Podcast Detail

Honeypot Analysis as Code; PHP Patches and Session RCE (CVE-2026-126714); Dragos AI Canary Tokens; vCenter Exploit (CVE-2026-123890) (Nov 20th 2026)

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/10188.mp3

Podcast Logo
Honeypot Analysis as Code; PHP Patches and Session RCE (CVE-2026-126714); Dragos AI Canary Tokens; vCenter Exploit (CVE-2026-123890) (Nov 20th 2026)
00:00

My Next Class

Click HERE to learn more about classes Johannes is teaching for SANS

This episode covers a SANS.edu intern's Git-and-pull-request workflow for AI-assisted honeypot analysis, today's PHP security updates including a session-handling remote code execution (RCE) bug, Dragos research that planted canary tokens in AI prompts, and an already exploited VMware vCenter vulnerability.

Treat Your Honeypot Analysis Like Code: Git, Pull Requests and a Human Gate for AI [Guest Diary]
SANS.edu undergraduate intern Kevin Truong explains how he treats honeypot analysis like code, using Git and pull requests so an AI assistant can propose query and report changes while a human analyst reviews and approves them. He says the same workflow can apply to other kinds of log analysis, not just honeypots.
https://isc.sans.edu/diary/33534

PHP Vulnerabilities and Patches
PHP released security updates in versions 8.5.13, 8.4.28, 8.3.37, and 8.2.36 that fix memory-corruption issues leading to remote code execution (RCE), including a session-handling flaw, CVE-2026-126714, that depthfirst documents as phpAfterShell. A separate moderate issue, CVE-2026-128064, can cause fopen() and similar stream functions to fetch an HTTPS or FTPS URL over plaintext when the scheme uses an uppercase S.
https://www.php.net/archive/2026.php
https://github.com/php/php-src/security
https://depthfirst.com/research/phpaftershell

Supply Chain Canaries
HD Moore, now with Dragos, describes planting canary tokens, also called honey tokens, in prompts sent to AI services in order to see whether those secrets later leak. Dragos sent approximately two million such requests and found that at least three AI providers exposed the tokens after apparent internal compromises, that inference-marketplace proxies inspected and used some of the tokens, and that some organizations did not honor their privacy policies.
https://www.dragos.com/blog/ai-supply-chain-security-canary-tokens

VMware vCenter Exploit
Mandiant and Google Threat Intelligence report that CVE-2026-123890, an unauthenticated authentication bypass in VMware vCenter Server, is already being exploited. The bug is in the vSphere stack's Directory Service and can yield an administrator session; Broadcom released patches in advisory VMSA-2026-0009 and lists no workaround.
https://cloud.google.com/blog/topics/threat-intelligence/armory-castle-exploits-vsphere
https://support.broadcom.com/web/ecx/support-content-notification/-/external/content/SecurityAdvisories/0/39495

No Podcast Thursday and Friday
Next week is the Thanksgiving holiday, so Stormcast will only be published Monday through Wednesday.

My Upcoming Classes
https://www.sans.org/profiles/dr-johannes-ullrich

Podcast Transcript

 Hello and welcome to the Friday, November 20th, 2026
 edition of the SANS Internet Storm Center's
 Stormcast. My name is Johannes Ullrich, recorded today from
 Jacksonville, Florida. And this episode is brought to you
 by the SANS.edu, Graduate Certificate Program in
 Cybersecurity Leadership. In diaries today, we have one of
 our undergraduate interns writing about how to, well,
 treat your analysis of Honeypot data like code. What
 I like about this diary by Kevin Truong is how he
 integrates human and AI analysis. Now, the trick that
 Kevin uses to actually keep himself organized here is that
 he uses Git in order to manage all of the queries and such.
 And essentially, whenever AI, for example, makes a change to
 a set of queries to analyze a particular incident, well,
 that's being treated as a pull request. And that way, of
 course, you have nice accountability as to who did
 what, what was contributed by the AI, what was contributed
 by the analyst. And of course, then the analyst can approve
 these pull requests and review them. So interesting approach
 here that I think can also be applied to other types of log
 analysis, not just to Honeypots. Well, as expected,
 we did get updates for PHP today. I mentioned the pre
 -announcement yesterday. Now we know what's in this
 particular set of patches. A couple of memory issues again.
 And with that, remote code execution vulnerabilities,
 there's also a remote code execution vulnerability, in the
 session handling code that I sort of think is maybe the
 most exploitable here, but still not easily exploited.
 There's a good blog post with details regarding this issue
 that I'll link to in the show notes. There's one non
 -critical vulnerability that I thought that was kind of odd
 and interesting. And that's if you are using the PHP file
 open command to connect to a remote URL, which is perfectly
 fine and often done in PHP. If you're connecting to an HTTPS
 website, but you're spelling the HTTPS in the URL in
 uppercase, it may actually connect via HTTP instead. Have
 to look at the code and see what exactly happened here.
 But I suspect that HTTPS uppercase is not sort of
 recognized as a valid protocol. And then it's
 defaults back to HTTP. Would be interesting to see if any
 other similar libraries and such that are not in PHP are
 having similar issues. Often you sort of see either code
 being copied or different developers thinking similarly.
 So again, an odd vulnerability. I don't think
 it deserves sort of to be rated critical or anything
 like this, but definitely something that I didn't
 expect. I've got an interesting blog post by HD
 Moore, who is now with Dragos. And what HD Moore did is
 essentially include canary tokens in requests to various
 AI tools. Now canary tokens, honey tokens is another name
 for them, are essentially deception technique where you
 are publishing tokens, API tokens that are basically
 intended to be stolen. And then you monitor if someone
 uses them. Now they did this with over a million, I
 believe, different requests. And well, what they found was
 somewhat enlightening overall. So first of all, they found
 that at least three of the AI providers they used exposed
 these tokens likely because of an internal compromise. For
 example, they had connections to their AWS account using
 these canary tokens. Also inference marketplace
 providers. These are essentially proxies that are
 forwarding your requests and basically attempt to optimize
 your token use and your billing. Well, being a proxy,
 of course, they have an insight into what you're
 sending to these providers. And that's exactly the problem
 that happened here that by inspecting those requests,
 well, they actually were also leaking or using some of that
 tokens. And then finally, there were basically organizations
 that did accept these requests while outright lying in their
 privacy policy. And yes, definitely inspecting and
 using data that was included in the request. I always like
 the idea of canary tokens and definitely recommend that you
 may include some of them with some of the AI providers that
 you're using to see if anything is leaking. And that
 has been a problem where data being submitted to providers
 is either used for training. That's almost sort of the
 least harmful part here. But then as indicated here, may
 also be used then to reach out to resources that you
 mentioned in the prompt without really telling you
 about that. Well, that's it for today. At least that's all
 I had time for today. There was one more vulnerability I
 almost included in vSphere. Apparently, it's already being
 exploited according to a blog by Google and Mandiant. I'll
 add the URL in the show notes in case you're running
 vSphere. You may want to take a look at that. Otherwise,
 hope you like this podcast. Thanks for liking it. Thanks
 for subscribing. Thanks for listening. And next week, of
 course, there will be the Thanksgiving holiday. So we'll
 only have a podcast Monday through Wednesday. Thanks and
 talk to you again on Monday. Bye.