Description

In JetBrains IntelliJ IDEA before 2026.3 blind SSRF was possible via remote references in JSON Schema validation

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "4.3",
  "pubDate": "2026-11-06T17:17:23.423Z",
  "pubdate": "2026-11-06T17:17:23.423Z",
  "executiveSummary": "A blind Server-Side Request Forgery (SSRF) vulnerability exists in JetBrains IntelliJ IDEA versions prior to 2026.3. The flaw resides within the IDE's JSON Schema validation subsystem, which improper handles external network references. An attacker can exploit this weakness by embedding remote schema references within JSON files processed by the application, causing the IDE to automatically execute arbitrary outbound HTTP or HTTPS requests without proper authorization or filtering.\nThe primary impact of this vulnerability involves out-of-band network request manipulation from the environment running the IDE. While classified as a blind SSRF—meaning the response data from target endpoints is not directly rendered back to the user interface—an attacker can leverage this behavior to probe internal networks, map private infrastructure, detect active services behind firewalls, and interact with accessible local network resources or cloud metadata services. Exploitation typically requires a victim to open or trigger validation on a untrusted project or JSON document containing malicious schema pointers within an affected version of JetBrains IntelliJ IDEA.",
  "technicalDetails": "The root cause of this vulnerability lies in the automatic resolution mechanism of remote URI references during JSON Schema evaluation within JetBrains IntelliJ IDEA versions prior to 2026.3. Standard JSON Schema specifications utilize the '$ref' property to allow schema definitions to import external schema files. In vulnerable versions of the IDE, the integrated JSON validation engine resolves these URI references dynamically over the network. The implementation lacks sufficient network access controls, domain allowlists, or protocol restrictions, enabling the parser to initiate outbound connections to arbitrary network destinations specified in schema definitions.\nThe attack flow proceeds through a sequence of automated parsing actions triggered by user interaction with untrusted codebases or files:\n1. Schema Construction: The attacker crafts a malicious JSON file or schema document containing a remote reference pointer (e.g., '$ref': 'http://169.254.169.254/latest/meta-data/' or an attacker-controlled external host).\n2. File Ingestion: The target user opens the project or imports the file into JetBrains IntelliJ IDEA, triggering the automatic background indexing and validation subsystems.\n3. Reference Resolution: The JSON schema parser parses the document and encounters the remote '$ref' URI key. Without verifying if the target destination is trusted or local, the parser delegates an HTTP/HTTPS fetch operation to the application's network layer.\n4. Outbound Request Execution: JetBrains IntelliJ IDEA generates an out-of-band network request to the specified target IP address or hostname from the developer host environment.\n5. Blind Data Transmission & Probing: Because the validation component processes the response internally without outputting the full payload body to the developer console, the SSRF remains blind. However, the attacker can observe inbound connections on controlled infrastructure to confirm active targets, measure request timing to discover open internal ports, or trigger side-effect actions on internal services accessible only from the host system.\nThis vulnerability operates under the execution context of the local IDE user, requiring no special network privileges prior to ingestion. The flaw affects all installations of JetBrains IntelliJ IDEA before version 2026.3 where JSON Schema validation is active."
}