CVE-2026-114938 Java4Shell (CVSS 9.2) Base64 heap overflow Red/Blue toolkit. JDK/CPU exposure checks, bounded local proof, crash and mitigation tests, remote version scan, --profile heap hints, --persist/--exfil helpers, SOCKS5/Tor, stealth jitter, JSON reports. Experimental probe; experimental RCE chain. Only authorized security testing.
# CVE-2026-114938 – Java4Shell OpenJDK AVX-512 VBMI Base64 Heap Overflow Toolkit
[](https://www.python.org/downloads/)
[](https://nvd.nist.gov/vuln-metrics/cvss/v4-calculator)
[](https://github.com/tc4dy/CVE-2026-114938-Java4Shell)
[](https://openjdk.org/)
> **Ethical Use Only – Authorized Security Testing**
>
> This repository provides tools for **authorized security professionals, blue teams, and penetration testers** only.
> Unauthorized access to computer systems is illegal under **CFAA (US), Computer Misuse Act (UK), TCK 243/244 (Turkey), and similar laws worldwide**.

> photo from endor labs
---
## [*] Vulnerability Overview
**CVE-2026-114938** is a heap-based buffer overflow in OpenJDK HotSpot JVM's AVX-512 VBMI Base64 decoding intrinsic. A malformed Base64 input ending with a padding character (`"A"*64 + "="` on JDK 21/25/27, `"A"*128 + "="` on JDK 17) causes an integer underflow in the output size calculation. The negative size is interpreted by the `bzhiq` instruction as a large unsigned value, producing a fully-open write mask. The subsequent `evmovdqub` writes 64 bytes to a destination buffer that has no room for them, resulting in a **62-byte out-of-bounds write**.
The overflow happens before `IllegalArgumentException` is thrown, so the heap is already corrupted when the application sees the error. The same corruption primitives lead to heap disclosure ("JavaBleed") and ultimately to JIT code cache overwrite, which is where code execution lives.
The upstream fix is JDK-8393212, a three-line guard that routes tails shorter than 4 bytes to the safe scalar decoder before the `bzhiq` underflow can occur.
### > How it works
1. **JIT compilatio