JeecgBoot CVE-2026-108609

Requires a valid authenticated JeecgBoot account (PR:L) and knowledge or enumeration of another user's id value, which is supplied directly in the userId parameter of GET /airag/voice/listByUser; the AI voice (airag/voice) module must be present and in use so that voice-generation history is stored in Redis. No user interaction or elevated privileges are needed, and exploitation works over the network against default configurations once authenticated. The practical limiters are the authentication requirement and the need to know/guess valid victim user ids; impact is confined to reading voice-history data (TTS text, voice settings, timestamps, audio file names/paths) with no write or service-disruption capability.

This is a moderate, confidentiality-only access-control flaw, not a top-tier priority. The CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) and CWE-639 agree with the description: an already-authenticated low-privilege user can read other users' data over the network with low complexity and no user interaction, and the impact is limited to low confidentiality (VC:L) with no integrity or availability effect. The 5.3 score is consistent with this scope. Publicly available exploit code exists (a Python POC on GitHub and a VulnCheck advisory), which raises practical exploitability, but the vulnerability is NOT in CISA KEV and no EPSS score was provided, so there is no evidence of active or widespread exploitation - treat the POC as proof of feasibility, not of in-the-wild attacks. The meaningful limiter is the PR:L requirement: an attacker must already hold a valid low-privilege account and must know or enumerate a victim's user id, which caps this as an insider/post-authentication data-exposure issue rather than a pre-auth breach. Note the input tag 'Authentication Bypass' conflicts with the actual nature of the flaw, which is an authorization/IDOR issue requiring authentication (PR:L) - this is a labeling discrepancy, not an authentication bypass.

A logged-in JeecgBoot user (e.g., a low-privilege tenant on a shared deployment) requests GET /airag/voice/listByUser with another user's id in the userId parameter and receives that user's text-to-speech inputs, voice settings, timestamps, and generated audio file paths from Redis. A public Python POC (poc_airag_voice_history_read.py) automates this request. Because the vector is network-based, low-complexity, and needs no victim interaction, enumerating or guessing sequential user ids lets one account harvest many users' voice history.