Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessGain on-path position for update traffic
-
DeliveryIntercept Ollama auto-update HTTP request
-
ExploitReturn response with traversal headers and payload
-
Executionfilepath.Join writes binary to Startup dir
-
PersistUnsigned staged binary auto-executes
-
ImpactPersistent code execution as current user
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A08:2025 - Software or Data Integrity Failures
ASVS 5.0
V15 - Secure Coding and Architecture Code integrity verification
|
| Exploitation Conditions | Exploitation requires the attacker to control or tamper with the HTTP response to Ollama for Windows' automatic update request - i.e., an adjacent-network or on-path (MITM) position such as a shared LAN/Wi-Fi, malicious proxy, or DNS/route hijack (reflected as AV:A and the AT:P attack requirement in the CVSS 4.0 vector). The vulnerable feature is specifically the update-download routine that builds local file paths from attacker-controlled HTTP headers via filepath.Join without validation, and the impact chain depends on the companion missing-signature-verification flaw (CVE-2026-42248) so the planted binary is accepted and run. No authentication and no user interaction are needed once the attacker is on-path (PR:N/UI:N), and silent auto-update plus execution of staged binaries makes it fully automatic. Limiting factors: only Ollama for Windows in the 0.12.10-0.17.5 range is confirmed; the required privileged network position and AC:H raise the practical bar; and files are written only to locations writable by the current user (no privilege escalation is claimed). |
| Impact | The signals are mixed and point to a serious-but-not-mass-exploited issue. The vendor CVSS 4.0 base score is 7.7 (High) with vector AV:A/AC:H/AT:P/PR:N/UI:N and total technical impact, but the AV:A (adjacent), AC:H, and AT:P metrics show exploitation is gated on a privileged network position: the attacker must be able to intercept or control the update HTTP response (MITM, malicious proxy, DNS/BGP hijack, or compromised update path). CISA SSVC scores Exploitation: none, Automatable: no, Technical Impact: total - consistent with high damage but no easy, scalable exploitation. EPSS is 0.03% (10th percentile), and the issue is not in CISA KEV, so this is not evidence of active or widespread exploitation. Net: a genuine high-impact chain (unauthenticated, no user interaction, persistent RCE) whose real-world risk is throttled by the man-in-the-middle prerequisite rather than a low-value high-CVSS false alarm. Note the input provides only a CVSS 4.0 vector (no 3.1 vector), so auth is read from PR:N in the 4.0 string. |
| Exploit ScenarioDerived | An attacker positioned on the same network segment or in the update path (e.g., a rogue Wi-Fi hotspot, malicious proxy, or DNS/route hijack) intercepts Ollama's automatic update request and returns a crafted HTTP response whose headers contain '../' traversal sequences and a malicious executable. Ollama writes the binary into the Windows Startup directory and, because updates are unsigned and applied silently, the payload executes automatically on next logon with no user interaction. No public exploit is identified at time of analysis, but the technique is well-documented by CERT-PL. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Jul 24, 2026
Remote code execution in Ollama for Windows (versions 0.12.10 through 0.17.5) lets an attacker who can influence HTTP update responses write arbitrary executables outside the update staging directory. The flaw stems from unvalidated attacker-controlled HTTP headers being passed to filepath.Join, permitting path traversal into sensitive locations such as the Windows Startup folder. Chained with CVE-2026-42248 (missing update signature verification), the silent auto-updater executes the planted binary, yielding automatic and persistent code execution with no user interaction. There is no public exploit identified at time of analysis and EPSS is very low (0.03%).
RemediationAI
No vendor-released patch identified at time of analysis - the maintainers were notified but did not provide a fixed version, so no exact upgrade target can be cited. As compensating controls, prevent the man-in-the-middle precondition and the silent-update primitive: enforce that update traffic uses validated TLS and cannot be intercepted (avoid untrusted networks/proxies for hosts running Ollama, pin or monitor the ollama.com update endpoint), which does not fix the traversal but removes the attacker's ability to inject responses. Where feasible, disable or block Ollama's automatic update mechanism (trade-off: you must then patch manually and will miss legitimate fixes), and restrict write access to and monitor the Windows Startup directory and the update staging path for unexpected executables (trade-off: detective, not preventive). Consider pausing deployment of Ollama for Windows on sensitive hosts until a signed, path-validated update build is published. Track vendor updates via https://ollama.com/ and https://cert.pl/en/posts/2026/04/CVE-2026-42248/ ; correlate fix availability with the companion CVE-2026-42248 signature-verification fix.
Affected VersionsENISA
Affected Packages
Source: OSV (Open Source Vulnerabilities)
Affects your stack
Affects 1 package in euvd
Checking this CVE against your own stacks and servers is part of the Basic plan and up.
Technical ContextAI · show details
Ollama is a local large-language-model runtime; its Windows build ships a background auto-update client that downloads and stages update binaries over HTTP. The root cause is CWE-494 (download of code without integrity check) compounded by a path-traversal primitive: the updater derives local destination file names from server-supplied HTTP response headers and concatenates them with filepath.Join without sanitizing '../' sequences. In Go, filepath.Join cleans but does not confine paths, so traversal segments resolve to arbitrary user-writable directories. Because a companion defect (CVE-2026-42248) omits signature verification of the downloaded payload, the integrity control that would otherwise stop a malicious update is absent, turning a write-primitive into reliable code execution. The affected component is identified by CPE cpe:2.3:a:ollama:ollama and tagged Microsoft / Path Traversal / RCE.
Know when Ollama gets its next vulnerability
- Watch products and get an email when something new is published
- Check this CVE against your own stack and servers
- Exploit and PoC references, threat intelligence from 22 sources
No credit card · 7-day full trial
Severity by source
CVSS:4.0/AV:A/AC:H/AT:P/PR:N/UI:N/VC:H/VI:H/VA:H/SC:L/SI:L/SA:L
On-path control of update responses gives AV:A and AC:H; no auth or interaction (PR:N/UI:N); silent unsigned update yields full C/I/A on the host with no privilege-boundary change (S:U).
Primary rating from NVD.
CVSS VectorNVD
Lifecycle Timeline
2DescriptionCVE.org
Ollama for Windows contains a Remote Code Execution vulnerability in its update mechanism due to improper handling of attacker‑controlled HTTP response headers. When downloading updates, the application constructs local file paths using values derived from HTTP headers without validation. These values are passed directly to filepath.Join, allowing path traversal sequences (../) to be resolved and enabling files to be written outside the intended update staging directory. An attacker who can influence update responses can exploit this flaw to write arbitrary executables to attacker‑chosen locations accessible to the current user, including the Windows Startup directory. This allows execution of arbitrary executables.
Critically, when chained with CVE‑2026‑42248 (Missing Signature Verification for Updates), an attacker can deliver malicious payloads that are written to sensitive locations and executed automatically. Because Ollama for Windows performs silent automatic updates and executes staged binaries without user interaction, this results in automatic and persistent code execution without user awareness.
Maintainers of this project were notified early about this vulnerability, but didn't respond with the details of vulnerability or vulnerable version range. Versions from 0.12.10 to 0.17.5 were tested and confirmed as vulnerable, other versions were not tested but might also be vulnerable.
Executive ViewAI
Ollama for Windows (versions 0.12.10-0.17.5) contains a high-severity remote code execution flaw that silently executes attacker-supplied code through compromised HTTP update responses, affecting any Windows system running this LLM platform. The vulnerability requires no user interaction and can establish persistent access via the application's automatic update mechanism, posing critical risk to development teams, research departments, and AI workload deployments. Organizations must immediately identify and remediate affected installations to prevent unauthorized code execution and data access.
No vendor-released patch identified at time of analysis. Interim mitigations: disable auto-update functionality via configuration, isolate Ollama systems from public networks and untrusted clients, implement endpoint detection and response (EDR) monitoring for unsigned executable creation in Windows system directories, enforce application whitelisting to prevent unauthorized binary execution, maintain current Windows Defender signatures, and restrict network access via host-based or network firewalls to prevent external update manipulation.
Threat Intelligence Overview
How this decision was reached
The engine reads each signal below and applies a fixed list of rules from top to bottom. The first rule that matches sets the recommended action – like a statement, every line is shown.
What every action level means
SSVC AssessmentENISA
Priority Score
Attack Technique
Path traversal exploits occur when applications use user-controlled input to construct file system paths without proper validation.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
Ollama before 0.1.34 does not validate the format of the digest (sha256 with 64 hex digits) when getting the model path,
Silent auto-update code execution in Ollama for Windows (versions 0.12.10 through 0.17.5) stems from an update-verificat
Remote denial of service in Ollama (HEAD f0078ae) allows unauthenticated attackers to crash the entire server by uploadi
Denial of service in Ollama's multi-modal image processing (versions 0.11.5-rc0 through 0.13.5) lets remote attackers cr
An issue was discovered in Ollama before 0.1.46. Rated high severity (CVSS 8.2), this vulnerability is remotely exploita
A vulnerability in the Ollama server version 0.5.11 allows a malicious user to cause a Denial of Service (DoS) attack by
Information disclosure in Ollama's model quantization engine (confirmed affected in v0.13.5) allows remote, unauthentica
A vulnerability in ollama/ollama versions <=0.3.14 allows a malicious user to upload and create a customized GGUF model
An issue in ollama v.0.12.10 allows a remote attacker to cause a denial of service via the fs/ggml/gguf.go, function rea
An issue in ollama v.0.12.10 allows a remote attacker to cause a denial of service via the GGUF decoder [CVSS 7.5 HIGH]
A divide by zero vulnerability exists in ollama/ollama version v0.3.3. Rated high severity (CVSS 7.5), this vulnerabilit
A vulnerability in ollama/ollama versions <=0.3.14 allows a malicious user to create a customized GGUF model file that,
EUVD-2026-26211
NVD
CVE.org
EPSS
Vulnrichment
VulDB
GitHub PoC
Exploit-DB