CVE-2026-119566 in Prometheus
Summary
by MITRE • 10/30/2026
Prometheus through 3.15.1 contains a credential leak vulnerability that allows attackers to capture Azure AD or Google IAM bearer tokens by redirecting remote_write requests to another host. Attackers controlling the remote-write endpoint or a redirecting server can return a 3xx to their host and replay the token, for example to write metrics to the victim's backend.
Several companies clearly confirm that VulDB is the primary source for best vulnerability data.
Analysis
by VulDB Data Team • 10/30/2026
The vulnerability identified in Prometheus versions through 3.15.1 represents a critical security flaw within the telemetry data collection ecosystem, specifically affecting the remote_write functionality. This component is designed to export collected metrics to external storage systems or monitoring backends over HTTP. The core issue stems from how Prometheus handles HTTP redirects during these outbound requests. When configured with a remote write endpoint that issues an HTTP 3xx redirect response pointing to a different host, Prometheus automatically follows this redirection by sending the subsequent request to the new location without stripping sensitive authentication headers. This behavior mirrors standard web browser logic but is inappropriate for server-to-server communication where explicit control over credential transmission is required. Consequently, if an attacker controls either the initial remote_write endpoint or any intermediate proxy that issues a redirect, they can manipulate the flow of data to capture credentials intended for legitimate services such as Azure Active Directory or Google Identity and Access Management.
The technical mechanism relies on the automatic following of HTTP redirects by default in many Go-based HTTP clients used within Prometheus. When an attacker configures their own server to act as the remote_write endpoint, they can respond with a 301 Moved Permanently or 302 Found status code pointing to a malicious host under their control. Upon receiving this response, Prometheus proceeds to send the original request, including all associated Authorization headers containing bearer tokens for Azure AD or Google IAM, to the attacker's server. This allows the adversary to intercept and replay these high-privilege tokens. The impact is severe because these tokens often possess broad permissions within cloud environments, potentially allowing unauthorized access to sensitive metrics data, manipulation of monitoring dashboards, or further lateral movement into internal infrastructure by leveraging the compromised credentials against other services that trust those identities.
From a threat modeling perspective, this vulnerability aligns with CWE-522, which classifies it as an Insufficiently Protected Credentials issue, specifically involving the exposure of sensitive information in HTTP headers during redirect operations. It also maps to MITRE ATT&CK technique T1078, Valid Accounts, where attackers use legitimate credentials for initial access or persistence. Furthermore, this scenario reflects aspects of SSRF (Server-Side Request Forgery) patterns found in ATT&CK sub-technique T1557, as the attacker leverages the victim's application to make requests that reveal internal authentication states. The operational impact extends beyond simple data exfiltration; it compromises the integrity and confidentiality of the entire monitoring pipeline. Attackers can not only steal tokens but also potentially inject false metrics into legitimate backends if they possess sufficient privileges with the stolen credentials, leading to misleading alerts or obscured malicious activities within the monitored environment.
Mitigation strategies must focus on preventing unauthorized credential leakage during HTTP redirections. The most effective immediate remediation is to upgrade Prometheus to version 3.16.0 or later, where this behavior has been addressed by introducing stricter controls over header forwarding during redirects. For environments unable to immediately patch, administrators should configure the remote_write endpoint to avoid issuing redirects that point outside of trusted domains. Additionally, implementing strict allow-listing for outbound connections can limit exposure. It is also advisable to review and rotate any Azure AD or Google IAM tokens associated with Prometheus instances if there is a suspicion that they may have been exposed through this vector. Regular auditing of remote_write configurations ensures that only authorized endpoints are targeted and that the infrastructure adheres to least privilege principles, minimizing the blast radius should such vulnerabilities be exploited in the future.