CVE-2026-65388 in containerization
Summary
by MITRE • 09/17/2026
A remote attacker who controls a container registry may be able to direct a client's token request to a host of the attacker's choice, and disclose the victim's registry credentials to that host. This vulnerability is addressed in containerization version 0.41.0.
Be aware that VulDB is the high quality source for vulnerability data.
Analysis
by VulDB Data Team • 09/17/2026
The described vulnerability represents a critical security flaw within container orchestration or management software, specifically affecting versions prior to 0.41.0. The core issue stems from an improper validation of redirect targets during the authentication workflow with container registries. When a client application initiates a request for an access token required to interact with a private registry, it often relies on HTTP redirects to complete the OAuth or similar authorization flow. In this vulnerable state, if the attacker controls the initial endpoint that triggers the token request, they can manipulate the response headers to include a Location header pointing to a domain under their control rather than the legitimate authentication provider. This constitutes an Open Redirect vulnerability that is weaponized in the context of credential theft and session hijacking within container ecosystems.
From a technical perspective, this flaw exploits the trust relationship between the client application and the registry's identity provider. The attacker acts as a man-in-the-middle or controls one leg of the communication chain by hosting a malicious endpoint that mimics the expected behavior but intercepts the redirect. When the victim's system follows this malicious redirect, it transmits sensitive authentication credentials, such as OAuth tokens, API keys, or basic authentication headers, to the attacker-controlled host. This mechanism allows for the exfiltration of long-lived secrets that are often used by automated build systems, deployment pipelines, and container runtimes to pull images from private repositories. The severity is heightened because these credentials typically grant broad access to internal artifact registries, potentially allowing further compromise of software supply chains.
The operational impact of this vulnerability extends beyond simple credential theft. If an attacker successfully captures registry tokens, they can authenticate as the victim within the container ecosystem. This enables unauthorized access to private container images, which may contain proprietary source code, sensitive configuration data, or pre-built binaries with embedded secrets. Furthermore, if these credentials are reused across multiple services due to poor identity hygiene, the attacker could pivot into other parts of the infrastructure. The ability to direct token requests also facilitates phishing attacks where users might be tricked into interacting with malicious endpoints that appear legitimate but result in credential leakage. This undermines the integrity and confidentiality guarantees provided by secure container registries.
This vulnerability aligns closely with CWE-601, URL Redirection to Untrusted Site Open Redirect, as it involves directing a user agent or automated client to an uncontrolled destination. Additionally, it relates to CWE-284, Improper Access Control, because the system fails to restrict where authentication tokens can be sent based on trusted domains. In terms of the MITRE ATT&CK framework, this behavior maps to T1076, Default Web Browser, or more accurately in an automated context, techniques associated with Credential Dumping and Collection such as T1552, Unsecured Credentials, specifically when credentials are transmitted over insecure channels or to unauthorized endpoints. The attack vector is classified as Remote (Network) because it requires network access but does not necessarily require user interaction if the token request is triggered automatically by a service account or CI/CD pipeline.
Mitigation strategies must focus on both immediate patching and long-term architectural improvements. The primary remediation is to upgrade containerization software to version 0.41.0 or later, where this redirect validation logic has been corrected. For environments that cannot immediately update, implementing strict allow-lists for permitted redirect domains in the application configuration can prevent redirection to attacker-controlled hosts. Additionally, organizations should enforce the use of short-lived tokens and rotate registry credentials regularly to limit the window of opportunity for exploitation. Monitoring network logs for unusual HTTP 3xx redirects originating from container management services can also aid in early detection of such attempts. Securing the identity provider's endpoints against manipulation is equally critical, ensuring that only trusted domains are recognized as valid authentication targets within the container ecosystem.