CVE-2026-90620: HexStrike AI Auth Bypass Vulnerability

CVE-2026-90620 Overview

CVE-2026-90620 is a missing authentication vulnerability [CWE-287] in 0x4m4 HexStrike AI, affecting builds up to commit d689933ff579d839c676c82b231f8e98326c5f04. The flaw resides in an unspecified function within hexstrike_server.py inside the API Command Endpoint component. Remote attackers can reach the endpoint over the network without credentials, user interaction, or elevated privileges. The exploit technique has been publicly disclosed. Because the project uses continuous delivery with rolling releases, no discrete affected or fixed version numbers are published. The maintainer was notified through a GitHub issue but has not responded at the time of publication.

Critical Impact
Unauthenticated network attackers can invoke API command functionality on exposed HexStrike AI servers, affecting confidentiality, integrity, and availability of the host application.

Affected Products

0x4m4 HexStrike AI (rolling release) up to commit d689933ff579d839c676c82b231f8e98326c5f04
hexstrike_server.py API Command Endpoint component
Any deployment tracking the upstream main branch prior to a maintainer-issued fix

Discovery Timeline

2026-09-14 - CVE-2026-90620 published to NVD
2026-09-14 - Last updated in NVD database

Technical Details for CVE-2026-90620

Vulnerability Analysis

HexStrike AI exposes an API Command Endpoint implemented in hexstrike_server.py. The endpoint does not enforce an authentication check before processing incoming requests. Because HexStrike AI is designed to orchestrate offensive security tooling, the command surface reachable through this endpoint has direct operational value to an attacker. Remote clients can interact with server functionality that should be restricted to authorized operators. The disclosed weakness maps to [CWE-287] Improper Authentication.

Root Cause

The root cause is the absence of an authentication layer on a network-facing API route inside hexstrike_server.py. Requests reach handler logic without a session token, API key, or transport-layer client validation. As a result, the server treats anonymous callers as trusted operators.

Attack Vector

Exploitation requires only network reachability to the HexStrike AI server. An attacker sends HTTP requests to the unauthenticated API Command Endpoint and receives handler responses without providing credentials. No user interaction is needed and complexity is low. Public disclosure of the exploit technique lowers the barrier for opportunistic scanning and abuse. See the GitHub Issue Discussion and the VulDB Vulnerability Details for the technique write-up.

No verified proof-of-concept code is included here. Refer to the linked references for technical specifics.

Detection Methods for CVE-2026-90620

Indicators of Compromise

Inbound HTTP requests to the HexStrike AI API Command Endpoint originating from untrusted networks or unexpected client addresses.
Successful command responses returned to clients that never presented an authentication header, session cookie, or API key.
Unexpected child processes or offensive tool invocations launched by the HexStrike AI server process outside operator activity windows.

Detection Strategies

Enable access logging on hexstrike_server.py and alert on any request to command routes lacking an authenticated principal.
Deploy a reverse proxy in front of HexStrike AI that enforces authentication and logs every request path for retrospective review.
Baseline expected operator IP ranges and generate alerts on API traffic from outside that allow list.

Monitoring Recommendations

Forward server and proxy logs to a centralized analytics platform and monitor request-rate spikes against the API Command Endpoint.
Track outbound network connections from the HexStrike AI host to identify tool executions triggered by unauthenticated callers.
Review host process trees for unexpected shell, scanner, or exploitation binaries spawned by the server process.

How to Mitigate CVE-2026-90620

Immediate Actions Required

Restrict network exposure of HexStrike AI so the API Command Endpoint is not reachable from untrusted networks or the public internet.
Place the service behind an authenticating reverse proxy that requires mutual TLS, an API key, or single sign-on before forwarding requests.
Audit recent access logs for anonymous requests to command routes and investigate any unexpected tool executions on the host.

Patch Information

No vendor patch is available at the time of publication. The project uses continuous delivery with rolling releases and, according to the CVE record, has not yet responded to the issue report tracked in the GitHub Issue Discussion. Operators should monitor the upstream repository for a commit that introduces authentication on the API Command Endpoint and rebuild from source once available.

Workarounds

Bind the HexStrike AI listener to 127.0.0.1 or a private management interface and access it only through an authenticated tunnel such as SSH or a VPN.
Enforce host-based firewall rules that permit inbound traffic to the API port only from designated operator workstations.
Run HexStrike AI inside an isolated network segment or container with egress restrictions to limit blast radius if the endpoint is abused.

bash

# Configuration example: restrict HexStrike AI API exposure with iptables
# Allow only a trusted operator subnet to reach the API port
iptables -A INPUT -p tcp --dport 8888 -s 10.10.20.0/24 -j ACCEPT
iptables -A INPUT -p tcp --dport 8888 -j DROP