CVE-2026-16128: zevorn rt-claw SSRF Vulnerability

CVE-2026-16128 Overview

CVE-2026-16128 is a server-side request forgery (SSRF) vulnerability affecting zevorn rt-claw versions up to 0.2.0. The flaw resides in the receiver_thread function of claw/services/swarm/swarm.c within the http_request component. An attacker can manipulate HTTP request handling to force the service to issue outbound requests to attacker-chosen destinations. The vulnerability is remotely exploitable without authentication or user interaction, and a public exploit has been released. The maintainers were notified through an issue report but have not responded at the time of disclosure.

Critical Impact
Remote unauthenticated attackers can coerce the rt-claw service into issuing arbitrary HTTP requests, enabling internal network reconnaissance and interaction with internal-only endpoints.

Affected Products

zevorn rt-claw versions up to and including 0.2.0
Component: http_request handler in claw/services/swarm/swarm.c
Function: receiver_thread

Discovery Timeline

2026-07-18 - CVE-2026-16128 published to the National Vulnerability Database (NVD)
2026-07-20 - Last updated in NVD database

Technical Details for CVE-2026-16128

Vulnerability Analysis

The vulnerability is classified as [CWE-918] Server-Side Request Forgery. The receiver_thread function processes incoming HTTP requests handled by the http_request component of the swarm service. Attacker-supplied input influences the target of a subsequent outbound HTTP request without adequate validation of the destination URL or host. This allows an unauthenticated remote attacker to redirect the service's HTTP client to arbitrary internal or external addresses.

Because the exploit has been published, weaponization risk is elevated even though the EPSS probability remains modest at 0.288%. Confidentiality, integrity, and availability impacts are all rated low, reflecting bounded but real damage potential per successful request.

Root Cause

The root cause is missing validation of the URL or host parameters consumed by the http_request code path inside receiver_thread. Without an allow-list of destinations, without checks against loopback, link-local, and private IP ranges, and without protocol restrictions, the request handler acts as an open HTTP proxy for the attacker. Reference the GitHub issue tracker for the reported behavior.

Attack Vector

Exploitation occurs over the network. The attacker sends a crafted HTTP request to the rt-claw service containing a manipulated destination parameter. The receiver_thread parses the request and initiates an outbound HTTP call to the attacker-controlled target. Typical abuse paths include probing internal services, retrieving cloud metadata endpoints, and bypassing perimeter firewalls by using the vulnerable host as a pivot. See the VulDB vulnerability details and the GitHub PoC repository for further technical context.

No verified, sanitized exploit code is included here. The vulnerability mechanism is exercised by supplying a crafted URL in the HTTP request field consumed by the http_request component.

Detection Methods for CVE-2026-16128

Indicators of Compromise

Outbound HTTP connections from rt-claw hosts to internal IP ranges such as 10.0.0.0/8, 172.16.0.0/12, and 192.168.0.0/16 that deviate from normal baseline traffic.
Outbound requests from rt-claw processes to cloud metadata addresses such as 169.254.169.254.
Unusual HTTP requests to the rt-claw listener containing URL-like parameters or embedded schemes such as http://, file://, or gopher://.

Detection Strategies

Inspect application logs from the swarm service for HTTP request bodies that contain user-supplied URLs targeting private, loopback, or metadata addresses.
Build network detections that alert when the rt-claw process initiates outbound connections to destinations outside an approved allow-list.
Correlate inbound requests to the rt-claw listener with immediately subsequent outbound HTTP calls, which is the characteristic SSRF pattern.

Monitoring Recommendations

Enable egress logging on hosts running rt-claw and forward logs to a centralized analytics platform for baseline comparison.
Monitor for spikes in outbound HTTP traffic volume from rt-claw hosts, particularly to unusual destinations.
Track process-to-network relationships so that any HTTP client activity by the rt-claw binary is visible and reviewable.

How to Mitigate CVE-2026-16128

Immediate Actions Required

Restrict network egress from hosts running rt-claw to an explicit allow-list of required destinations.
Place the rt-claw listener behind authenticated network controls and remove any exposure to untrusted networks or the public internet.
Review historical logs for evidence of prior SSRF exploitation using the indicators listed above.

Patch Information

At the time of publication, the zevorn/rt-claw project has not responded to the reported issue and no vendor patch is available. Track the GitHub issue tracker and the project repository for a fix. Once released, upgrade beyond version 0.2.0 to the patched release and validate that URL validation is applied in receiver_thread.

Workarounds

Deploy an egress firewall rule that blocks rt-claw processes from reaching RFC1918 ranges, 127.0.0.0/8, 169.254.0.0/16, and other sensitive internal networks.
Front the rt-claw HTTP endpoint with a reverse proxy that validates or strips URL-typed parameters before forwarding requests.
If the affected feature is not required, disable the swarm http_request code path or stop the service until a vendor patch is available.

bash

# Example iptables egress restriction for a host running rt-claw
# Replace <RTCLAW_UID> with the UID under which rt-claw runs
iptables -A OUTPUT -m owner --uid-owner <RTCLAW_UID> -d 127.0.0.0/8 -j REJECT
iptables -A OUTPUT -m owner --uid-owner <RTCLAW_UID> -d 10.0.0.0/8 -j REJECT
iptables -A OUTPUT -m owner --uid-owner <RTCLAW_UID> -d 172.16.0.0/12 -j REJECT
iptables -A OUTPUT -m owner --uid-owner <RTCLAW_UID> -d 192.168.0.0/16 -j REJECT
iptables -A OUTPUT -m owner --uid-owner <RTCLAW_UID> -d 169.254.0.0/16 -j REJECT