffected versions include ggml-org llama.cpp at commit bec4772f6 and potentially surrounding revisions that lack input validation within the RPC server graph computation handler."
}