Spring Tools EUVD-2026-50973

Exploitation requires that Spring Boot DevTools remote support has been explicitly enabled for a remote application target (Docker container or Cloud Foundry app) via the Spring Tools Boot Dashboard, causing the vulnerable Spring Tools (≤5.2.0) to generate the shared secret with a non-cryptographic PRNG; the attacker must also be able to reach the deployed application's DevTools remote-restart endpoint over the network. Limiting factors: DevTools remote is a development-only feature that is off by default and is officially discouraged in production, exploitation depends on successfully predicting the weak PRNG output (reflected as AC:H), and the vendor scores the vector as adjacent (AV:A) rather than fully remote, so exposure is bounded by how network-reachable the target endpoint is.

The vendor CVSS 3.1 vector AV:A/AC:H/PR:N/UI:N/S:C/C:H/I:H/A:H (8.3 High) reflects an unauthenticated but non-trivial attack (AC:H for predicting the weak secret) whose successful exploitation crosses a security boundary (S:C) to fully compromise the deployed application. No EPSS score, KEV entry, POC, or SSVC decision was provided in the input, so the only quantitative signal is CVSS itself. Real-world risk is materially constrained by the fact that this only matters when DevTools remote support is deliberately enabled against a live target - a configuration Spring explicitly warns should never be used in production - so this is more a targeted/opportunistic developer-supply-chain concern than a mass-exploitation threat. One tension to verify: the vendor chose AV:A (adjacent), yet the described targets (Cloud Foundry apps, Docker containers) are frequently network-reachable, which would argue for AV:N and a slightly higher exposure than the adjacent metric implies - confirm reachability of the DevTools remote endpoint in your environment. Absence of EPSS/KEV/POC data is missing information, not evidence of low risk.

A developer enables DevTools remote support from the Spring Tools Boot Dashboard against a Cloud Foundry app, and Spring Tools generates a predictable shared secret. An attacker who can reach the app's DevTools remote endpoint predicts or brute-forces the weakly-random secret, authenticates a restart upload, and pushes malicious classes that execute as the application, achieving code execution. No public POC is identified at time of analysis, and the attack complexity is high because it hinges on successfully predicting the PRNG output.