Unleash CVE-2026-63004

Exploitation requires an authenticated session holding the root-level CREATE_ADDON or UPDATE_ADDON permission within Unleash - this is an administrative capability not granted to standard users by default. The attacker must be able to create or modify an addon/integration configuration (Webhook, Slack, Microsoft Teams, Datadog, or New Relic) supplying an attacker-controlled URL, and a feature-flag event must subsequently fire to trigger the outbound HTTP request from the server. For the most severe credential-theft scenario, the Unleash server must be deployed in an environment where cloud metadata endpoints (e.g., AWS IMDS at 169.254.169.254) or other sensitive internal HTTP services are reachable from the server's network interface. Limiting factors: high-privilege permission required (PR:H); impact on confidentiality is indirect - the attacker receives only event status for blind network probing unless header secrets (Authorization, DD-API-KEY, customHeaders) are configured in the integration and forwarded to an attacker-observed destination; non-cloud on-premises deployments with egress filtering face substantially reduced practical impact.

The CVSS 3.1 vector (AV:N/AC:L/PR:H/UI:N/S:C/C:L/I:L/A:N, score 5.5) accurately reflects that exploitation requires high privileges, but the Changed scope signals the most consequential aspect: the server crosses network trust boundaries on behalf of the attacker. Real-world risk is highly deployment-context-dependent - a cloud-hosted Unleash instance with an accessible AWS IMDS, GCP metadata server, or internal Kubernetes API elevates the practical confidentiality impact well above the C:L rating, because forwarded Authorization or DD-API-KEY headers to an attacker-controlled host can yield persistent credential access. Conversely, on-premises deployments with strict egress filtering face substantially reduced exposure. EPSS data was not provided in the available intelligence, and Unleash is not listed in the CISA KEV catalog, indicating no confirmed widespread active exploitation at time of analysis. The PR:H requirement meaningfully limits the attack surface to administrative accounts, though in multi-tenant or SaaS deployments where addon permissions are granted broadly, this boundary may be weaker than the score implies. No conflicting signals were identified between the CVSS vector and the description.

An authenticated Unleash operator with CREATE_ADDON permission creates a Webhook integration configured with a URL pointing to the AWS Instance Metadata Service (http://169.254.169.254/latest/meta-data/iam/security-credentials/) and includes a custom Authorization header. When a feature flag event fires and the integration triggers, the Unleash server issues the request to the metadata endpoint; the attacker observes the integration event's success or failure status as a blind probe to map reachable internal addresses, and any configured Authorization or API key values are forwarded to an attacker-controlled host for harvesting. No public POC has been identified, but the attack mechanics are straightforward given the description.