ruoyi-vue-pro CVE-2026-110610

Requires an authenticated application account (PR:L) with access to the yudao-module-ai image feature's Midjourney imagine endpoint (AiImageController), and exploitation depends on supplying the referImageUrl parameter, which is passed to Hutool HttpUtil.downloadBytes() without SSRF filtering. Real-world impact is amplified when the deployment runs in a cloud environment exposing a metadata service (e.g., 169.254.169.254, IMDSv1) or hosts internal-only HTTP services reachable from the application host. Limiting factors: valid credentials are mandatory, no unauthenticated path exists, and the flaw yields information disclosure/internal reach rather than direct code execution.

This is a medium-severity (CVSS 4.0 base 5.3) server-side request forgery issue that is a moderate, not top-tier, priority. The provided CVSS 4.0 vector (AV:N/AC:L/AT:N/PR:L/UI:N) indicates network-reachable, low-complexity exploitation but requires authentication (PR:L), meaning the attacker must already hold a valid application account - a significant limiting factor that pushes this below unauthenticated internet-facing SSRFs. Impact metrics are deliberately modest: VC:L/VI:N/VA:N with subsequent-system SC:L/SI:L reflect that the flaw enables internal probing and reaching cloud metadata endpoints rather than direct full compromise; the real danger is escalation (e.g., harvesting cloud IAM credentials from metadata services), which the base score does not fully capture. No EPSS score, no KEV listing, and no POC were provided in the input, so there is no signal of active or probable exploitation - this was reported by VulnCheck as a research finding. There is no conflict among the signals: all point to a genuine but authentication-gated internal-network risk. Treat as a real but lower-urgency fix, prioritized higher only in environments running in cloud with reachable metadata endpoints (IMDSv1).

An authenticated low-privileged user of a ruoyi-vue-pro deployment calls the Midjourney imagine endpoint and sets referImageUrl to an internal URL such as http://169.254.169.254/latest/meta-data/ or http://127.0.0.1:port. The server's Hutool HttpUtil.downloadBytes() fetches the attacker-chosen URL, letting the attacker port-scan internal-only HTTP services or retrieve cloud instance metadata. No public exploit code was identified at time of analysis, though the vulnerable code paths are referenced directly on GitHub.