Risky Business · Software Engineering & Infra · November 2026
Patrick asked about the Java bug dubbed Java4Shell, an out-of-bounds write in OpenJDK's accelerated base64 decoder. James says the headline sounds worse than the real-world reach because of how exotic the required configuration is.
Hopefully. It'll be in a report, whether it gets actioned, completely different matter.
…but it looks like a pretty bad, yeah, pretty bad Java bug, which, you know, you wonder how long it is before this gets turned into a reliable exploit targeting like Palo Alto Panorama or something, because that's where all of these bugs seem to wind up. But what do we know about it so far, James?
Look, yeah, you know, as you said, Calif, they do great work, but there's also a lot of publicity attached to the work when they publish it. And you really need to look through and determine: is this as bad as it sounds, or are there mitigating circumstances here that limit the impact? In this one, there's quite a few, what I would call, load-bearing ands. You know, there's you've got to be using OpenJDK and on x86, and that x86 processor has the AVX512 extension set, and it has VBMI enabled, and that's available to the guest OS, and the JVM has turned that on, and the path has to be attacker. You know, the base64 decoder encode has to be attacker reachable and also exploitable. I think the sheer number of ends in that is going to severely limit the impact here.
Yeah.
But that said, even though this is like an exotic config with a lot of load-bearing ands, the original writer from Calif does say that in the right circumstances, they managed to achieve RCA on a bunch of enterprise stuff, not least of which was Key Cloak, Atlassian Jira, Confluence, JFrogata…
I see what you mean about Calif, though, because they are a little bit like, you know, shooting glitter cannons everywhere and like, you know, bring out the dancers. Yeah, the