SGLang Router Socket Deserialization RCE Vulnerability

SGLangs multimodal generation runtime scheduler's ROUTER socket binds to 0.0.0.0 by default and contains a sink that calls pickle.loads() on incoming messages, enabling RCE when exposed to the internet.

Casky was already ahead

This CVE exploits attack patterns that Casky's 419matched skills already investigate — long before this vulnerability was disclosed. Claude's reasoning model maps these techniques to MITRE ATT&CK, so practitioners who ran these skills have already seen the threat behaviour in their findings.

Analysis

SGLang's multimodal generation runtime scheduler contains a critical remote code execution vulnerability stemming from unsafe deserialization. The ROUTER socket binds to 0.0.0.0 by default, exposing the service to network access, and processes incoming messages through pickle.loads() without validation. This combination allows unauthenticated attackers to execute arbitrary code on affected systems. The vulnerability impacts organizations deploying SGLang for AI inference workloads, particularly those with internet-facing or insufficiently segmented deployments. With a CVSS score of 9.8, this represents a critical risk requiring immediate patching and network isolation.

Casky's 419 matching security skills leverage Claude's extended reasoning to detect the attack patterns associated with this vulnerability across three key MITRE ATT&CK techniques: Initial Access (TA0001) through exposed network services, Execution (TA0002) via malicious pickle payload delivery, and Resource Development (TA0043) where attackers probe for vulnerable SGLang instances. Practitioners using Casky would observe findings that correlate network reconnaissance of port 5555 (ROUTER socket), detection of pickle-serialized payloads in network traffic, and process execution anomalies following socket connections. The platform's AI-driven analysis can identify attack chains—from reconnaissance to code execution—that traditional signature-based detection might miss, helping security teams understand not just that SGLang is vulnerable, but how threat actors would exploit it in their environment.

Live Threat IntelligenceComing soon

Casky Risk Score

—

EPSS Probability

—

Shodan Exposure

—

GreyNoise Signal

—

Composite risk scoring from EPSS, CISA KEV, Shodan, and GreyNoise — 21 security APIs correlated into a single Casky Risk Score. Coming in Casky Pro. Join early access →

419 Casky skills cover this attack pattern

These skills use Claude AI's reasoning model to surface findings in the same attack categories as CVE-2026-7301.

analyzing-android-malware-with-apktool

malware analysis · medium

Run

analyzing-apt-group-with-mitre-navigator

threat intelligence · low

Run

analyzing-bootkit-and-rootkit-samples

malware analysis · medium

Run

MITRE ATT&CK Techniques

TA0001 TA0043 TA0002

Investigate the attack patterns behind CVE-2026-7301

Casky has 419 skills that investigate the attack patterns behind CVE-2026-7301. Run one and get CVSS-scored findings in 3 minutes.

Run the skill that detects this →

X Instagram LinkedIn