What just happened? Part 1

The last two weeks have been rough with multiple RCE vulnerabilities, incredibly weak private keys on certs and a massive crypto theft, it's been hard to keep up.

Joe Garifo

11/3/20265 min read

How are you doing? You alright? Maybe a little shellshocked? These last two weeks have been a heck of a ride in InfoSec and IT realms. I remember BEAST, Heartbleed, Drupalgeddon, POODLE, Sweet32, EternalBlue, Ripple20, Log4j and many more over the years, but what if these all happened in a two week span? That's pretty much what we just went through. I'm gonna hit on a lot of the incidents here, not in super deep detail, but I will try linking to a number of articles from across the web that have more details.

Two weeks ago, here in The Netherlands, we were a week into the Ahold Delhaize recovery from a cyber security incident. Websites and apps were coming back online, supply chain was returning to normal at Albert Heijn, grocery deliveries were still cancelled. The US was was slightly impacted back this breach because they own Giant and Stop & Shop on the east coast. Aside from the note of suspicious activity that was the result of malicious actions of others, we don't really have any details on what happened, what was seen and who might be suspected of it. Ahold Delhaize released a statement here about it and I'll keep an eye out for more technical details of it.

October 19th was also the day things really started taking off for Plex2Shell, an issue that didn't hit the corporate world, but a lot of "prosumer" folks out there running Plex servers to share media with friends and family. A combination of low to medium vulnerabilities could be chained together to get unauthenticated remote code execution on Plex servers. The fix was released in August, this wasn't a zero day catching people off guard. But a Plex server is usually one of those leave it alone unless something breaks things for people, probably running on a device tucked away to keep down on the noise. These Plex servers are an enticing target because they can have decent GPUs in the them which can be used for cryptomining and because they have traffic flowing through them already, residential proxies with access being sold to redirect traffic. Peak Cyber has a fantastic write-up here about it.

October 20th is when reports start flying about certificate authority eMudhra having eight intermediate signing certs getting compromised. The private keys, the part critical to signing certificates that can't be made by someone else, are supposed to be a 384-bit number. That's big number and practically impossible for someone to search through that whole range of numbers to find the key. eMudhra's private keys ranged from 4-237. Not bits, the actual numbers, which maxes out at an 8-bit number. They commented on Mozilla's forums about the issue, that secondary weaker keys were generated when moving between HSMs and those weaker keys were published. There has also been some speculation that someone misunderstood how the keys worked and instead of using a 256-bit key, they picked numbers 1-256. Their comment refutes this theory a bit, but still leaves some questions. This one is a big one with long term concerns because these keys were trusted by pretty much every device and software released since mid-2025. Yes, OS and browser updates have been released with updated certificate revocation lists that don't trust these certs anymore, you have Apple not updating any of their command line tools to not trust them, IoT devices may or may not ever get these updates and are thus susceptible to man in the middle attacks from these compromised certs. Scott Helme has a great write up here going much more in depth on this issue.

As I am working late on the 20th, some of my security channels start buzzing about PHP RCE vulns leaking before scheduled release on the 22nd. I wake up on the 21st to see EURO-Cert recommending pulling any PHP based sites offline until they can be patched. People are waiting for this patch to come out and it turns out that Chainguard, one of the coordinated disclosure partners, released their patched binaries early to their customers. While there is speculation and some evidence these vulns were being exploited in the wild before the disclosure anyway, this release from Chainguard definitely accelerated these vulnerabilities being exploited on vast scale. At the same time, OpenAI released a statement talking about the coordinated disclosure and what a success it was, so unfortunately a lot of accusations fell on them for being the source of the leak. I'll admit, I did suspect them at first two, but I corrected myself one the PHP Foundation put out a statement clearly naming Chainguard as the culprit.

And during all of that Node.js releases a patch to address the eMudhra certs and for CVE-2026-111497, a vulnerability that leads to remote code execution. It starts getting called Node2Shell, notice a trend? And so now you have people waiting for patches from vendors for various applications and servers running Node.js that people can't update themselves. Then people start to realize it's not really Node.js that's the issue, the CVE is for the Brotli package, a compression library used to speed up HTTP traffic. It starts getting called HTTP2Shell because this affects many, many tools using this HTTP compression and decompression library. Tanium has a nice timeline and write up on their blog but it is now slightly out of date as this is firmly on multiple KEV lists now.

We also saw GitLab release a patch for a Remote Code Execution vulnerability in their regex engine. Which was the second one patched by them in October. Here on November 3rd, ShadowServer is still reporting thousands of instances still vulnerable to it two weeks later.

That was merely up to Wednesday two weeks ago. We had multiple RCE exploits in major platforms and leaked certificate signing keys. I am gonna break here and work on the write up for the next week of events, but I do have some parting thoughts. I have mostly negative feelings on LLMs and how they are being forced upon people as "inevitable", but the use of finding vulnerabilities in code is mostly a good thing. But a lot of these vulnerabilities are found in projects maintained by a handful of people and they support most of the internet. And the companies that make billions due to this code contribute very little to support the code. And I know OpenAI finding the vuln is helpful, but how much did they help to fix the code? We're also continuing to see that if there is gain to be had, people will go for it, like with Plex2Shell. Turning media servers into cryptominers or reselling bandwidth for proxy services. I know Anthropic has said they think the next 18 months is going to be hard but then we'll reach vulnerability 0. I believe the next 18 months is going to be hard, I don't doubt that for a moment, but I don't think we'll ever reach vulnerability 0. There is too much legacy software out there that won't get patched and the LLMs have shown they're great at writing code, not so much at shipping software. Back when I was at Tenable and LLMs were taking off, I wondered how long it would be until an LLM taking a pass over code was part of the CI/CD pipeline and we got here about four years later.

In part two, we'll cover the Tailscale SSH vulnerability, a UEFI boot vulnerability and the crypto heists at Binance and Polymarket. It's been wild, and hopefully things slow down, but yesterday did see more grocery chains in Europe have their freezers set to defrost remotely.