A memory-corruption bug in OpenJDK's Base64 decoder, tracked as CVE-2026-114938 and carrying a CVSS score of 9.2, allows unauthenticated remote code execution through a trigger as short as 65 characters. Patches have been released; the flaw reaches applications including Jenkins, Jira and Keycloak that run on the affected runtime.

What's affected
  • Runtime: OpenJDK, the most widely deployed Java runtime, which underpins a large share of enterprise Java applications.
  • Applications: Jenkins, Jira and Keycloak are among the products confirmed as reachable by the flaw, according to Security Affairs, which covered the disclosure on 18 November.
  • Attack requirements: No credentials. The bug is reachable pre-authentication, meaning an attacker needs only network access to the affected service.
How the trigger works

The flaw sits in OpenJDK's Base64 decoding routine. Calif, the research team that disclosed the issue and coined the "Java4Shell" name, showed that a specially crafted Base64 string of just 65 characters can corrupt memory during decoding and ultimately hand an attacker control of code execution on the host.

Because the corruption occurs in a decoding routine that sits on many request-handling paths, an attacker does not need to defeat an application's own authentication layer first. The malicious input only has to reach code that decodes it.

Administrators should identify every host and container running an unpatched OpenJDK build, apply the vendor updates, and rebuild any container images that bake in a vulnerable runtime, then restart the affected services. Where immediate patching is not feasible, compensating controls such as WAF rules or request filtering that reject anomalous Base64 payloads can reduce exposure in the meantime — a stopgap, not a substitute for patching.

Why this is not Log4Shell redux

The "Java4Shell" nickname invites comparison with Log4Shell, the 2021 flaw in the Log4j logging library. The parallel is mostly about blast radius: Log4Shell was severe because Log4j was embedded almost everywhere, just as OpenJDK is the default Java runtime across enterprise infrastructure.

The mechanics, however, are different. Log4Shell exploited JNDI lookups — an application feature that fetched remote resources — whereas CVE-2026-114938 corrupts memory at the decoder level. Defences that worked in 2021, such as blocking outbound JNDI or LDAP connections, do not address this bug. Patching the runtime is the mitigation.

This story matters to IT teams precisely because the vulnerable component sits below the application layer. Upgrading Jenkins or Jira alone may not be sufficient if the underlying OpenJDK installation remains unpatched; conversely, patching the runtime can close the hole for every application running on top of it in one pass.

Security Affairs reported the disclosure on 18 November, attributing the research and the vulnerability's name to Calif. The report makes no claim of exploitation in the wild, and none should be assumed — but with a pre-authentication path to remote code execution and patches already available, there is little reason to delay. As with any freshly disclosed high-severity issue, defenders should confirm the CVE identifier, CVSS score and affected product versions against the primary advisory or vendor bulletins before finalising remediation scope.

Source: Security Affairs — "65 Characters to RCE: Inside the Java4Shell Vulnerability"


OpenJDK 的 Base64 解碼器存在一個記憶體損毀缺陷,編號為 CVE-2026-114938,CVSS 評分為 9.2,可讓未經驗證的遠端透過僅 65 個字元的觸發載入執行任意程式碼(RCE)。相關修補程式已經發布;受影響的執行環境所驅動的應用程式包括 Jenkins、Jira 及 Keycloak。

受影響範圍
  • 執行環境: OpenJDK,即部署最廣泛的 Java runtime,支撐著絕大部分企業 Java 應用程式。
  • 應用程式: 根於 11 月 18 日報道是次披露的 Security Affairs,Jenkins、Jira 及 Keycloak 均屬已證實可被此缺陷觸及的產品。
  • 攻擊條件: 無需任何憑據。該缺陷可在驗證前階段(pre-authentication)觸及,即攻擊者只需網絡連接受影響的服務即可。
觸發機制如何運作

缺陷位於 OpenJDK 的 Base64 解碼例程之中。披露是次問題並命名為「Java4Shell」的研究團隊 Calif 展示,只需一條精心構造、僅 65 個字元的 Base64 字串,便可在解碼過程中損毀記憶體,最終令攻擊者取得主機上的代碼執行控制權。

由於損毀發生在眾多請求處理路徑共用的解碼例程中,攻擊者無需先繞過應用程式自身的驗證層。惡意輸入只需到達負責解碼的代碼即可。

管理員應識別所有執行未修補 OpenJDK 版本的主機及容器,套用供應商更新,並重新構建任何內置有漏洞 runtime 的容器映像,隨後重啟受影響的服務。若無法即時修補,可暫時採用 WAF 規則或請求過濾等補償性控制措施,拒絕異常的 Base64 負載以降低暴露風險——這是權宜之計,並不能取代修補。

為何這不是 Log4Shell 翻版

「Java4Shell」這個別稱令人聯想到 2021 年 Log4j 日誌庫的漏洞 Log4Shell。兩者相似之處主要在於影響範圍:Log4Shell 之所以嚴重,是因為 Log4j 幾乎無處不在,正如 OpenJDK 是企業基礎設施中的預設 Java runtime。

然而,兩者的運作機制並不相同。Log4Shell 利用的是 JNDI 查詢——一項用於取回遠端資源的應用程式功能;而 CVE-2026-114938 則是在解碼器層面損毀記憶體。2021 年有效的防禦手段,例如封鎖出站 JNDI 或 LDAP 連接,並不能解決這個缺陷。修補 runtime 才是根本的緩解方法。

是次事件之所以值得 IT 團隊關注,正因為有漏洞的元件位於應用程式層之下。若底層的 OpenJDK 安裝仍未修補,單靠升級 Jenkins 或 Jira 可能並不足夠;反過來說,修補 runtime 可以一次性堵塞其上所有應用程式的漏洞。

Security Affairs 已於 11 月 18 日報道是次披露,並將研究工作及漏洞名稱歸功於 Calif。報告未有聲稱漏洞已在野外被利用,外界亦不應作此假設——但既然已存在驗證前即可達致遠端代碼執行的途徑,且修補程式已經備妥,實在沒有理由延遲處理。與任何新披露的高嚴重性問題一樣,防禦方在確定修補範圍前,應先對照主要安全公告或供應商通報,核實 CVE 編號、CVSS 評分及受影響的產品版本。

來源: Security Affairs — "65 Characters to RCE: Inside the Java4Shell Vulnerability"

新聞來源 / Original News Source