PatchSiren

PatchSiren cyber security CVE debrief

CVE-2026-92972 sgl-project CVE debrief

CVE-2026-92972 is a high-severity vulnerability in SGLang through 0.5.19 that allows attackers to poison the KV transfer routing table via an unauthenticated PUT /route endpoint, potentially causing denial of service or disclosure of KV transfer metadata including session identifiers and tensor-parallel topology parameters. This vulnerability affects SGLang deployments and requires immediate attention from defenders and administrators to assess exposure and prioritize remediation. The vulnerability is caused by the lack of authentication on the prefill bootstrap service, allowing attackers to supply arbitrary rank_ip and rank_port values to redirect decode workers to attacker- -

Vendor
sgl-project
Product
sglang
CVSS
HIGH 8.8
CISA KEV
Not listed in stored evidence
Original CVE published
2026-09-17
Original CVE updated
2026-09-22
Advisory published
2026-09-17
Advisory updated
2026-09-22

Who should care

Defenders and administrators using SGLang should assess exposure and prioritize remediation due to the high-severity nature of this vulnerability. SGLang users must verify their version, restrict access to the prefill bootstrap service, and monitor for suspicious activity. This vulnerability impacts SGLang operators, platform administrators, vulnerability management teams, and security teams responsible for ensuring the security and integrity of their -

Why it matters

CVE-2026-92972 is a high-severity vulnerability in SGLang that allows attackers to poison the KV transfer routing table, causing denial of service or disclosure of metadata. Defenders and administrators using SGLang should assess exposure and prioritize remediation.

  • Denial of service due to redirecting decode workers to attacker-controlled endpoints
  • Disclosure of KV transfer metadata including session identifiers and tensor-parallel topology parameters

Technical summary

The vulnerability is caused by an unauthenticated PUT /route endpoint on the prefill bootstrap service in SGLang through 0.5.19. This allows attackers to supply arbitrary rank_ip and rank_port values to redirect decode workers to attacker-controlled endpoints, potentially causing denial of service or disclosure of KV transfer metadata including session identifiers and tensor-parallel topology parameters. The vulnerability affects SGLang deployments and requires immediate attention from defenders and administrators.

Defensive priority

Defenders should prioritize verifying and remediating this vulnerability, especially in environments using SGLang.

Recommended defensive actions

  • Verify SGLang version and check for updates
  • Restrict access to the prefill bootstrap service
  • Monitor for suspicious activity on the KV transfer routing table
  • Implement compensating controls for exposed systems
  • Conduct exposure review for managed environments
  • Review asset inventory for affected systems
  • Track exceptions and retest remediated assets

Evidence notes

The vulnerability is described in the CVE record and NVD entry. The source references provide additional details about the vulnerability and its impact. Evidence is limited to public CVE and NVD records. Defenders should verify SGLang version and check for updates, restrict access to the prefill bootstrap service, and monitor for suspicious activity on the KV transfer routing table. No additional evidence is available beyond CVE and NVD records.

Sources and references

Verified primary and authoritative sources

  • CVE-2026-92972 CVE Program record

    Publisher, destination, and source semantics verified

    URL: https://www.cve.org/CVERecord?id=CVE-2026-92972

    CVE Program - Official CVE Program record with source-provided CVE metadata.

  • CVE-2026-92972 NVD vulnerability detail

    Publisher, destination, and source semantics verified

    URL: https://nvd.nist.gov/vuln/detail/CVE-2026-92972

    NIST National Vulnerability Database - Official NIST NVD detail page and source-specific vulnerability assessment.

Supplemental references

Methodology and review provenance

AI-assisted synthesis based on stored public vulnerability evidence. System validation, approval state, and publication status do not by themselves establish human review of this revision. PatchSiren helps prioritize defensive review and does not prove exposure or remediation on any system.