Java Alternative

Two doors for the November 2026 Java emergency โ€” find out whether Java4Shell can reach you, then patch it right.

Educational only. Built from the same verified research as the Lodestar channel video “CISA says patch Java NOW โ€” is the internet actually exposed?”

The Week Everything Broke โ€” a short story

It started with the skies. On Friday, November 13, 2026, at about 3:30 p.m. Eastern, data began flooding the FAA's flight-data network โ€” roughly 1,000 connection attempts every four to five seconds. At 3:54 p.m. the FAA stopped every departing flight in the United States. The nationwide ground stop lasted until 8:15 p.m. โ€” four hours and twenty-one minutes. About 3,500 aircraft were already airborne; every one landed safely. The instability reached 17 of the FAA's 22 en-route control centers. Two sites were the suspected sources โ€” a radar site in Amarillo, Texas, and a radio site in El Dorado, Arkansas, where contractor L3Harris had been deploying a new router. The broadcasting function blamed for the flood was later disabled on more than 60 similar Juniper routers. The FAA's words: based on all available evidence, there is no indication this was a cyberattack.

The cancellation count is where the weekend's retellings went wrong, so this page keeps both counters separate. The FAA's own later count for Friday is 12,462 cancellations, including 9,082 at major airlines โ€” with the FAA's own caveat that not all were directly caused by the ground stop. FlightAware counts 9,159 for the same day. Two honest counters, two totals; they are not averaged here.

The day before, Microsoft 365 had fallen over. Microsoft's preliminary post-incident review โ€” reported, with the final review due around November 20 โ€” describes a latent defect from a July fix in its networking gateway servers, a Teams failover onto those gateways, and about 14% of gateway capacity out of service. At the worst point more than 80% of the gateway servers were unresponsive; active impact ran 23 hours 50 minutes, November 12โ€“13, with Outlook hit hardest and longest.

Then, on Tuesday, November 17, came the Java act. CISA โ€” the U.S. cybersecurity agency โ€” issued a standalone alert: patch a critical flaw in Java, CVE-2026-114938, immediately. The flaw, nicknamed Java4Shell, is a heap out-of-bounds write in the Base64 decoder of the x86-64 HotSpot Java Virtual Machine. A 65-character input โ€” sixty-four letter A's and one equals sign โ€” makes the engine allocate a 50-byte array and then write a full 64 bytes into it: 62 bytes past the end. It was found by researcher Quang Luong at the startup Calif, who told OpenAI's Codex to audit the engine โ€” about an hour to find it, about two more hours to a first exploit, under $200 in tokens. (Press reports name the model "GPT-6 Cyber"; Calif's own write-up names only Codex, so the model name is labelled reported, not confirmed.)

Java4Shell was the second act, not the first. A month earlier the same firm had found two critical flaws in PHP โ€” PHP2Shell. And patching those PHP flaws is what broke something at the NSA this same week: the Director of National Intelligence said the NSA's classified NSANet went down on November 17 because systems did not come back up after the PHP patching. The NSA's own statement calls it scheduled routine maintenance. Restoration has not been confirmed. A third AI-found flaw, in the Brotli compression library โ€” a library, not a language โ€” rounds out the pattern: the auditors and the attackers now share the same tools, and the rate is the story.

Two honest scores close the story, and they are not merged: Oracle scores Java4Shell 8.1 (CVSS v3.1, attack complexity High, "difficult to exploit"); Calif scores it 9.2 (CVSS v4.0, no vector published). It is not in CISA's Known Exploited Vulnerabilities catalog. As of November 17 there are zero confirmed production compromises and zero reported in-the-wild exploitation. The crash is independently confirmed; Calif's product-level code-execution claims are its own, unconfirmed; Wiz says it proved code execution against Jenkins, Confluence and Tomcat in its labs. The doors on this page are the response: check whether you are exposed, then patch.

Every event in this story is drawn from the verified research behind this site: the FAA's own Q&A, Microsoft's preliminary post-incident review as reported, CISA's November 17 alert, Oracle's November 2026 Critical Security Patch Update, the CVE record, and Calif's write-up โ€” each claim labelled verified or reported in the underlying research, and disagreements kept separate, never averaged.

Door 1 โ€” Am I affected? The three-part check

What happened: a critical memory-corruption flaw โ€” nicknamed Java4Shell (CVE-2026-114938) โ€” in the Base64 decoder of the x86-64 HotSpot JVM lets specially crafted Base64 input crash the JVM and, per CISA, allow unauthenticated remote code execution. CISA urges immediate patching. But exposure is narrower than the headlines: it needs all three parts below at once. Miss any one, and this flaw cannot reach you.

Part 1 โ€” An affected HotSpot build

The CVE record lists these exact builds as affected:

ProductAffected build
Oracle Java SE 8 (Enterprise Performance Pack only)8u511b33-perf
Oracle Java SE 1717.0.21.0.3
Oracle Java SE 2121.0.13.0.3
Oracle Java SE 2525.0.5.0.3
Oracle Java SE 2727.0.1.0.3
Oracle GraalVM for JDK 17 / 2117.0.21.0.3 / 21.0.13.0.3
Oracle GraalVM Enterprise Edition21.3.20.0.3

In range terms: OpenJDK 17.0.6 and later in the 17 line, and the 21 / 25 / 27 lines before their November 17, 2026 fixes. Builds from the same OpenJDK code โ€” Amazon Corretto, Eclipse Temurin, Microsoft Build of OpenJDK, Azul Zulu, BellSoft Liberica, SAP SapMachine, Red Hat OpenJDK โ€” carry the same code; check your vendor's advisory.

Part 2 โ€” An x86-64 CPU with AVX-512 VBMI

The vulnerable fast lane only runs on x86-64 processors with AVX-512 VBMI: Intel Xeon Ice Lake through Granite Rapids, and AMD Zen 4 or newer (EPYC and Ryzen). On Linux, check the flags:

grep -m1 flags /proc/cpuinfo

All four of these must be present: avx512f avx512bw avx512vbmi bmi2. Missing any one means the vulnerable routine is not in use on that machine. Note the cloud catch: the same container image can be safe on an ARM/Graviton host and exposed on an Ice Lake or Zen 4 host โ€” check per host, and re-check after instance-type changes.

Part 3 โ€” Attacker-controlled data reaching the decoder

The flaw triggers when data an attacker controls is decoded by the JVM's basic or URL-safe java.util.Base64 decoder โ€” the everyday path for login headers, tokens, SAML, JWTs and file uploads. Datadog's testing found the MIME decoder path bypasses the accelerated routine. A Java program that never decodes a stranger's Base64 is not reachable by this flaw.

Not affected โ€” stated plainly

  • JavaScript โ€” a different language and a different engine. Not affected. (The names confuse everyone; the code shares nothing.)
  • Standard Java 8 and Java 11 โ€” not listed for this CVE. (The one exception is the Java 8 Enterprise Performance Pack build in the table above.)
  • Java 17.0.0โ€“17.0.5 โ€” before the vulnerable optimization shipped.
  • ARM64 anything โ€” AWS Graviton, Ampere, Apple Silicon. The discoverer states ARM64 CPUs are not affected.
  • x86-64 without AVX-512 VBMI โ€” older Xeons and pre-Zen-4 AMD chips do not run the fast lane.
  • Android โ€” no source names Android or its ART runtime as affected, and the flaw's own preconditions (HotSpot, x86-64 AVX-512) exclude it. That is reasoning from the preconditions โ€” an inference, not a vendor statement.

Not sure? Answer three questions

Door 2 โ€” Patch: the fix, the killswitch, and the controls

The fix โ€” update to a fixed build, then restart every JVM

Oracle's fix ships in the Critical Security Patch Update โ€“ November 2026 (655 patches in total across Oracle's portfolio โ€” press reports saying 652 are wrong โ€” with 11 for Java SE). Fixed build numbers below are labelled REPORTED: they come from Qualys, Datadog and Censys summaries of Oracle's release notes, because Oracle's own advisory page does not print fixed build numbers.

If you runFixed build (REPORTED)
Oracle Java SE 1717.0.21.1
Oracle Java SE 2121.0.13.1
Oracle Java SE 2525.0.5.1
Oracle Java SE 2727.0.1.1
Oracle Java SE 8 Enterprise Performance Pack8u513-perf
Amazon Corretto 17 / 21 / 25 / 27 (REPORTED, from Corretto changelogs via Censys)17.0.21.1.2 / 21.0.13.1.2 / 25.0.5.1.2 / 27.0.1.1.4

After updating, restart every JVM. The vulnerable code lives inside running processes; an updated file that a running engine never reloads protects nothing. Patch internet-facing systems first.

The honest open gap

As of November 17, 2026, no confirmed fixed build exists yet for Eclipse Temurin, Microsoft Build of OpenJDK, Red Hat OpenJDK, Azul Zulu, BellSoft Liberica, SAP SapMachine, or the Debian/Ubuntu distribution packages. If that is your Java, the gap is real: check your vendor again, and use the killswitch below meanwhile. No fixed GraalVM build numbers were named in any source read either.

The killswitch โ€” one flag, if you cannot patch today

From the discoverer, and independently test-confirmed by Datadog, Qualys and Wiz. Apply one flag where the JVM starts, then restart every JVM:

Preferred:

-XX:-UseBASE64Intrinsics

This disables the hardware-accelerated Base64 intrinsic entirely; Java falls back to its standard decoder. The performance cost is narrower โ€” heavy Base64 users may slow down.

Alternative:

-XX:UseAVX=2

This disables AVX-512 across the whole JVM โ€” a broader performance cost. x86-64 only: an ARM64 JVM does not recognize this flag and will fail to start.

Step by step

  1. Find where your JVM gets its options โ€” the service unit, start script, container entrypoint, or JAVA_TOOL_OPTIONS / JDK_JAVA_OPTIONS environment variable, whichever your deployment actually uses.
  2. Add exactly one flag from above to that place. Do not edit the application code; this is a runtime switch.
  3. Restart every JVM โ€” every node, every container, every bundled runtime you found in your inventory. A flag in a file protects nothing until the process restarts.
  4. Verify on the running process, not the start script:
    jcmd <PID> VM.version
    jcmd <PID> VM.flags -all
    The output should show UseBASE64Intrinsics = false (preferred flag) or UseAVX = 2 (alternative flag).
  5. Write down what you did, and when. Versions found, flags set, processes restarted. A dated record is the difference between an incident handled and an incident guessed at.

CISA's compensating controls (while you patch)

  • Validate and restrict Base64 input at the application boundary.
  • Enforce input-size limits. One caution from Qualys: size limits alone will not stop this โ€” the trigger is only about 65 characters.
  • Limit unauthenticated access to Java services where you can.
  • Take internet-facing identity, CI and collaboration consoles (Keycloak, PingFederate, Jenkins, Jira, Confluence, Artifactory) off the open internet or behind a VPN until patched โ€” reported guidance from Qualys and Wiz.

The bundled-JDK warning

Java hides. CISA's alert is explicit: check JDKs and runtimes bundled with third-party applications, included in container images, and shipped inside appliances โ€” they "may not appear as separate entries in software inventories." Patching the operating system's Java does not patch the Java inside your identity console, build server or appliance. Those fixes come from their vendors, on their vendors' schedules.

Official sources

CISA alert (Nov 17, 2026): cisa.gov โ€” Critical Java Vulnerability Requires Immediate Patching
Oracle Critical Security Patch Update โ€“ November 2026: oracle.com/security-alerts/cspunov2026
CVE record: cve.org โ€” CVE-2026-114938
Discoverer's write-up: calif.io/java4shell
Qualys guide: blog.qualys.com ยท Datadog: securitylabs.datadoghq.com ยท Censys: censys.com/advisory

Detection โ€” what to watch for

State it plainly: no IOC list exists for Java4Shell. No published IP addresses, domains, hashes or file artefacts from CISA or any other source read. Detection today is behavioural.

  • CISA (verified): monitor for rogue runtimes, unexpected JVM crashes, and suspicious Java processes.
  • Qualys (reported): unexpected JVM crashes or restarts on internet-facing Java services; Java processes that spawn a shell or make new outbound connections; Java runtimes on a host that your inventory does not account for.
  • Crash artefacts (reported): JVM fatal-error logs (hs_err_pid*.log), SIGSEGV in the G1 garbage collector, and bursts of IllegalArgumentException from java.util.Base64 clustered by source โ€” the decoder throws after corrupting memory, so the exception is a tell, not an all-clear.
  • A caution: a lone trigger-shaped string in a log is not proof of compromise. It fits scanning, research, or a failed attempt just as well.
  • Network rules (reported): Cisco Talos published Snort rules matching malformed Base64 in specific HTTP request fields (Snort2 SIDs 1:67502โ€“1:67513; Snort3 1:301754 and 1:67502โ€“1:67513).

Support this work

bc1q96sz892qrnl77yyemkxq2ku3stl2qn4e5kheex

Voluntary support for an educational project. Not a charity, not tax-deductible, nothing promised in return. This is a dedicated public-facing address for this project โ€” one address only, never a primary wallet. Or open it in your wallet: bitcoin: link

About this site

What it is for: education. Java Alternative exists to explain, in plain language, what happened in the November 2026 Java security emergency, who is actually exposed, and how to patch it โ€” so that server operators, companies and developers can make their own informed decisions.

How it was made: this site was created with AI assistance, from a deep, source-checked research pass (CISA's alert, Oracle's advisory, the CVE record, the discoverer's write-up, and independent security research, each claim labeled by strength โ€” verified on the primary source, or reported by a named secondary). AI makes mistakes, like humans do. Fixed builds and vendor positions change by the day. Do your own research: verify anything you act on against the official source linked beside it.

What it is not: this is educational content only โ€” not security advice for your specific systems, not financial advice, not legal advice. Nothing here is a recommendation to buy, sell, or migrate anything.

Companion to the Lodestar channel video “CISA says patch Java NOW โ€” is the internet actually exposed?” Research snapshot: November 17โ€“18, 2026.

Support this work

bc1q96sz892qrnl77yyemkxq2ku3stl2qn4e5kheex

Voluntary support for an educational project. Not a charity, not tax-deductible, nothing promised in return. This is a dedicated public-facing address for this project โ€” one address only, never a primary wallet. Or open it in your wallet: bitcoin: link