Java4Shell: Critical OpenJDK Base64 flaw (CVSS 9.2) and how Cloudflare WAF blocks it
Cloudflare Blog · rss · 2026-11-18
Calif disclosed Java4Shell (CVE-2026-114938), a critical (CVSS 4.0: 9.2) memory-corruption flaw in OpenJDK's Base64 decoder: malformed padding in the hand-written AVX-512 assembly computes a negative output length, writing 62 bytes past the buffer and enabling Heartbleed-like heap leaks and remote code execution.
Who's affected: OpenJDK 17.0.6+ on x86-64 with AVX-512 VBMI (Intel Sapphire Rapids, AMD Zen 4+); ARM64 is unaffected. Pre-auth RCE was demonstrated against Keycloak, Jira, Confluence, Jenkins and others in default configs. Application-level validation doesn't help since corruption happens inside the decoder.
Mitigations:
- JVM flags -XX:-UseBASE64Intrinsics or -XX:UseAVX=2 remove the vulnerable path (restart required, some perf cost)
- Cloudflare shipped managed rules detecting malformed Base64 (Pro+ plans, Log mode by default), but the only complete fix is updating the Java runtime
- Bundled JDKs in containers and running JVMs are not updated by system package upgrades; every service needs a restart
Notably, Calif built its first working local exploit in 3 hours with the help of AI agents, so other exploit chains are expected quickly.
More from Safety
- LeCun signs open letter from 59 European leaders urging EU not to sleep through AI revolution — ylecun · 2026-11-18
- 20% of UK's Milburn Review report was AI-generated, sparking disclosure debate — birchlse · 2026-11-18
- Rep. Liccardo proposes liability safe harbor for enterprises using CAISSI-certified safest AI labs — ShakeelHashim · 2026-11-18
- OpenAI study: persona-style CoT training drops monitorability from 0.94 to 0.22 — dair_ai · 2026-11-18
- UK AI minister says Britain doesn't need its own frontier AI model, betting on trusted access — rosstaylor90 · 2026-11-18
- Trump administration weighs requiring US firms to disclose use of Chinese AI models — Polymarket · 2026-11-18