Handler on Duty: Didier Stevens
Threat Level: green
Podcast Detail
SANS Stormcast Tuesday, October 20th, 2026: Plex2Shell RCE Exploited; Broadcom eCos Wi-Fi RCE; OVHcloud CLI Compromised; Exchange Issues After October Updates; AI Agent Containment; VBA in DWG YARA
If you are not able to play the podcast using the player below: Use this direct link to the audio file: https://traffic.libsyn.com/securitypodcast/10142.mp3
Plex2Shell RCE Exploited; Broadcom eCos Wi-Fi RCE; OVHcloud CLI Compromised; Exchange Issues After October Updates; AI Agent Containment; VBA in DWG YARA
00:00
My Next Class
Click HERE to learn more about classes Johannes is teaching for SANS
Today's episode covers an actively exploited unauthenticated remote code execution chain in Plex Media Server, unpatched Broadcom eCos Wi-Fi flaws that allow code execution on access points without connecting, an infostealer slipped into the OVHcloud command line tools, Exchange Server instability after the October 2026 Windows updates, how to contain AI agents, and a YARA rule for orphaned VBA projects in AutoCAD DWG files.
Small Agents, Smaller Blast Radius: Controlling What AI Can Do
Mark outlines techniques to restrict what AI agents can do, including limiting permissions, using containers, and planning for AI vendors retiring older models.
https://isc.sans.edu/diary/Small%20Agents%2C%20Smaller%20Blast%20Radius%3A%20Controlling%20What%20AI%20Can%20Do/33430
Orphaned VBA Projects Inside DWG Files: YARA Rule
Didier Stevens follows up on hidden Visual Basic for Applications projects in AutoCAD DWG files with a YARA rule to detect them.
https://isc.sans.edu/diary/Orphaned%20VBA%20Projects%20Inside%20DWG%20Files%3A%20YARA%20Rule/33434
Broadcom eCos Wi-Fi Remote Code Execution Vulnerability
Quentin Kaiser shows that an attacker in Wi-Fi range can execute code on Broadcom eCos access points without WPA authentication and steal the WPA key. Broadcom has not issued a CVE, and many ISP-provided access points are affected.
https://quentinkaiser.be/security/2026/10/16/one-parser-two-bugs/
OVHcloud Command Line Tools Compromised
For a few hours on Friday, compromised versions of the OVHcloud CLI shipped an infostealer. Update to version 0.16.0, released after the compromise, and pin package hashes.
https://github.com/ovh/ovhcloud-cli/issues/300
Plex2Shell Vulnerability (Exploited!)
A chain starting with server-side request forgery (SSRF) to loopback lets attackers steal admin tokens and run code on Plex Media Server without authentication (CVE-2026-96652, CVE-2026-96654, CVE-2026-96655, CVE-2026-96656). It is already exploited; patch now and assume compromise if your server is exposed.
https://zmain.info/blog/plex2shell
Exchange Servers May Become Unstable After the October 2026 Windows Updates
Microsoft recommends a registry change rather than uninstalling the update on on-premises Exchange Server.
https://techcommunity.microsoft.com/blog/exchange/exchange-servers-may-become-unstable-after-installation-of-october-2026-windows-/4565382
NGINX ngx_http_core_module Vulnerability (CVE-2026-15176)
https://my.f5.com/manage/s/article/K000162416
My Upcoming Classes
https://www.sans.org/profiles/dr-johannes-ullrich
| Application Security: Securing Web Applications, APIs, and Microservices | Washington | Dec 14th - Dec 18th 2026 |
| Application Security: Securing Web Applications, APIs, and Microservices | Online | US Eastern | Feb 8th - Feb 12th 2027 |
| Application Security: Securing Web Applications, APIs, and Microservices | Online | India Standard Time | Mar 15th - Mar 19th 2027 |
| Application Security: Securing Web Applications, APIs, and Microservices | Orlando | Apr 12th - Apr 16th 2027 |
| Application Security: Securing Web Applications, APIs, and Microservices | Online | US Mountain | Apr 21st - Apr 25th 2027 |
| Application Security: Securing Web Applications, APIs, and Microservices | Baltimore | May 17th - May 21st 2027 |
Podcast Transcript
Hello and welcome to the Tuesday, October 20th, 2026 edition of the SANS Internet Storm Center's Stormcast. My name is Johannes Ullrich, recording today from Jacksonville, Florida. And this episode is brought to you by the SANS.edu Graduate Certificate Program in Cyber Defense Operations. Well, with no podcast yesterday, we do have a little bit catching up to do. So let's start with diaries. Mark on Friday wrote about containing your AI agents. Certainly one issue that a lot of organizations are struggling with is how to properly restrict what your AI agent is able to do. And Mark is outlining some techniques you can do to, for example, restrict what permissions an AI agent had, how to set up containers and restrictions around that, but also, well, how to deal, for example, with AI companies changing models and no longer supporting older models that you may have originally used and that you may feel comfortable with with your AI agents. I think this also fits a little bit into something that I've seen Heather Barnhart say about AI agents, in particular when it comes to forensics, well, treat them as an intern. And here, would you actually let an intern on their first day have root access to all of your systems? Well, probably not. Well, remember last week Didier wrote about how to identify and extract data from orphaned VBA projects that may be residing inside of your AutoCAD DWG files. Well, Didier wrote a little follow-up to that with a YARA rule to help you identify these files. So as a heavy AutoCAD user, certainly worthwhile. Take a look at that and see if there is anything malicious or so hidden in these supposedly deleted, but really just unlinked Visual Basic projects. And last week, Quentin Kaiser did present at Hexacon in Paris, not so, wrote up a detailed blog post about a couple of serious vulnerabilities in Broadcom's eCos platform. This is the embedded customizable operating system, basically the operating system that Broadcom's Wi-Fi access points are running under. And the problem here is that an attacker in reach, in Wi-Fi reach of the access point, is able to execute arbitrary code on the access point without having to connect to the Wi-Fi network. So without having to actually complete any kind of WPA authentication. This is not one of your sort of web admin vulnerabilities, but this is a fundamental Wi-Fi 802.11 implementation vulnerability in this particular Broadcom platform. Now, what makes this even kind of worse is that Broadcom apparently isn't sort of acknowledging these vulnerabilities. Quentin tried to report it to Broadcom. There is no CVE being issued for it. Apparently, there was at one point sort of a silent patch for at least some of the vulnerabilities. Now, Quentin also did present a proof of concept that was able to, again, execute code on the access point. And then, for example, retrieve the encryption key for WPA. So then an attacker could just join the network and perform additional attacks. Or, well, with remote code execution on an access point, of course, you may also be able to do things like intercept traffic or just exfiltrate data that is being sent across the network. A large number of access points are vulnerable. And, of course, it's not always that obvious what exact chipset and such they're running. But apparently also many ISPs that are providing access points are delivering vulnerable access points to their clients. And if you're using the OVH cloud, you're likely going to also use the command line utilities that are published by OVH. These command line utilities were recently compromised. And that was about Friday for a few hours. An infostealer was added to these command line utilities. A couple different versions were compromised here. So if you are using the tools, again, make sure that you update that you're on the latest version. Version 0.16.0 is the version that was released post-compromise. So stick to that and, again, also pin on hashes. That apparently is also how this was discovered by customers of OVH realizing that the hash changed of existing versions for these command line utilities. And also on Friday, we do have a blog post about a new vulnerability in Plex, the media server. Now, the reason it's a problem is that many people are exposing Plex to the Internet to be able to remotely access their media library. Plex is pretty good about authentication for these remote requests. But the chain of vulnerabilities here essentially allows you to execute arbitrary code on a Plex server without authentication. It starts all out with a good old server-side request forgery where you're able to send a remote request to Plex and then essentially have it proxied to loopback. Well, Plex is a lot more relaxed about authentication for requests coming from loopback. And that is then being used to retrieve admin tokens and, in the end, execute arbitrary code. So this is something you definitely want to patch now. It's already being exploited. If your server is exposed to the Internet, assume compromise. And the Microsoft Exchange team blog is reporting that after applying the October 2026 Windows update to on-premise Exchange, it may become unstable. They recommend a fix that they're showing in their blog post and do not recommend that you uninstall the patch. Instead, just update the registry setting that they are recommending. Well, and that's it for today. So thanks for listening. Thanks for liking. Thanks for recommending this podcast. And talk to you again tomorrow. Bye. Bye. Bye.





