Originally published on The Daily Flare.
Oracle issued a critical Java security patch on November 17 after Calif Research disclosed an AI-assisted discovery of a remote-code-execution flaw in OpenJDK.
The vulnerability, tracked as CVE-2026-114938 and nicknamed Java4Shell, affects the Base64 decoder in the x86-64 HotSpot Java Virtual Machine, part of the Java Development Kit. The US Cybersecurity and Infrastructure Security Agency (CISA) classified it as a critical memory-corruption flaw and said specially crafted input could crash the JVM or allow unauthenticated remote code execution.
The defect sits in hand-written processor instructions used to speed up Base64 decoding, according to CyberScoop. Calif said the problem occurs when the decoder receives 64 “A” characters followed by an equals sign, the padding character used in Base64. The code miscalculates the input length and writes 64 bytes into a memory buffer that can be as small as 50 bytes, overwriting neighboring data before the program rejects the input.
That damage can crash a program, but researchers said an attacker may be able to turn it into control of the server. Calif described using heap grooming, a technique for arranging data in memory, to make the overwrite more useful to an attacker. The optimization involved runs on processors supporting AVX-512; CyberScoop reported that ARM64 systems are unaffected.
Calif reported unauthenticated code execution in products including Keycloak, Jira, Confluence, Jenkins and PingFederate, and demonstrated a working break-in against Keycloak, an open-source login system. The researchers also showed code execution requiring a valid account in Kafka, Solr and Elasticsearch. Researcher samczsun demonstrated that Java programs could be compromised by connecting to a malicious server.
Calif researcher Quang Luong found the bug in about an hour while auditing OpenJDK with OpenAI’s Codex, the security firm said. The computing tokens cost less than $200, according to CyberScoop’s account of the research. Calif also said AI agents produced most of its exploit chains with little human involvement. The researchers cautioned that there is no universal exploit and that each target requires tailored steps.
Oracle shipped patched Java builds on November 17 as part of its November critical patch update, which covered 652 vulnerabilities across the company’s products, CyberScoop reported. CISA urged organizations to identify affected Java installations and apply relevant Java or vendor fixes promptly, prioritizing systems exposed to the internet.
In its CISA alert, the agency warned that Java runtimes may be bundled inside third-party applications, container images and appliances, and may not appear as separate items in software inventories. For systems that cannot be patched immediately, it recommended restricting and validating Base64 input at application boundaries, setting input-size limits and limiting unauthenticated access. It also urged defenders to watch for rogue runtimes, unexpected JVM crashes and suspicious Java processes.
The flaw matters because Java is used across banking, aviation, telecommunications, hospitals and government systems. Calif said OpenJDK runs at 90% of Fortune 500 companies. The BBC cited computer scientist Toby Walsh, who said Java’s role in managing memory made the weakness particularly concerning. Walsh said the flaw had been “hiding in plain sight for 12 years.”
Governments warned companies to install the fix, Reuters reported on November 18. South Korean authorities held an emergency meeting about possible Java vulnerabilities, according to Yonhap, as cited by Reuters.
Top comments (0)