Live Exposure Defense: From CVE to Confirmed Exposure in 12 Hours – See more

Back
New CVE Detected

CVE-2026-102634 – Unauthenticated Denial of Service – SGLang ≤ 0.5.20

Be the first to know when new zero-days emerge:

Subscribe to Email Alerts
Subscribe to RSS Feed

Summary

CVE-2026-102634 is a denial-of-service vulnerability in SGLang, the open-source LLM inference/serving engine, affecting all versions through 0.5.20. The flaw resides in SGLang’s prefill/decode (PD) disaggregation mode when using the Mooncake KV transfer backend, where duplicate bootstrap_room identifiers in /generate requests are not validated. An unauthenticated remote attacker can exploit this to crash scheduler processes or hang other users’ in-flight requests, and the issue carries a High severity rating of 8.7.

Technical details

  • Root cause: The bootstrap_room field in GenerateReqInput is bound directly from the client-supplied request body with no ownership or uniqueness validation before being used to register KV transfer sessions (CWE-694, Use of Multiple Resources with Duplicate Identifier).
  • Trigger conditions: The target must be running SGLang in prefill/decode disaggregation mode with the Mooncake KV transfer backend, and exposed directly (not solely behind a router/gateway, which overwrites room values and is not affected).
  • Attack vector: An attacker sends two or more concurrent, unauthenticated /generate requests carrying an identical bootstrap_room value. Both prefill and decode sides register sessions under the same room without checking whether it is already in use.
  • Impact: When the first request completes, its room teardown affects the second session’s in-flight transfer, causing unhandled exceptions (KeyError on the prefill side, ValueError: Unexpected poll case on the decode side) that can crash scheduler processes. Other legitimate users’ requests sharing affected infrastructure can hang for roughly 60 seconds until a transfer timeout is reached, and repeated abuse can force engine termination.

Affected software

  • SGLang (sgl-project/sglang), all versions from 0 through 0.5.20 (inclusive), when deployed with prefill/decode disaggregation and the Mooncake KV transfer backend.

Severity

  • CVSS score: 8.7 (High)
  • CVSS 4.0 vector: CVSS:4.0/AV:N/AC:L/AT:N/PR:N/UI:N/VC:N/VI:N/VA:H/SC:N/SI:N/SA:N

Mitigation and recommended actions

  • Immediate: No fixed release has been published as of this writing; track the upstream SGLang repository and issue tracker for a patched version and upgrade as soon as one is available.
  • If no patch is available:
    • Avoid exposing prefill/decode engine endpoints directly to untrusted networks; place a router/gateway in front of disaggregated deployments, since the router overwrites bootstrap_room values and is not affected by this issue.
    • Restrict access to /generate endpoints on PD-disaggregated SGLang deployments using network-level controls (firewalls, allow-lists, authentication proxies) until a fix is applied.
    • Monitor scheduler processes for unexpected crashes or restarts and for anomalous concurrent requests carrying identical bootstrap_room values.

References

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report

How IONIX’s External Exposure Management Platform Detects and Validates
Zero-Days to Shrink MTTR

1

Map your entire attack surface (continously)

IONIX uses multi-factor discovery methods, including DNS analysis, certificate mapping, metadata inspection, and more, to automatically map every internet-facing asset across your environment. This includes cloud instances, third-party platforms, shadow IT, and even forgotten infrastructure that traditional tools miss.

2

Monitor for new CVEs

Dozens of threat intel feeds using agentic technology are continuously analyzed to detect the appearance of proof-of-concept code, exploit kits, and indicators of active targeting. IONIX goes further by applying AI to proactively evaluate whether emerging vulnerabilities are likely to be exploited, even before PoCs go public.

3

Identify Potential External Exposures

Not all CVEs matter. IONIX filters vulnerabilities by asking attacker-centric questions: Can it be reached from the internet? Does it require authentication? Is it being exploited in the wild? This dramatically reduces noise and focuses teams on threats that can actually be weaponized.

4

Create Safe, Scalable Exploit Validations

IONIX transforms real-world PoCs into safe, non-intrusive test payloads that can be run in production environments without disruption. These simulations are precisely targeted to the systems that are vulnerable, ensuring rapid validation without unnecessary load.

5

Execute Exploit Validations

By combining context about software stack, versioning, exposure status, and reachability, IONIX ensures that only the right payloads are executed against the right assets, maximizing efficiency and minimizing risk.

6

Drive Fast and Actionable Remediation

Results are routed through integrations with ticketing, SOAR, and SIEM tools. Issues are written in plain language, bundled into remediation clusters, and prioritized based on asset criticality, exploitability, and blast radius. This shortens mean time to remediation (MTTR) and empowers teams to act with confidence.

Are you exposed?

Get a free report of your organization’s exposure to this CVE and threat

  • Mapping of all Assets with this Technology
  • Identification of potentially exposed assets to this CVE
  • Confirmation of verified exploitable assets
Get Exposure Report
Close

Subscribe to Threat Center RSS

Copy/paste the link below into your preferred RSS reader or follow these instructions to subscribe to Slack alerts.

Open RSS feedSlack setup instructions

Close

Get Real-Time CVE Alerts to Your Email

Be the first to know when new zero-days emerge