DeepChat CVE-2025-67744

Exploitation requires two specific conditions present in DeepChat before 0.5.3: (1) the Mermaid diagram rendering component configured unsafely (allowing script execution within rendered diagrams), and (2) the Electron IPC renderer being exposed to the DOM, which lets injected JavaScript invoke IPC and reach host command execution. The concrete prerequisite for an attack is that the victim renders attacker-controlled Mermaid diagram content - delivered via a chat message, shared content, or AI/tool output - inside the DeepChat desktop app. The principal limiting factor is UI:R: the victim must actually load/render the malicious diagram; there is no server-side or fully passive trigger, and the flaw is fixed in 0.5.3, so patched installs are not exploitable.

This is a genuine high-priority issue, not an inflated-CVSS artifact. The CVSS 3.1 vector (AV:N/AC:L/PR:N/UI:R/S:C/C:H/I:H/A:H, base 9.6) describes a network-reachable, low-complexity, unauthenticated flaw whose only gating factor is user interaction (UI:R), and the scope change (S:C) is justified because an XSS in the Electron renderer escalates across a trust boundary into full host-level RCE. The description's mechanism (unsafe Mermaid config + an Electron IPC renderer exposed to the DOM) is internally consistent with that vector, so there is no network-vs-local contradiction to flag. Publicly available exploit code exists (referenced GitHub Security Advisory GHSA-w8w8-82pv-5rg9), which raises practical risk, but note that POC availability is NOT the same as active exploitation. This CVE is NOT listed in CISA KEV in the provided data, and no EPSS score was supplied, so widespread or confirmed in-the-wild exploitation cannot be asserted - the honest read is 'serious, weaponizable, publicly documented, but no evidence of mass exploitation yet.' The realistic risk ceiling is moderated by the UI:R requirement and by the product's limited deployment footprint (an open-source desktop AI client rather than ubiquitous server software).

An attacker crafts a malicious Mermaid diagram payload and delivers it to a DeepChat user - for example embedded in a shared chat message, a document, or content returned by a connected AI tool/agent. When the victim's DeepChat client (prior to 0.5.3) renders the diagram, the unsafe Mermaid configuration executes attacker JavaScript, which reaches the exposed Electron IPC bridge and escalates to arbitrary OS command execution on the victim's machine. Public exploit code for this chain exists in the referenced GitHub advisory.