CVE-2025-27787 in Applio
Summary
by MITRE • 03/19/2025
Applio is a voice conversion tool. Versions 3.2.8-bugfix and prior are vulnerable to denial of service (DoS) in restart.py. `model_name` in train.py takes user input, and passes it to the `stop_train` function in restart.py, which uses it construct a path to a folder with `config.json`. That `config.json` is opened and the list of values under "process_pids" are read. Next all the process IDs listed in the JSON are killed. Using one of the arbitrary file writes, one can write to `logs/foobar` a `config.json` file, which contains a list of process IDs. Then one can access this endpoint to kill these processes. Since an attacker can't know what process is running on which process ID, they can send a list of hundreds of process IDs, which can kill the process that applio is using to run, as well as other, potentially important processes, which leads to DoS. Note that constructing a path with user input also enables path traversal. For example, by supplying "../../" in `model_name` one can access `config.json` freom locations two folders down on the server. As of time of publication, no known patches are available.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 08/01/2025
The vulnerability CVE-2025-27787 affects Applio voice conversion software versions 3.2.8-bugfix and earlier, presenting a critical denial of service risk through improper input validation and insecure path construction. This flaw exists in the restart.py module where user-supplied input from train.py's model_name parameter is directly utilized without adequate sanitization or validation. The system constructs file paths using this untrusted input to locate and access configuration files, creating multiple attack vectors including path traversal and arbitrary file manipulation. The core technical issue manifests when the stop_train function processes the user-provided model_name value to build a file path leading to a config.json file, which then gets parsed to extract process identifiers for termination.
The operational impact of this vulnerability extends beyond simple service disruption to potentially compromise system stability and availability. An attacker can leverage arbitrary file write capabilities to place a malicious config.json file at logs/foobar, containing a crafted list of process IDs that will be terminated when the vulnerable endpoint is accessed. This creates a scenario where hundreds of process IDs can be targeted simultaneously, potentially including the critical Applio processes themselves along with other important system services. The vulnerability's path traversal capability amplifies the attack surface, allowing adversaries to access configuration files from arbitrary locations on the server filesystem by manipulating the model_name parameter with directory traversal sequences like "../../". This combination of path traversal and process termination creates a powerful denial of service mechanism that can destabilize the entire application environment.
The security implications of this vulnerability align with CWE-22 Path Traversal and CWE-400 Uncontrolled Resource Consumption, both of which are catalogued in the Common Weakness Enumeration database and represent fundamental flaws in input validation and resource management. From an attack framework perspective, this vulnerability maps to ATT&CK technique T1499.004 for Network Denial of Service and T1059.001 for Command and Scripting Interpreter, as it enables attackers to execute destructive commands against system processes. The lack of available patches at the time of publication compounds the risk, leaving affected systems vulnerable to exploitation without immediate remediation options. Organizations using Applio must implement immediate mitigations including input validation controls, path sanitization, and restricted file access permissions to prevent unauthorized manipulation of configuration files and protect against this critical denial of service vulnerability.