CVE-2026-45796: Unauthenticated SSRF in Coder Remote Development Platform

Technical summary

The vulnerability exists in the `/api/v2/workspaceagents/azure-instance-identity` endpoint, which is designed to verify Azure instance identities via PKCS#7 signature validation. An attacker can submit a crafted signature that causes the Coder server to issue GET requests to attacker-controlled destinations. The server's error handling inadvertently leaks information through HTTP status codes and failure messages, allowing semi-blind SSRF exploitation. The flaw stems from insufficient validation of the signature submission mechanism and failure to properly isolate the Azure identity verification process from direct network requests. CWE-918 (Server-Side Request Forgery) classification applies.

Business impact

Organizations using Coder for remote development environment provisioning face two key risks. First, attackers can perform internal network reconnaissance to map infrastructure, identify cloud metadata endpoints (a common attack vector), and discover vulnerable internal services without triggering typical authentication logs. Second, in multi-tenant deployments, this could be leveraged to probe other organizations' infrastructure if deployed in shared cloud environments. The attack requires no authentication, making it accessible to any external attacker who can reach the Coder API. For organizations with strict network segmentation or those handling sensitive development environments, this reconnaissance capability represents a meaningful risk to network security posture.

Affected systems

Coder versions prior to 2.24.5, 2.29.13, 2.30.8, 2.31.12, 2.32.2, and 2.33.3 are vulnerable. This indicates multiple release tracks—the version numbering suggests at least five parallel maintenance branches. Organizations running any Coder instance below their respective branch's patched version are affected. The vulnerability is present regardless of whether Azure identity authentication is actively used, though it is only exploitable if the vulnerable endpoint remains exposed.

Exploitability

Exploitability is high from an attacker's perspective. No authentication is required, the attack surface is a public API endpoint, and exploitation requires only HTTP POST requests with a crafted PKCS#7 signature—a low technical barrier. However, impact is constrained by the semi-blind nature of the SSRF: attackers receive no response body, only timing and error-message information. This limits direct data exfiltration but enables effective host discovery and service fingerprinting. The CVSS 3.1 score of 6.5 (MEDIUM) reflects this balance—easy exploitation but limited direct impact. No indication of active exploitation in the wild has been reported, and the vulnerability was not added to the KEV catalog.

Remediation

Organizations should immediately upgrade to patched versions: 2.24.5, 2.29.13, 2.30.8, 2.31.12, 2.32.2, or 2.33.3, depending on their current release track. For deployments where immediate patching is not feasible, implement network-level controls: restrict access to the `/api/v2/workspaceagents/azure-instance-identity` endpoint using ingress firewall rules or API proxy ACLs, allowing only trusted internal sources if the Azure identity mechanism is in use. If Azure identity authentication is not required, disable or remove the endpoint entirely from your deployment.

Patch guidance

Identify your current Coder version and determine which release track it belongs to. Cross-reference against the patched versions provided: 2.24.5, 2.29.13, 2.30.8, 2.31.12, 2.32.2, and 2.33.3. Upgrade within your track to the patched version. Test the upgrade in a non-production environment first, particularly for remote development environments where outages directly impact developer productivity. Review your Azure instance identity authentication configuration post-upgrade to ensure it functions as expected. No breaking changes have been reported in these patch releases.

Detection guidance

Monitor for POST requests to `/api/v2/workspaceagents/azure-instance-identity` from unexpected sources or with unusual request patterns. Specifically, look for requests with invalid or suspicious PKCS#7 signatures, repeated requests from the same external IP, or requests followed by error responses. Correlate with subsequent outbound HTTP requests from the Coder server to internal or non-standard destinations. Enable HTTP logging on your Coder deployment and on any upstream proxy or WAF to capture the full request/response flow. Analyze error response rates for this endpoint—a spike in 400/403 errors may indicate exploitation attempts. Use DNS and HTTP traffic analysis to detect outbound reconnaissance requests generated by exploited SSRF vulnerabilities.

Why prioritize this

This vulnerability merits prompt but measured attention. The MEDIUM severity (6.5 CVSS) and lack of KEV listing indicate it is not an emergency, but the unauthenticated, network-accessible nature of the flaw means it should be prioritized within your standard patch cycle. For organizations running Coder instances exposed to the internet or accessible via VPN, this should be addressed within 1–2 weeks. For internal-only deployments with strict network segmentation, prioritization can extend to the standard quarterly patch window. The reconnaissance capability makes this a risk to network mapping and cloud metadata discovery, not direct data loss.

Risk score, explained

The CVSS 3.1 score of 6.5 reflects unauthenticated network access (AV:N), low complexity (AC:L), no privileges required (PR:N), no user interaction (UI:N), and impact limited to confidentiality and integrity of reconnaissance data only (C:L, I:L, A:N). The 'I:L' (integrity low) designation reflects that attackers can influence the Coder server's outbound requests and potentially poison internal caches or monitoring systems, though no direct system modification occurs. The absence of availability impact acknowledges that while DoS is theoretically possible via resource exhaustion, the flaw itself does not grant denial-of-service capabilities. This score appropriately balances the ease of exploitation against the constrained impact.

Frequently asked questions