CVE-2026-82033
Description
A flaw was found in DeepSeek Harness. A missing authentication check in its Remote Procedure Call (RPC) control plane allows an unauthorized local attacker to access internal session interfaces. By interacting with these unauthenticated endpoints, an attacker can bypass human approval safeguards, monitor active session traffic, and execute arbitrary commands with the privileges of the victim user.
Statement
This vulnerability is rated as Important severity because any unauthenticated local process can interact with the control interface to achieve arbitrary code execution under the victim's account without requiring prior privileges. Although exploitation requires existing local system access, the risk is classified as Important rather than Moderate due to the immediate bypass of approval safeguards and full compromise of confidentiality and integrity within the running user's context. Red Hat products do not distribute or support DeepSeek Harness, so standard enterprise deployments are not affected.
Mitigation
Mitigation for this issue is either not available or the currently available options do not meet the Red Hat Product Security criteria comprising ease of use and deployment, applicability to widespread installation base, or stability.
--- CME Mitigation Guidance (2026-11-06) ---
Disclaimer: The following mitigation solutions are provided as a courtesy, pending the application of an official patch. This content was generated utilizing AI tools in conjunction with data from https://cmetaxonomy.org. Customers are advised to independently evaluate and test these temporary mitigations within their own infrastructure.
The highest-value action is to enforce browser token verification across loopback HTTP and WebSocket endpoints to prevent unauthenticated access to DeepSeek Harness session RPC interfaces: upgrade to DeepSeek Harness 0.1.2-alpha.2, which introduce browser token generation and verification for the browser Host API and reject unauthenticated HTTP and WebSocket RPC requests, closing all 3 injection vectors (unauthenticated session.prompt RPC invocation, ungated event mux stream eavesdropping, tool-approval safeguard bypass via POST /api/respond). Until patched, this is a concern on Hosts running DeepSeek Harness with its web profile HTTP and WebSocket RPC control plane exposed on loopback, so Restrict local process execution permissions on the host and prevent untrusted local users from communicating with loopback HTTP and WebSocket RPC interfaces. As a compensating measure, Enforce authorization on sensitive URL paths like /api/respond and internal RPC endpoints before requests reach application logic, Deploy EDR agents to detect and block suspicious child processes or arbitrary commands spawned via the agent's bash tool, Configure auditd to log process executions and system calls resulting from commands initiated through the RPC interface, and Use Falco or eBPF runtime security to detect anomalous shell executions and unauthorized process spawning triggered by RPC invocations. These controls limit blast radius and contain exploitation, but they do not fix the underlying missing authentication defect — apply the upstream patch to fully remediate.
Highest-impact CME: CME-909 Default-Deny API Authorization Policy (high)
Understanding the Weakness (CWE)
Access Control,Other
Technical Impact: Gain Privileges or Assume Identity; Varies by Context
Exposing critical functionality essentially provides an attacker with the privilege level of that functionality. The consequences will depend on the associated functionality, but they can range from reading or modifying sensitive data, accessing administrative or other privileged functionality, or possibly even executing arbitrary code.
Frequently Asked Questions
Not sure what something means? Check out our Security Glossary.
Want to get errata notifications? Sign up here.
