Quest Bot EUVD-2026-114279

Attacker must (1) be an ordinary member of a Discord server that has the bot installed - the image-to-GIF command's developer-only gate is disabled in normal deployments, so no elevated permissions are needed; (2) control a domain with authoritative DNS set to a very short cache lifetime so it can answer a public address during the validation lookup and an internal address (e.g. loopback) during the actual request; and (3) target a service on the bot's host/internal network that speaks HTTPS on port 443 AND returns a genuine image for content to come back. Limiting factors that reduce realistic risk: exploitation requires winning a DNS-rebinding timing race (CVSS AC:H/AT:P), requests are forced to HTTPS on port 443 only, only responses that are genuine images are returned (otherwise the request is blind), and the common cloud metadata service (plain HTTP on a different port) is explicitly NOT reachable through this path.

Real-world risk is low and the 'monitor' priority label is appropriate. The CVSS 4.0 base score is only 2.3, and the vector (AV:N/AC:H/AT:P/PR:L/UI:N) shows that while the command is network-reachable, exploitation is gated by high attack complexity and a specific attack requirement - the attacker must control a domain, win a DNS-rebinding timing race, and the bot host must actually run an internal service that speaks HTTPS on port 443 and returns a genuine image. Impact is deliberately constrained: VC:L/VI:N/VA:N plus SC:L reflects blind internal probing and limited retrieval of image responses, not arbitrary internal-data disclosure, and the advisory explicitly notes the common cloud metadata service (plain HTTP on a non-standard port) is NOT reachable. There is no CISA KEV listing and no EPSS score was supplied in the input, so no quantitative exploitation-probability signal is available - absence of EPSS should not be read as either high or low likelihood. A PoC methodology is described in the GHSA advisory but is not confirmed as working weaponized exploit code. Net: this is a genuine SSRF trust-boundary crossing that any ordinary Discord server member can reach by default, which raises exposure, but the combined complexity, timing, and transport/content constraints keep it a low-priority, monitor-grade issue rather than an urgent one.

An attacker who is an ordinary member of any Discord server running Quest Bot registers a domain with a very short DNS TTL and points it alternately at a public IP (to pass validation) and at the bot host's loopback/internal address (for the actual request). They invoke the image-to-GIF command with that domain, and after a few timing attempts the bot's second DNS lookup resolves to the internal address, causing the bot to fetch an internal HTTPS service and return any genuine image it finds. A PoC reference describing this exact methodology exists in the GHSA advisory, but it is not verified as working exploit code; the high attack complexity means several attempts may be needed to align the rebind timing.