Podcast Detail

SANS Stormcast Tuesday, November 17th, 2026: LiteLLM/Starlette RCE Scans (CVE-2026-42271, CVE-2026-48710); Danfoss AK-SM800A RCEs; cyrus-sasl SRP Bypass (CVE-2026-129483); NoisyTCP (CVE-2026-99288)

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/10182.mp3

Podcast Logo
LiteLLM/Starlette RCE Scans (CVE-2026-42271, CVE-2026-48710); Danfoss AK-SM800A RCEs; cyrus-sasl SRP Bypass (CVE-2026-129483); NoisyTCP (CVE-2026-99288)
00:00

My Next Class

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

Internet Storm Center honeypots are seeing LiteLLM and Starlette exploit scans, Midnight Blue found unpatched remote code execution flaws in Danfoss AK-SM800A refrigeration controllers, a Cyrus SASL Secure Remote Password bypass can allow logins without a password, and NoisyTCP is a Linux kernel TCP denial-of-service bug.

Scans for LiteLLM/Starlette (CVE-2026-42271, CVE-2026-48710)
Internet Storm Center honeypots are seeing scans that chain a LiteLLM remote code execution bug in a test URL, CVE-2026-42271, with CVE-2026-48710, a Starlette Host header validation flaw that bypasses access control. Patches are available, and Horizon3.ai has a write-up of the chain.
https://isc.sans.edu/diary/Scans%20for%20LiteLLM%20Starlette%20%28CVE-2026-42271%2C%20CVE-2026-48710%29/33522
https://horizon3.ai/attack-research/vulnerabilities/cve-2026-42271-chained-with-cve-2026-48710/

Danfoss AK-SM800A Remote Code Execution Vulnerabilities
Midnight Blue found six pre-authentication remote code execution flaws, including a Host header buffer overflow, in Danfoss AK-SM800A refrigeration controllers after about an hour of manual firmware review. The devices lack standard protections such as stack canaries and address space layout randomization (ASLR), no patches are available yet, and the controllers should be kept isolated from the internet.
https://www.midnightblue.nl/blog/multiple-rce-danfoss-aksm800a

cyrus-sasl SRP Server Authentication Bypass (CVE-2026-129483)
A Secure Remote Password (SRP) bug in the Cyrus SASL library, often used to authenticate Internet Message Access Protocol (IMAP) and other mail services, lets an attacker log in as an existing user with no password. SRP is usually not enabled, and mail servers that offer it typically advertise the algorithm in their banner, so a quick scan can show whether it is in use.
https://github.com/cyrusimap/cyrus-sasl/issues/896

NoisyTCP: Linux Kernel TCP Pre-auth Remote Corruption (CVE-2026-99288)
NoisyTCP is a Linux kernel Transmission Control Protocol (TCP) denial-of-service vulnerability that can be triggered against any listening TCP socket, with no indication that code execution is possible. A patch is in the kernel.org tree, exploitability varies from about 14 hours on a quiet bare-metal server to much faster on a virtual machine, and disabling SYN cookies is a possible emergency workaround but is not recommended as a proactive fix.
https://seclists.org/oss-sec/2026/q4/715

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

Podcast Transcript

 Hello and welcome to the Tuesday, November 17th, 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 Bachelor's Degree Program in Applied
 Cybersecurity. Well, in Diaries today we have scans
 for LiteLLM. Actually, the attack that these scans
 attempt to pull off relies on two vulnerabilities. One is in
 LiteLLM and then the second vulnerability is in Starlette.
 That's a framework that's being used for web
 applications, a Python framework that LiteLLM does
 rely on. So what happens here, and I think the real
 vulnerability here is the Starlette one, the LiteLLM. Yes,
 it's a remote code execution. It's in this test URL, but it
 does assume authentication, access control. And of course,
 having remote code execution behind a proper access control
 may be something that you really want to offer in
 particular here, as the URL implies, sort of some kind of
 test feature that you have in your system. But the problem
 is that the access control part is broken in Starlette. And
 the problem is, yet again, well, trusting headers. So the
 way the access control works is that you can restrict
 certain URLs. Of course, that's normal. But the URL is
 composed of the host name that's pulled from the host
 header, and then the URL that you have in the first line of
 your HTTP request. The problem is that Starlette never really
 verifies the host name, that it actually is a valid host
 name. So what happens is that the attacker is able to add
 characters like question marks, or in our case, square
 brackets to the host name. And that then leads to a URL that
 will pass the validation, even though, well, that's not
 really the URL that you're supposed to access. So that's
 really the real sort of root cause here is that the host
 header, the host name of the host header is never properly
 validated. Anyway, there are patches available. The
 vulnerability is not new, sort of was described in July. And
 also then soon after that, some exploits were seen for
 it. Horizon3.ai has a good sort of brief post about
 describing this vulnerability. And recently, of course, there
 was a lot of news about thermostat or controllers for
 refrigeration systems getting compromised, in particular, as
 far as grocery stores were concerned. But apparently also
 some in hospitals like that were controlling a temperature
 of blood supply were hacked as well in this particular
 attack. Now, there's always kind of the assumption and
 some examples of some obvious vulnerabilities in these kind
 of controllers. But now Midnight Blue has published a
 brief blog post looking at actual firmware images from
 Danfoss. Danfoss is the manufacturer of one of the
 controllers that was affected by these attacks. And, well,
 they say within an hour, without the help of AI, just
 basically by doing some manual sort of browsing through the
 code, they found a number of vulnerabilities. They focus
 more on sort of buffer overflows. And they found, for
 example, in the host header, a buffer overflow that then
 leads to remote code execution. And it's actually
 relatively straightforward to exploit because these systems
 don't do stack canaries. They don't do address space layout
 randomization. So pretty much the standard compiler
 protections that you have in a lot of modern systems are not
 enabled in any of these thermostat controllers. And
 exploitation should be relatively straightforward. So
 definitely keep those systems isolated. I don't think there
 are any patches available yet for the vulnerabilities
 discovered by Midnight Blue. Not really sure if they even
 bothered to disclose them to the vendor. But definitely
 isolation is probably your best bet in particular since
 they just sort of scratched the tip of the iceberg here.
 There's probably a lot more vulnerabilities that they
 haven't looked at yet. And we've got an interesting
 authentication bypass vulnerability in Cyrus SASL.
 Now, this library is often used to authenticate email
 services like particular IMAP. I've seen it often being used
 and used it myself. The vulnerability, luckily, is
 only in the SRP algorithms. So that stands for Secure Remote
 Password. Sounds like something that you may want
 enabled. But luckily, it's usually not enabled because
 server client support is a little bit limited for this
 particular algorithm. The problem here is that an
 attacker essentially can log in with no password at all.
 They just have to set their side of the secret for a
 particular value that will then essentially shortcut the
 authentication algorithm. This is particularly described
 actually in the RFC describing SRP as something that you have
 to be aware of. But I guess wasn't properly implemented
 here. From a discovery point of view, the mail server, if
 that's how you use it, that's how it's usually being used.
 It typically lists these algorithms as part of the
 banner. And it should be pretty easy to identify any
 vulnerable systems, which of course helps defenders as well
 as attackers. You may want to do a quick scan of your
 network, making sure that it doesn't happen to be enabled
 in any of your mail servers. And then we got yet another
 TCP vulnerability in the Linux kernel. This one was named
 NoisyTCP by the discoverer and has been reported to the
 Linux kernel and a patch should be available in the
 kernel from kernel.org. Not sure whether or not it made it
 into various distributions yet. Good thing about this
 vulnerability first. Well, this is a denial of service
 vulnerability. So no indication that code execution
 is possible. But like any of these sort of kernel TCP
 vulnerabilities, all it needs is a listening TCP socket.
 Doesn't really matter what software is listening on that
 socket. Exploitability, according to the author, is a
 little bit mixed on a bare metal server. That does not
 have a lot of traffic. Apparently, it took about 14
 hours to trigger exploitation. On the other hand, on a virtual
 system, it took only a few minutes. So lots of
 dependencies here. Also, how many receive queues and such
 you have on the system. This is related to SYN
 cookies. So as a quick fix, you could just disable SYN
 cookies. I probably wouldn't recommend that sort of
 proactively. But if you do see an attack against one of your
 systems using these vulnerabilities, that's
 probably the sort of appropriate quick fix in order
 to block exploitation. Well, and that's it for today. So
 thanks for listening. Thanks for liking. Thanks for
 recommending and subscribing to this podcast. And talk to
 you again tomorrow. Bye.