Security firm Xint says it has uncovered five critical vulnerabilities in the NGINX web server, the most serious of which would let an unauthenticated attacker achieve remote code execution — provided HTTP/2 is enabled on the target. In Xint's account, reaching root on the host requires chaining those flaws with a separate local-privilege-escalation vulnerability in the kernel's SCTP subsystem, introduced in Linux 4.15, for which no fix is currently available. LWN.net summarised the disclosure and links to Xint's own technical write-up as the primary source.
Every vulnerability claim in this article is attributed to Xint and should be treated as vendor self-disclosure pending independent confirmation. Xint's write-up appears to be the primary public source for these findings; at the time of publication we have not been able to confirm an upstream nginx.org advisory, a CVE assignment, or a coordinated-disclosure timeline, and the per-vulnerability mechanics have not been independently validated by our team.
What Xint is claiming
The five flaws, as summarised from Xint's write-up, form a progression rather than five independent bugs. The entry point is the unauthenticated code-execution flaw in request handling, which requires HTTP/2 to be in use. From there, Xint says, the remaining flaws carry an attacker through to privilege escalation on the host — but only when chained with the kernel SCTP bug, and only where the SCTP module is reachable in the running kernel. The end state of the full chain is root on the machine.
That dependency matters for risk assessment: the NGINX-side flaws alone do not deliver root, and the kernel leg has no fix available yet. Equally, the initial remote code execution needs no credentials, session, or prior foothold beyond network reachability to an HTTP/2-enabled instance.
On remediation, Xint explicitly recommends updating to NGINX 1.31.9 — just released — and blocking the SCTP module. Those two actions, taken together, address both the web server leg and the kernel leg of the reported chain.
Operator response
For teams running self-managed NGINX, the response breaks into three parallel workstreams.
Patch and upgrade. With a fixed version named, the first move is concrete: inventory every NGINX deployment, note how it was obtained — distribution packages, upstream source builds, in-house builds, or vendor appliance images — and upgrade to 1.31.9 or the corresponding distributor fix through the normal errata channel. Where builds are compiled in-house, engineering needs a rebuild plan now, not after an exploit appears.
Block the SCTP leg. Because the kernel component of the chain has no fix, operators should follow Xint's advice to block the SCTP module on hosts running NGINX: remove or disable the module where business requirements allow, and confirm it is not loaded. This is the only currently available measure against the unpatched half of the chain.
Containment and hardening. Until every instance is upgraded, WAF and reverse-proxy rules that normalise request paths, reject malformed URI encodings, and rate-limit anomalous traffic from unauthenticated sources can reduce the reachable attack surface — time-buying containment, not remediation, since a rule can be bypassed. Durable defence comes from reducing privilege: running worker processes under a dedicated non-root user, disabling unnecessary modules, limiting file-system and syscall exposure through sandboxing, and keeping configuration minimal all shrink what an attacker gains even if code execution succeeds.
The harder cases
A meaningful slice of NGINX deployments cannot be updated directly by the operator: embedded builds inside hardware appliances, long-term-support container images pinned for reproducibility, and vendor-managed stacks where patch cadence sits with a third party. For these, the immediate action is to identify affected assets, confirm with the vendor whether 1.31.9-equivalent fixes are forthcoming, and apply network-level compensating controls — including SCTP blocking where host access permits — in the interim.
For internet-facing instances, assuming exposure until upgrade status is positively confirmed costs little and avoids the trap of treating a deployment as unaffected simply because no alert has fired.
Caveats
Xint's write-up is the primary and, at present, the only detailed source we have been able to identify for these claims; the full technical analysis is available directly from Xint, and LWN.net's coverage points back to it. Per-vulnerability specifics remain unvalidated by our team. Organisations running self-managed NGINX should monitor nginx.org and their distribution security channels for an upstream advisory, and treat any CVE assignment as the point at which formal patch prioritisation and follow-up coverage begin.
網絡安全公司 Xint 表示,已在 NGINX web server 揭露五個嚴重漏洞,其中最嚴重者可讓未經身份驗證的攻擊者實現遠端代碼執行——前提是目標已啟用 HTTP/2。根據 Xint 的說法,要取得主機的 root 權限,需將這些漏洞與 kernel SCTP 子系統中另一個本地權限提升(local-privilege-escalation)漏洞串連,該漏洞於 Linux 4.15 引入,目前尚無修復可用。LWN.net 已總結是次披露,並連結至 Xint 自身的技術文章作為主要來源。
本文所有漏洞說法均歸於 Xint,在獲得獨立證實之前,應視為供應商自我披露。Xint 的技術文章似乎是這些發現的主要公開來源;在本文刊出時,本團隊未能確認上游 nginx.org 是否已發布 advisory、是否已有 CVE 分配或協調披露(coordinated disclosure)時間表,各漏洞的具體運作機制亦未經本團隊獨立驗證。
Xint 的說法
五個漏洞(根據 Xint 技術文章的總結)構成一個遞進過程,而非五個獨立漏洞。入口是請求處理中未經身份驗證的代碼執行漏洞,必須正在使用 HTTP/2。Xint 指出,其餘漏洞可帶領攻擊者一路提升至主機權限——但僅限與 kernel SCTP 漏洞串連,且僅限於運行中的 kernel 可觸及 SCTP module 的情況。整條鏈的最終狀態是取得機器的 root 權限。
此依賴關係對風險評估十分重要:單靠 NGINX 端的漏洞並不能取得 root,而 kernel 一環目前亦未有修復可用。同樣,最初的遠端代碼執行毋須憑證、session,或任何超出網絡可達性(即可連接到已啟用 HTTP/2 的實例)的先行立足點。
在修復方面,Xint 明確建議更新至 NGINX 1.31.9——剛剛發布——並封鎖 SCTP module。這兩項措施合併實施,可同時處理所報告鏈的 web server 一環與 kernel 一環。
維運人員的應對
對於自行管理 NGINX 的團隊,應對可分為三條並行工作線。
修補與升級。 既然已指明修復版本,第一步十分具體:盤點每一個 NGINX 部署,記錄其取得方式——發行版套件、上游 source build、自行編譯版本或供應商 appliance image——並透過正常的 errata 渠道升級至 1.31.9 或發行商對應的修復版本。若 build 由內部編譯,工程團隊現在就需制定重新編譯計劃,而非等到漏洞被利用後才行動。
封鎖 SCTP 一環。 由於鏈中的 kernel 部分沒有修復,維運人員應跟隨 Xint 的建議,在運行 NGINX 的主機上封鎖 SCTP module:在業務允許的情況下移除或停用該 module,並確認其未被載入。這是目前針對尚未修補的那一半鏈條唯一可用的措施。
遏制與加固。 在所有實例完成升級之前,能夠正規化請求路徑、拒絕格式錯誤的 URI 編碼、並對來自未經身份驗證來源的異常流量實施 rate limit 的 WAF 及 reverse-proxy 規則,可縮減可被觸及的攻擊面——這是爭取時間的遏制(containment),而非修復(remediation),因為規則可能被繞過。長久防禦則來自降低權限:以專用的非 root 用戶執行 worker process、停用不必要的 module、透過 sandboxing 機制限制檔案系統及 syscall 的暴露範圍,以及保持配置最簡化,即使代碼執行成功,也能縮小攻擊者所能獲得的成果。
較棘手的情況
相當一部分 NGINX 部署並非維運人員可直接更新:內置於硬件 appliance 的嵌入式 build、為確保可重複性而固定版本的長期支援容器鏡像,以及 patch 節奏由第三方主導的供應商管理堆疊。對此,即時行動是識別受影響的資產、向供應商確認 1.31.9 的等效修復是否即將推出,並在過渡期間實施網絡層面的補償性控制(compensating controls)——在主機存取權限許可的情況下包括封鎖 SCTP。
對於面向互聯網的實例,在升級狀態獲得確實證實之前預設已暴露,代價甚微,同時可避免因未有警報觸發便將部署視為不受影響的陷阱。
注意事項
Xint 的技術文章是上述說法的主要、且目前本團隊所能識別的唯一詳細來源;完整技術分析可直接向 Xint 獲取,LWN.net 的報導亦指向該處。各漏洞的具體內容仍未經本團隊驗證。自行管理 NGINX 的機構應監察 nginx.org 及其發行版安全渠道,以獲取上游 advisory,並將任何 CVE 分配視為正式開始排定修補優先次序及後續報導的時間點。
