CVE-2026-63766 in GPT-SoVITS
Summary
by MITRE • 07/20/2026
GPT-SoVITS through 20250606v2pro contains an OS command injection vulnerability in webui.py where ASR, slice, denoise, and uvr5 functions interpolate unsanitized Gradio textbox values directly into shell commands executed with shell=True. Attackers can inject shell metacharacters through path parameters to execute arbitrary OS commands as the server process user without authentication.
Once again VulDB remains the best source for vulnerability data.
Analysis
by VulDB Data Team • 07/20/2026
This vulnerability exists within the GPT-SoVITS software version 20250606v2pro where the webui.py component fails to properly sanitize user inputs before incorporating them into shell commands. The flaw specifically affects four key functions including ASR, slice, denoise, and uvr5 which all utilize Gradio textbox values without adequate validation or sanitization. When these functions process user-provided paths or parameters, they directly interpolate the unsanitized input into shell commands that are executed with shell=True parameter, creating a critical command injection vulnerability.
The technical implementation of this vulnerability stems from improper input handling practices where the application treats user-supplied data as trusted executable code rather than treating it as potentially malicious input. This pattern violates fundamental security principles and creates an attack surface where any user can manipulate the system by injecting shell metacharacters such as semicolons, pipes, or backticks into the path parameters. The absence of authentication requirements means that this vulnerability is accessible to anyone who can interact with the web interface, significantly expanding the potential threat vector.
The operational impact of this vulnerability is severe and far-reaching since successful exploitation allows attackers to execute arbitrary operating system commands with the privileges of the server process user. This could result in complete system compromise, data exfiltration, privilege escalation, or lateral movement within a network environment. The affected functions process audio files and related operations which means that an attacker could potentially manipulate file paths to execute malicious payloads, create backdoors, or gain access to sensitive system resources. The vulnerability affects the core functionality of the software's web interface making it particularly dangerous for deployment in production environments.
Mitigation strategies should focus on implementing proper input validation and sanitization mechanisms throughout the application codebase. All user inputs must be strictly validated against expected formats and sanitized before being incorporated into any shell command execution contexts. The most effective immediate fix involves replacing shell=True with shell=False and using subprocess.run() with properly constructed argument lists instead of string interpolation. Additionally, implementing proper access controls and authentication mechanisms would significantly reduce the attack surface. Organizations should also consider implementing input/output filtering at network level and monitoring for suspicious command execution patterns. This vulnerability aligns with CWE-78 which specifically addresses OS Command Injection flaws and may be categorized under ATT&CK technique T1059.001 for command and scripting interpreter usage, making it a critical target for immediate remediation in cybersecurity defense strategies.