CVE-2026-53869: Hermes Agent DNS Rebinding WebSocket Vulnerability (v0.16.0)
Hermes Agent versions before 0.16.0 have a DNS rebinding vulnerability affecting WebSocket communication endpoints. Attackers can trick the application into accepting requests from attacker-controlled domains by manipulating DNS responses, then use those connections to inject commands or extract sensitive terminal output. The vulnerability exists because security validation middleware is skipped for WebSocket upgrade requests on specific API endpoints.
Source data · NVD / CISA · public domain
- CVSS
- 3.1 · 7.5 HIGH · CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H
- Weaknesses (CWE)
- CWE-306
- Affected products
- 0 configuration(s)
- Published / Modified
- 2026-06-17 / 2026-06-18
NVD description (verbatim)
Hermes Agent before 0.16.0 contains a DNS rebinding vulnerability in WebSocket endpoints that allows remote attackers to bypass Host and Origin validation. FastAPI HTTP middleware does not execute for WebSocket upgrade requests on /api/pty, /api/ws, /api/pub, and /api/events endpoints, enabling attackers to exploit DNS rebinding and inject malicious commands or read terminal output.
5 reference(s) · View on NVD →
SEC.co analysis · AI-assisted, reviewed against source
Technical summary
CVE-2026-53869 is a DNS rebinding vulnerability in Hermes Agent's WebSocket implementation. FastAPI HTTP middleware that enforces Host and Origin validation does not execute for WebSocket upgrade requests on the /api/pty, /api/ws, /api/pub, and /api/events endpoints. An attacker can exploit this by initiating a DNS rebinding attack—first resolving attacker.com to the victim's IP, establishing a WebSocket connection, then updating DNS to resolve to the attacker's server for subsequent requests. This allows the attacker to bypass hostname/origin checks and interact with protected endpoints from the browser's perspective as same-origin. The vulnerability is classified as CWE-306 (Missing Authentication for Critical Function).
Business impact
This vulnerability enables unauthorized command injection and information disclosure on systems running vulnerable Hermes Agent instances. An attacker can execute arbitrary commands through the terminal API endpoint (/api/pty) or eavesdrop on session data through the publish/events endpoints (/api/pub, /api/events). For organizations using Hermes Agent for remote administration or monitoring, this represents a direct path to lateral movement, privilege escalation, or data exfiltration without needing credentials. The attack surface includes any user whose browser visits an attacker-controlled website while authenticated to a vulnerable Hermes Agent instance.
Affected systems
Hermes Agent versions prior to 0.16.0 are affected. The vulnerability specifically impacts the WebSocket endpoints: /api/pty (pseudo-terminal), /api/ws (generic WebSocket), /api/pub (publish), and /api/events (event streaming). Organizations should audit deployed instances to identify running versions and network exposure of these endpoints.
Exploitability
This vulnerability has a CVSS 3.1 score of 7.5 (HIGH) with a network attack vector, low complexity, and no authentication or user interaction required. However, practical exploitation requires the target user to visit attacker-controlled content while maintaining an active authenticated session with Hermes Agent. The attack is not trivial—it requires DNS rebinding conditions and browser-based attack delivery—but once conditions align, the impact is severe: unauthenticated command execution or data access.
Remediation
Upgrade Hermes Agent to version 0.16.0 or later. This version applies FastAPI middleware security checks to WebSocket upgrade requests, closing the validation bypass. Organizations should also consider network-level mitigations: restrict WebSocket endpoint access via firewall rules, use DNS pinning or DNSSEC validation on client systems, and deploy Content Security Policy (CSP) headers to limit cross-origin WebSocket initiation.
Patch guidance
Upgrade to Hermes Agent 0.16.0 immediately for affected deployments. Verify patch application by checking the running version via the agent's health/version endpoint. Test WebSocket connectivity to monitored endpoints after patching to ensure operational continuity. For organizations unable to patch immediately, isolate Hermes Agent instances to trusted networks only and require VPN access, reducing browser-based attack surface.
Detection guidance
Monitor for unusual WebSocket connection attempts to /api/pty, /api/ws, /api/pub, and /api/events endpoints with mismatched Host or Origin headers. DNS rebinding attacks may present as legitimate connections followed by rapid DNS resolution changes. Check Hermes Agent logs for commands or data access from unexpected sources. Review proxy/load balancer logs for WebSocket upgrade requests lacking proper authentication context. Network-level detection should flag DNS queries where the same hostname resolves to multiple IPs within a short timeframe targeting your infrastructure.
Why prioritize this
This vulnerability merits immediate attention due to its HIGH CVSS score, unauthenticated attack vector, and critical impact (command injection and data disclosure). Although practical exploitation requires specific conditions—victim visiting attacker content while authenticated—the barrier is not prohibitively high in environments where users access untrusted websites. Given the nature of Hermes Agent (remote administration/monitoring), compromise enables full system control and is difficult to detect post-exploitation.
Risk score, explained
The CVSS 3.1 score of 7.5 reflects network-based access, low attack complexity, and no authentication requirement. The score emphasizes availability impact (denial via command execution) but should be contextualized as also carrying high confidentiality and integrity risk through command injection and data exfiltration capabilities. Organizations with Hermes Agent exposed to untrusted users or networks should treat this as a critical remediation priority despite the 7.5 rating.
Frequently asked questions
Does this vulnerability require the user to click a link or take action?
The attacker must trick a user into visiting attacker-controlled content while they maintain an active authenticated session with Hermes Agent. No additional clicks are typically needed once the page loads; the attack executes via JavaScript in the browser. Users who never visit untrusted websites or who log out of Hermes Agent sessions before browsing elsewhere face lower risk.
Can this be exploited from outside our network if Hermes Agent is behind a firewall?
If Hermes Agent is restricted to internal network access only, the risk is significantly lower. However, if the agent is exposed to the internet or accessible via VPN/remote access, and a user visits an attacker website from that same network context, exploitation remains possible. The vulnerability is primarily a concern for agents with web-facing endpoints.
Does version 0.16.0 fully resolve the DNS rebinding issue?
Yes, version 0.16.0 applies FastAPI middleware validation to WebSocket upgrade requests, enforcing Host and Origin checks before connection establishment. This closes the bypass. Verify deployment of 0.16.0 or later and confirm through version checks and security logs post-patch.
What does CWE-306 mean in this context?
CWE-306 (Missing Authentication for Critical Function) indicates that the /api/pty, /api/ws, /api/pub, and /api/events endpoints lacked proper authentication enforcement due to middleware bypass. The vulnerability allows critical functions (command execution, data access) to be invoked without valid authentication tokens.
This analysis is based on the CVE record published 2026-06-17 and modified 2026-06-18. No public exploit or proof-of-concept is known to be in active use at the time of writing. Organizations should verify all patch version numbers and deployment guidance against official Hermes Agent vendor advisories. This page does not constitute professional security advice; consult your security team or a qualified vendor for environment-specific risk assessment and remediation planning. CVSS scores are indicators of severity; actual risk depends on network architecture, authentication controls, and user behavior. Source: NVD (public-domain), retrieved 2026-07-27. Analysis generated by SEC.co (claude-haiku-4-5).
Weaknesses (CWE)
Related vulnerabilities
- CVE-2018-25437HIGHCherryFramework Themes Information Disclosure Vulnerability
- CVE-2023-54350HIGHWordPress Augmented-Reality Plugin Remote Code Execution
- CVE-2026-10243HIGHSmart Parking System 1.0 Authentication Bypass – Remote Admin Access
- CVE-2026-10281HIGHEnderfga claw-orchestrator Authentication Bypass – Patch Available
- CVE-2026-10617HIGHGoClaw Webhook Authentication Bypass – Remote Exploitation
- CVE-2026-12199HIGHNLTK WordNet Browser Remote Shutdown Vulnerability (7.5 CVSS)
- CVE-2026-24088HIGHQualcomm Bootloader Cryptographic Verification Flaw (CVSS 8.2)
- CVE-2026-24090HIGHQualcomm Partition Table Cryptographic Flaw Enables Boot Modification