On 18 November (Wednesday), the U.S. Cybersecurity and Infrastructure Security Agency (CISA) added four actively exploited vulnerabilities to its Known Exploited Vulnerabilities (KEV) catalog — and the spread across four layers of a routine IT estate is the part local administrators should pay attention to. The additions cover Synology DiskStation Manager (DSM) network storage, BerriAI's LiteLLM proxy, the legacy Apache Axis SOAP framework, and Vanderbilt University's Research Electronic Data Capture (REDCap) platform, with exploitation activity linked to ransomware, cryptomining, and LLMjacking campaigns.

The most immediately uncomfortable item for a growing number of teams is the one aimed at AI builders: a flaw in LiteLLM, the proxy layer that sits in front of LLM API keys and cloud credentials.

Why these four matter

The spread is the story. Within a single advisory, administrators are looking at four distinct layers of an ordinary IT estate:

  • NAS appliances — Synology DSM remains a common fixture in offices and homelabs, and is frequently exposed to the internet for remote file access.
  • AI infrastructure — LiteLLM sits in front of LLM API keys and cloud credentials. LLMjacking, in this context, refers to the theft of those keys or credentials to run attacker-side inference, often resold to third parties — not simply illicit mining of GPU cycles.
  • Legacy middleware — Apache Axis powers older SOAP integrations that are rarely inventoried and even more rarely patched.
  • Research platforms — REDCap is widely deployed in universities and clinical research groups, where availability and data integrity are both sensitive.
How the KEV list works

CISA's Binding Operational Directive 22-01 requires U.S. federal civilian agencies to remediate KEV-listed flaws by a stated due date. Outside that mandate, the list functions as a de facto global patch-prioritisation benchmark: inclusion signals both exploitation evidence and a concrete remediation deadline, which is why security teams worldwide track it even when the directive does not apply to them.

What administrators should do

CISA has not published the specific CVE identifiers, affected versions, or per-flaw due dates in the summary available to us. Administrators should not wait on second-hand reporting for that data: each flaw's CVE identifier, required remediation action, and due date is recorded directly in CISA's KEV catalog entry for the November addition, which should be treated as the source of record.

Priorities below reflect our own assessment of exposure and blast radius, not CISA's due dates.

Product Layer Who's exposed Priority
BerriAI LiteLLM AI gateway Teams proxying LLM APIs, anyone holding API keys High — credential theft enables LLMjacking
Synology DSM NAS Internet-facing file servers, remote-work setups High — direct ransomware/cryptomining linkage
Apache Axis Legacy SOAP Older integrations, forgotten middleware Medium — low visibility, high dwell time
REDCap Research data Universities, clinical research groups Medium — sensitive datasets, institutional impact

A practical triage sequence:

  1. Inventory the four products across on-premises and cloud deployments.
  2. Pull vendor advisories for fixed versions, then cross-check against the KEV entry for due dates.
  3. Review exposure — DSM and Axis instances reachable from the internet should be prioritised.
  4. Check for abuse, not just patch status: review LLM API key rotation, cloud credential use, GPU and egress spend anomalies, and NAS log integrity, since active exploitation may have preceded any patch.

The four names are already circulating across security feeds following the 18 November update. The harder question for local administrators is which of these is quietly running in a corner of your network right now — and whether it was there when CISA first saw it being exploited.


美國網絡安全和基礎設施安全局(CISA)於 11 月 18 日(星期三)將四項正被活躍利用的漏洞加入其已知被利用漏洞(KEV)目錄——而這四項漏洞橫跨日常 IT 環境的四個層面,正是本地管理員應予留意之處。新增漏洞涉及 Synology DiskStation Manager(DSM)網絡存儲、BerriAI 的 LiteLLM proxy、傳統 Apache Axis SOAP 框架,以及范德堡大學的 Research Electronic Data Capture(REDCap)平台,相關利用活動與勒索軟件、加密貨幣挖礦及 LLMjacking 行動有關。

對日益增多的團隊而言,最即時令人不安的是針對 AI 開發者的那一項:LiteLLM 中的漏洞——該 proxy 層位於 LLM API keys 及雲端 credentials 之前。

為何這四項漏洞值得關注

關鍵在於其分佈範圍。在同一份安全通告之內,管理員需要處理的是一個普通 IT 環境中四個截然不同的層面:

  • NAS 裝置——Synology DSM 仍是辦公室及家用實驗室的常見配置,且經常直接暴露於互聯網以供遠端檔案存取。
  • AI 基礎設施——LiteLLM 位於 LLM API keys 及雲端 credentials 之前。在此語境下,LLMjacking 指的是竊取這些 keys 或 credentials 以進行攻擊者一方的推理運算,並往往轉售予第三方——而非單純非法使用 GPU 運算資源。
  • 傳統中介軟件——Apache Axis 支撐較舊的 SOAP 整合,此類系統鮮有列入資產清單,遭到修補的機會更少。
  • 研究平台——REDCap 廣泛部署於大專院校及臨床研究機構,其可用性與數據完整性同樣敏感。
KEV 清單的運作方式

CISA 的 Binding Operational Directive 22-01 要求美國聯邦文職機構在指定期限前修補列入 KEV 的漏洞。在該指令管轄範圍之外,此清單實際上充當全球性的修補優先次序基準:被列入清單既代表已有利用證據,亦意味著有具體的修訂期限,這正是全球安全團隊即使不受該指令約束仍然跟蹤此清單的原因。

管理員應採取的措施

在我們取得的摘要中,CISA 尚未公布具體的 CVE 編號、受影響版本或各漏洞的修補期限。管理員不應等待二手報道來獲取這些資料:每項漏洞的 CVE 編號、所需修補行動及期限,均直接記錄於 CISA KEV 目錄中 11 月新增條目之內,應以該條目作為權威來源。

以下優先次序反映我們自身對暴露範圍及影響程度的評估,並非 CISA 公布的修補期限。

產品 層面 受影響對象 優先次序
BerriAI LiteLLM AI gateway 代理 LLM API 的團隊、任何持有 API keys 的人 高——credentials 被竊可導致 LLMjacking
Synology DSM NAS 面向互聯網的檔案服務器、遠端工作配置 高——與勒索軟件/加密貨幣挖礦直接關聯
Apache Axis 傳統 SOAP 較舊的整合系統、被遺忘的中介軟件 中——可見度低、潛伏時間長
REDCap 研究數據 大專院校、臨床研究機構 中——數據集敏感、機構影響面大

實用的處理次序:

  1. 盤點四項產品在本地部署及雲端部署中的使用情況。
  2. 查閱廠商通告以確認已修補版本,再對照 KEV 條目核實修補期限。
  3. 檢視暴露情況——可從互聯網到達的 DSM 及 Axis 實例應優先處理。
  4. 檢查是否已被濫用,而非僅查看修補狀態:檢視 LLM API key 輪換、雲端 credentials 使用情況、GPU 及出站流量開支異常,以及 NAS 日誌完整性,因為活躍利用可能早於任何修補行動發生。

自 11 月 18 日更新以來,這四個名字已在各安全情報渠道流傳。對本地管理員而言,更難的問題是:這些產品之中,有哪一個此刻正悄悄運行於你網絡的某個角落——以及當 CISA 首次發現其被利用時,它是否已在那裏。

新聞來源 / Original News Source