Attack ChainAIDerived
Hypothetical attack flow derived from CVE metadata – not a confirmed exploit chain
-
AccessReach exposed RVC WebUI
-
ExploitSubmit crafted ckpt_path value
-
ExecutionTrigger code injection in change_info_
-
ImpactExecute arbitrary code on host
Vulnerability Assessment
| Weakness |
CWE
OWASP Top 10:2025
A05:2025 - Injection
ASVS 5.0
V1 - Encoding and Sanitization Code injection prevention, sandbox exec
MITRE ATT&CK
2 techniques across 2 tactics: Execution, Initial Access The technique list with ATT&CK links and the kill-chain view is part of the Basic plan and up. |
| Exploitation Conditions | Exploitation requires the attacker to be able to reach the Gradio-based WebUI and control the ckpt_path argument passed to the change_info_ model-info function in webui.py. The WebUI binds to localhost (127.0.0.1) by default, so the 'remote' vector is only realistic when the operator has explicitly exposed the interface (e.g., Gradio --listen/share or a reverse proxy/tunnel) to an untrusted network. No authentication is enforced by the default interface (PR:N), but user/operator interaction is required to submit the crafted checkpoint path (UI:P). Principal limiting factors: default local-only binding, required UI interaction, absence of auth meaning risk is concentrated on exposed instances, and the niche, non-enterprise deployment base. |
| Impact | This is a medium-severity (CVSS 4.0 base 5.3) code injection flaw (CWE-94) in the RVC-Project Retrieval-based-Voice-Conversion-WebUI, with publicly available exploit code but no CISA KEV listing and no EPSS score provided. The signals are mixed: the CVSS 4.0 vector claims a network attack vector with no privileges (AV:N/PR:N) which, combined with a published PoC (E:P), looks alarming, yet the vendor rated the three impact metrics as only Low (VC:L/VI:L/VA:L) despite this being a code-injection/RCE class bug - code injection normally yields full technical impact, so the Low ratings appear conservative or under-scored. More importantly, the 'attack can be executed remotely' claim should be treated with skepticism: this is a self-hosted Gradio-based ML tool that by default binds to localhost, so genuine remote exploitation depends on the operator having deliberately exposed the interface to the network. The vulnerability was reported solely by VulDB, the project was notified via a GitHub issue but has NOT responded, and there is no independent vendor confirmation. Net assessment: a real, exploitable code-execution bug worth patching for anyone who exposes the WebUI, but NOT a widespread-exploitation priority - the combination of a niche hobbyist product, default local-only binding, required user interaction (UI:P), and no KEV/EPSS evidence points to limited, opportunistic risk rather than a mass-exploited emergency. |
| Exploit ScenarioDerived | An attacker who can reach an exposed RVC WebUI instance (or who tricks an operator into loading a malicious artifact) supplies a crafted ckpt_path value to the change_info_ function in webui.py, which is unsafely evaluated and results in code injection on the host. A public proof-of-concept (packaged as rvc-train-log-eval-poc.zip) has been published, lowering the bar for exploitation. Because the Gradio UI binds to localhost by default, practical remote exploitation requires the instance to have been intentionally exposed to the network. |
Get an e-mail when this CVE is exploited, patched or re-scored. Free account.
Alert meAnalysisAIv1 – Oct 19, 2026
Code injection in RVC-Project Retrieval-based-Voice-Conversion-WebUI through 2.3.260718 lets unauthenticated remote attackers who can reach an exposed Gradio interface manipulate the ckpt_path argument passed to change_info_ in webui.py and execute attacker-controlled code, although user/operator interaction is required and default deployments bind to localhost. Publicly available exploit code exists, but there is no CISA KEV listing and no EPSS score was provided, and the project has not responded to the issue report. Practical risk is concentrated on instances explicitly exposed via Gradio --listen/--share or a reverse proxy/tunnel, where the lack of default authentication makes the code-injection path reachable.
RemediationAI
No vendor-released patch identified at time of analysis; the project has not responded to issue #2866. Until a fix is published, keep the Gradio WebUI bound to 127.0.0.1 and avoid --listen, --share, or reverse-proxy exposure; this blocks remote exploitation but also blocks legitimate remote access. If remote access is required, place the interface behind an authenticating reverse proxy with an IP allowlist, terminate public sharing, and restrict access to trusted operators; this adds operational overhead but reduces exposure. Disable or restrict the change_info_ model-info functionality to trusted administrators, or validate ckpt_path against an allowlist of approved checkpoint directories and reject path traversal or code-like input; this can break workflows that legitimately load checkpoints from arbitrary locations. Monitor for anomalous checkpoint paths and review the public PoC at https://github.com/user-attachments/files/32909432/rvc-train-log-eval-poc.zip plus the VulDB/CTI references for updates.
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
RVC-Project Retrieval-based-Voice-Conversion-WebUI is a Python and Gradio application for retrieval-based voice-conversion model training and inference. The vulnerable function change_info_ in webui.py handles model-information requests and accepts a ckpt_path argument that identifies a checkpoint; because that value is not adequately validated or constrained before reaching a code-execution sink, an attacker who can submit a crafted path can alter the code executed by the application, which is the CWE-94 code-injection class. The CPE cpe:2.3:a:rvc-project:retrieval-based-voice-conversion-webui:*:*:*:*:*:*:*:* indicates the flaw applies across the product line through version 2.3.260718. The Gradio server binds to 127.0.0.1 by default, so remote exploitation requires the operator to expose it through --listen/--share or a reverse proxy/tunnel; no authentication is enforced by default (PR:N), but user or operator interaction is required to submit or load the crafted checkpoint path (UI:P/UI:R). The official CVSS 4.0 vector rates direct confidentiality/integrity/availability impact as low (VC:L/VI:L/VA:L), while the independent CVSS 3.1 assessment rates C:H/I:H/A:H; the difference reflects assumptions about successful exploitation under the required interaction and exposure conditions.
Know when Retrieval Based Voice Conversion Webui 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:N/AC:L/AT:N/PR:N/UI:P/VC:L/VI:L/VA:L/SC:N/SI:N/SA:N/E:P
Code injection yields arbitrary code execution, so C/I/A are High; default Gradio UI enforces no auth (PR:N) but the crafted ckpt_path must be submitted through the UI (UI:R).
Primary rating from Vendor (VulDB).
CVSS VectorVendor: VulDB
Lifecycle Timeline
3DescriptionCVE.org
A flaw has been found in RVC-Project Retrieval-based-Voice-Conversion-WebUI up to 2.3.260718. Impacted is the function change_info_ of the file webui.py. Executing a manipulation of the argument ckpt_path can lead to code injection. The attack can be executed remotely. The exploit has been published and may be used. The project was informed of the problem early through an issue report but has not responded yet.
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
Code injection occurs when an application accepts user input and passes it directly into a language interpreter or evaluator without proper sanitization.
Remote Code Execution represents the critical moment when an attacker successfully runs arbitrary code on a target system without physical access.
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
Retrieval-based-Voice-Conversion-WebUI is a voice changing framework based on VITS. Rated high severity (CVSS 8.9), this
EUVD-2026-104066
GHSA-mgjw-qc2q-q7rc
NVD
CVE.org
VulDB
Exploit-DB