UpTrain CVE-2025-27771

Exploitation requires a valid authenticated session against the UpTrain instance, corresponding to CVSS PR:L. The attacker must be able to reach the UpTrain dashboard API over the network (AV:N) and must have credentials sufficient to call the /add_prompts endpoint - any user with a valid authentication method qualifies, not just administrators. No special configuration, non-default feature enablement, or specific platform is required beyond having the UpTrain service running in version 0.7.1 or earlier. What limits exploitation: the requirement for valid credentials prevents unauthenticated internet-scale exploitation; Docker-based deployment (as recommended) constrains initial post-exploitation scope to the container unless the runtime is misconfigured with elevated privileges or host mounts.

The CVSS 4.0 base score of 7.4 (High) reflects a network-reachable, low-complexity attack requiring only low-privilege authentication (AV:N/AC:L/AT:N/PR:L/UI:N) with full confidentiality, integrity, and availability impact on the vulnerable system (VC:H/VI:H/VA:H). The E:P exploit modifier confirms proof-of-concept code is publicly available per GitHub Security Lab's GHSL-2024-200/201 advisory, elevating practical risk above the base score alone. There is no CISA KEV listing, indicating active mass exploitation has not been confirmed at time of analysis. The primary real-world risk mitigator is that exploitation requires a valid authenticated session - this limits opportunistic internet-scale exploitation but does not protect against insider threat or compromised credentials. The absence of any vendor patch significantly elevates urgency for exposed instances. For organizations where UpTrain is internet-facing or accessible to untrusted users, this should be treated as a high-priority finding. For internal-only deployments with strong access controls, risk is moderate but still warrants compensating controls given the lack of a fix.

An attacker who has obtained valid UpTrain credentials - whether through phishing, credential reuse, or insider access - sends a crafted HTTP POST request to the /add_prompts endpoint with a malicious payload embedded in the checks or metadata parameter. The server-side code processes the parameter without sanitization, causing the injected code to execute in the context of the running application process. Since proof-of-concept code exists per the GitHub Security Lab advisory (E:P), the attacker does not need to independently develop the exploit, lowering the skill barrier significantly.