An RCE in GNU's telnetd has no relationship to the sunsetting of telnet. Something could equally likely happen with SSH (but not really because the OpenBSD folks are paranoid by nature).
Apple removing the telnet client from OS X was a stupid move. How can you call yourself UNIX and not have a telnet client? It's like removing grep or ed.
Apple still includes uucp for some unknown reason.
Telnet continues to be widely used for select use cases and being told we're naughty by not including it feels punitive and just adds extra steps. What are you supposed to do, trash a $1m piece of industrial equipment because Apple wants to remind you Telnet is insecure?
New devices are still being released with Telnet where SSH is impractical or unnecessary.
(Remember hearing about this a long time ago (from some searching I think it was in 1999 via Slashdot) and verified some instance of it still exists/works.)
Telnet is used in legacy, IoT, embedded, and low-level industrial hardware. It's also intentionally enabled on devices where automation was written for telnet and it wasn't easy to switch to ssh.
If you investigate most commercial uses of ssh, the security is disabled or ignored. Nobody verifies host keys, and with automation where hosts cycle, you basically have to disable verification as there's no easy way around the host keys constantly changing. Without host key verification, there's kinda no point to the rest.
Even assuming the host keys were verified, the popular ssh conventions are to use either long-lived static keys (and almost nobody puts a password on theirs), or a password. Very few people use SSH with 2FA, and almost no-one uses ephemeral keys (OIDC) or certificates (which many people screw up).
So in terms of how people actually use it, SSH is one of the least secure transport methods. You'd be much more secure by using telnet over an HTTPS websocket with OAuth for login.
How do you automated, for example, "HTTPS over websocket with OAuth", without providing some kind of hard-coded, static or otherwise persistent authentication credentials to the calling system in some form (either certificate based auth, OAuth credentials, etc.)?
The problem with IoT and embedded secrets isn't really a solved problem, from what I can tell. I'm not sure that OAuth exactly solves the problem here. Though all your comments about SSH (especially host verification) holds true.
Just honestly trying to understand the possible solution space to the IoT problem and automated (non-human) authorization.
Probably one of the reasons this bug survived so long is that it isn't used much for priveleged access any more, but so you can play a moo or play you an ASCII movie, as people below you are replying.
> IMHO we need a good telnet replacement that sends signed data. Most people interpret signatures as allowed under FCC rules, just not encryption.
I know from bitter experience that IPsec is a “now you have two problems” kind of solution, but the Authentication Header is a thing and is supported by most (all?) implementations. Ham radio operators probably don’t have much use for the actual features of telnet compared to plain netcat, do they? (It’s mostly terminal feature negotiation and such.)
TIL that IPsec can be used without encryption. That should work pretty well.
Telnet is mostly used for auth and straightforward terminal/BBS access in my experience. There are some other alternatives like HamSSH but I don’t think it’s that common.
Yeah instead we use DH where we "voluntarily agree" (lol) that nonces are ephemeral and that this byzantine insane stack we're using (IAM, JWTs, DH, npm, etc) is somehow secure and evolved whereas in the past, human social trust networks replaced our current security theater.
What an amazing bug. I probably spent my first 10 years on the internet just using telnet. They were wild times. You could log ethernet traffic and see passwords. Towards the end of those we started to have a few more single-user machines, but the vast majority were old school many many user machines, where "root" was thought to be tightly restricted (of course, even then, in practice it wasn't if you were in the know).
When I was an intern for some reason they issued me a voip phone for my desk. One day I got bored and figured out I could telnet into it. Nothing interesting but it was still a fun moment for me!
A very very long time ago as an intern I was working on a perl cgi script and I would often test it with telnet. I was used to messing around with hayes commands so manually typing in HTTP commands seemed like a natural extension of that.
That's crazy. This is core business critical software but they just YOLO critical changes without any automated tests? this PR would be insta-rejected in the small SAAS shop I work at.
If you think you can do better you're welcome to do better. I say this without a hint of sarcasm. This is how open source works. It's a do–ocracy, not a democracy. Whoever makes a telnet server gets to decide how the telnet server works and how much testing it gets before release.
Culture has changed a lot since the 20th century and older projects can have antiquated norms around things like testing. I was just listening to a recent podcast talking about how worrisome it is that OpenSSL has a casual culture about testing[1] and was reminded about how normal that used to be. I think in the case of telnetd you also have the problem that it’s been deprecated for multiple decades so I’d bet that they struggle even more than average to find maintainer time.
Even with automated tests you'd need to think of this exploit right? Perhaps fuzzing would have got it. The mailing lists says they proved it successful on
Any business that has a telnet daemon able to be reached by an unauthenticated user is negligent. Just the fact that everything is in the clear is reason enough to never use it outside of protected networks.
> Someone upstream of a significant chunk of the internet’s transit infrastructure apparently decided telnet traffic isn’t worth carrying anymore. That’s probably the right call.
Does this impact traffic for MUDs at all? I know several MUDs operate on nonstandard Telnet ports, but many still allow connection on port 23. Does this block end-to-end Telnet traffic, or does it only block attempts to access Telnet services on the backbone relays themselves?
MUDs use plaintext TCP protocols that are accessible to a wide range of clients.
The Telnet protocol is well-defined and not completely plaintext. There are in-band signaling methods and negotiations. Telnet is defined to live on 23/tcp as an IANA well-known, privileged, reserved port.
MUDs do none of this. You can usually connect to a MUD using a Telnet client, but most players hate the experience and often deride this method in favor of a dedicated, programmable client.
The fact that MUDs inhabit higher 4-digit ports is an artifact from their beginnings as unprivileged, user-run servers without a standardized protocol or an assigned “well-known port” presence. If you want your MUD to be particularly inaccessible, you could certainly run on port 23 now!
As a MUD enthusiast of two decades, this is not accurate. Where are you getting this information?
Most MUDs implement RFC 854, and a number of non-standard Telnet option subnegotiation protocols have been adopted for compression (MCCP2), transmission of unrendered data (ATCP, GMCP, ZMP), and even a mechanism for enabling marking up the normal content using XML-style tags (MXP). These telopts build on the subnegotiation facility in standard Telnet, whose designers knew that the base protocol would be insufficient for many needs; there are a great number of IANA-controlled and standardized telopt codes that demonstrate this, and the MUD community has developed extensions using that same mechanism.
> You can usually connect to a MUD using a Telnet client, but most players hate the experience and often deride this method in favor of a dedicated, programmable client.
I think you are confusing "telnet" the program with "telnet" the protocol. I am speaking here of the protocol, defined at base in RFC 854, for which "telnet" the program is but one particularly common implementation. You look at any of those "dedicated, programmable clients" and they will contain an implementation of RFC 854, probably also an implementation of RFC 1143 (which nails down the rules of subnegotiation in order to prevent negotiation loops), and an implementation of the RFCs for several standard telopts as well as non-standardized MUD community telopts. I can speak for the behavior of MUSHclient in especial regard here, though I am also familiar with the underlying Telnet nature of Mudlet, ZMud, and CMUD, not to mention my very own custom-made prototype client for which I very much needed to implement Telnet as described above.
Yes, perhaps we should define “MUD” and your incomplete experience of “most”.
As a MUD enthusiast for 37 years, I learned to program in C and Unix through TinyMUD, MUCK, and MUSH derived servers. From the beginning, none of these codebases implemted Telnet. There was nothing but a raw transparent TCP connection. In fact, I facilitated the introduction of a grand innovation: the "port concentrator" system which multiplexed TCP connections. Unix processes had a hard rlimit of 64 file descriptors, which crimped our style as an emerging MMORPG. The multiplexer increased this to 4096, for the biggest games of the era.
You mention MUSHclient, and I do not know about later revisions of the TinyMUSH server, but I can assure you that every MUSH I found from Larry Foard on, was not implementing Telnet. (I was privileged to help Larry "test" new features as I red-teamed his server with bizarre edge cases!)
Likewise, after I handed off TinyMUCK 2.3 to the furries, it was not doing the Telnet protocol. When we backported stuff to MUCK 1.x, it was not doing Telnet. I wrote a bonkers Perl program to read MUCK databases and sort of implement the game. No Telnet there. I've got to wonder whether the Ubermud or MOO guys had folded it in; they were close collaborators with us, back in the day.
Now as for the Diku, LP, and other “combat” type games, I’ve no idea. Perhaps they did. We never cared. I was aware that some of them had a pesky “prompt” that violated the line-mode assumptions of conventional clients and needed workarounds.
telnet(1), the program, was historically the only program that implemented the protocol. If you use Tinyfugue or Tinywar or tinymud.el, they are not, and no, I am not confused, because I was giving an example of why the Telnet-implementation, the program, the client, was so inadequate for playing on MUD servers.
It wouldn’t have been difficult to retrofit the Telnet RFC 854 into any MUD server, but none of us wizards had any use for it, seeing that our clients were mature and capable of much more processing without it.
If modern MUD servers have mostly implemented Telnet, then that is cool, but what surprises me is that it is mandatory, and your clients don’t seem to interoperate without it? That is a strange reversal!
The modern MUSH forks do generally support telnet, but yes -- as a 29 year old who's been pathologically obsessed with "MUD archeology" off and on, I'll confirm -- historically, most MUDs did not do any sort of Telnet negotiation.
Further, most older clients did not anticipate any kind of Telnet negotiation from the server, and will print garbage to the screen if connecting to modern MUSHes that do. (I've tested tinywar, vt, and that one VMS client...)
MUCKs never, to my knowledge, implemented telnet, though. They barely support ANSI escapes, nevermind Telnet. :-)
> [...] no, I am not confused, because I was giving an example of why the Telnet-implementation, the program, the client, was so inadequate for playing on MUD servers.
Then this is at the heart of our disconnect, because the post of mine that you originally replied to --- as well as, unless I drastically misread, the original article under discussion --- was concerned with traffic on port 23, the Telnet protocol port, and not with any particular implementation communicating on that port. The concern of my original comment was that this might affect MUDs that operate on port 23. Perhaps you can understand my confusion when you reply stating categorically that most MUDs do not use the `telnet` program, when that wasn't really what was at concern (and therefore implied that my question had no basis).
It is a true fact that many MUDs operate on port 23. Many do not, but you can skim a MUD aggregator like MudConnect [0] to see that it is quite common. Aardwolf, Discworld MUD, and the IRE games --- which consistently topped TopMudSites (when that aggregator was still running, anyway) all operate on 23, potentially in addition to an unreserved port.
> what surprises me is that it is mandatory, and your clients don’t seem to interoperate without it? That is a strange reversal!
All telopts are disabled by default, per Telnet RFC; the only things you must absolutely parse under the RFC are the standard complement of NVT commands (such as IAC GA "Go Ahead"), even if they are otherwise implemented as no-ops.
Any 7-bit clean input stream is treated as pure data -- with the incidental exception of bare `\r`, which must always be followed either by `\n` or by `\0`; but Postel's Law has turned that into more of a guideline. So as long as the standard NVT encoding is assumed (which is just 7-bit ASCII) and the NVT core escape sequences are avoided, a modern Telnet-based MUD client can interoperate with a 7-bit clean plaintext MUD server without issue.
Some MUD clients will eagerly send IAC DO / IAC WILL subnegotiations, but general practice is to let the server offer first -- probably precisely to ensure compatibility with MUDs that don't implement Telnet subnegotiations.
> Now as for the Diku, LP, and other “combat” type games, I’ve no idea
Diku-family MUDs are certainly the ones I have the most experience with. I understand LP MUDs also generally have Telnet support; or at least, I recall seeing a patch for them that MUD owners often sought to apply to their games.
Wouldn't that imply that >80% of all monitored telnet sessions were exploit attempts for the specific CVE in question? Even with the scale of modern botnets, that seems unrealistic for a single vuln that was undisclosed at the time.
Am I the only one who feels like it isn't the responsibility of backbone ISPs to filter traffic like this? In the case of a DDoS situation I could get behind it, but in this case I feel as though it's not Cogent's problem if I want to use telnet from a device on Charter's network to a Vultr VPS, even if it may be ill-advised.
(Of course, the article only speculates that this traffic filtering is what's going on; there isn't any hard proof, but it feels plausible to me.)
Well, I mean, the first part is a song by Don McLean called American Pie. You might know that, unsure that everyone will pick it out though.
One of the most famous play choices at karaoke bars these days too. I think because the song is a long story, of sorts? But it's a terribly long song and I will leave to take a smoke break anytime it gets chosen. You're going to be there for a good 10 minutes before it concludes.
So maybe the AI prompt was something like, "take CVE-2026-24061 and compose a song lyric in the style of American Pie by Don Mclean". I wonder if you would get similar results with that prompt.
The rest of it seems to be substantially edited by an LLM too, or at least it's composed much like LLM outputs often are these days: “not a gradual decline, not scanner attrition, not a data pipeline problem, but a step function.”
"Not X, not Y, not Z" is a common LLM tic, and there's a few more like it in there.
The design of telnet and ssh where you have a daemon running as root is bad security that as shown here is a liability, a ticking time bomb ready to give attackers root.
Oldschool telnetd didn’t actually run as root; rather, it just set up a PTY for the incoming socket to talk to, and then fork-exec’ed a /bin/login subprocess to live inside that pty. /bin/login is setuid-root, so it’s “where the security lived.”
I think we all collectively decided that that was a bad idea at some point — probably because /bin/login was never designed under the assumption that it would have to deal with arbitrary binary network traffic being thrown at it (it really only expects keyboard input.) So we switched to doing auth directly in our network daemons, since at least then “people who are aware the code is network-facing” would be maintaining it.
I think a proper architecture would not even have a root account. The server would just expose an authenticated endpoint that allows for configuration and updates to be pushed for it.
Congratulations, you've created a server that lets people have shells running as the user running telnetd.
You presumably want them to run as any (non root) user. The capability you need for that, to impersonate arbitrary (non-root) users on the system, is pretty damn close to being root.
That still needs a way to change users, and OpenSSH already has privilege separation. That hardens the process somewhat to reduce the amount of code running in the process which can change the uid for a session but fundamentally something needs permission to call setuid() or the equivalent.
I'm not sure that you need root because of the port - I think login itself needs to run as root, otherwise it cant login to anything other than the account its running under.
Those are already unprivileged operations, but how does it start the initial process in that terminal with the correct privileges for a different user?
Any breach of the daemon will still give access to a system that can approve/deny user logins. Breaching the daemon therefore allows permission escalation, because you can simply jump to an account. Chain with any local vuln of your choice to completely own the box.
It doesn't matter what user it is running as.
If this was so easy to deal with, someone would have done it. Instead, we get endless HN comments about people that act like they can do better but never submit a PR.
Breaching the daemon only allows for the attacker to get access to the login. User accounts should still be secured requiring authentication.
>If this was so easy to deal with, someone would have done it.
Sadly this is not the case. There is a lot of inertia towards solutions like ssh or sudo. It may be easy to delete them, but actually getting such a changed accepted is no trivial task.
Not the parent poster, but I also still use telnet. For me it's "Ancient", I have a few retired SPARC and PA-RISC boxes that run their period appropriate OSes as a hobby. Telnet/rlogin is the more reliable method to get into them remotely (just over the LAN).
They're on a LAN behind a NAT Router/Firewall, and I don't always keep them powered up (I'm not that insane) so I really don't have a concern for them.
Some of the more modern/high-performance examples I have run NetBSD with modern sshd and modern ciphers, but you can tell it's a bit of a workout for them.
The difference between "telnet" the program and "telnet" the protocol is especially important in this discussion, I think.
A more "proper" tool for that is netcat -- I doubt SMTP supports the Telnet option negotiations subsystem. (I also doubt SMTP servers can interpret the full suite of Network Virtual Terminal (NVT) commands that the Telnet protocol supports.) There's clearly enough similarity between the two protocols that if you're just using it to transfer plaintext it will probably work out fine, but they are distinct protocols.
I used telnet(1) as a generic TCP text client for many years before switching to GNU/BSD netcat. Nowadays, netcat is more prominent then telnet, and telnet had its corner cases with control characters.
You want nc (usually with -v) or socat. telnet is muscle memory for a lot of people (myself included sometimes) but it's a strictly inferior choice these days for poking arbitrary plaintext services.
Not the original commenter, but I noticed it too. I guess it's hard since AI is trained on human content, so presumably humans write like this too, but a few that stood out to me:
> Five entire countries vanished from GreyNoise telnet data: Zimbabwe, Ukraine, Canada, Poland, and Egypt. Not reduced — zero.
> An attacker sends -f root as the username value, and login(1) obediently skips authentication, handing over a root shell. No credentials required. No user interaction.
> The GreyNoise Global Observation Grid recorded a sudden, sustained collapse in global telnet traffic — not a gradual decline, not scanner attrition, not a data pipeline problem, but a step function. One hour, ~74,000 sessions. The next, ~22,000.
> That kind of step function — propagating within a single hour window — reads as a configuration change on routing infrastructure, not behavioral drift in scanning populations.
(and I'm not just pointing these out because of the em dashes)
GPTZero (which is just another AI model that can have similar flaws and is definitely not infallible, but is at least another data point) rates my excerpts as 78% chance AI written, 22% chance of AI-human mix.
To me at least, the article still seems to be majority human-written, though.
1. TELNET is an IETF-standard protocol defined by RFCs.
2. Telnet is a well-known port assigned by the IANA (tcp/23).
3. telnet is a client program, originated on Unix, available on many systems, and likely from a quite homogeneous codebase.
4. telnetd is a server program, also originated on Unix for the purpose of implementing Telnet protocol as a login server. Also a homogeneous codebase or two.
TFA is about items 2 and 4, and 1/3 are completely unrelated.
IIRC, the only traffic that was monitored and detected here is the scanning. The vulnerability scanners that try and detect, for better or worse, what someone's running on port 23, fingerprint it, and figure out if it's a vulnerability.
Interestingly, filtering port 23 only mitigates the CVE by happenstance. It is merely by convention that telnetd runs on port 23, so that people can use it to log in remotely. There is no constraint that requires port 23. Any other service could usurp 23/tcp for itself if the admin decrees it. So, filtering port 23 is an effective mitigation for the defaults of someone running a vulnerable server on the standard port. But it is not a panacea, and it doesn't prevent anyone from using the telnetd server, or the telnet client, except for port 23.
But it also prevents you from offering any service on port 23/tcp, lest it be filtered. You wouldn't want to run a web server, sshd, a MUD, or anything else, because your connectivity would be negatively impacted for this reason. (The common experience is that a lot of Windows SMB/NetBIOS ports are blocked, and SMTP and port 80, on a lot of consumer ISPs, although this is contrasting the ISP situation to Tier-1 transit carriers now.)
...except that port 23 seems to now be filtered across the internet at large, leading to a huge drop-off in telnet traffic over the course of days if not hours. I think it's safe to say that even if you patch telnetd, being able to use telnet over the internet is not possible in many places (including Canada, according to the data).
An RCE in GNU's telnetd has no relationship to the sunsetting of telnet. Something could equally likely happen with SSH (but not really because the OpenBSD folks are paranoid by nature).
Apple removing the telnet client from OS X was a stupid move. How can you call yourself UNIX and not have a telnet client? It's like removing grep or ed.
There's always nc hostname 23 unless you need authentication
Thats what the mystery exceptions for the Open Group macOS UNIX certification was for!
There's no UNIX requirement for telnet.
Ubuntu does not include it by default (starting 16.04?). Most most distros don't.
Two wrongs don't make a right.
Apple still includes uucp for some unknown reason.
Telnet continues to be widely used for select use cases and being told we're naughty by not including it feels punitive and just adds extra steps. What are you supposed to do, trash a $1m piece of industrial equipment because Apple wants to remind you Telnet is insecure?
New devices are still being released with Telnet where SSH is impractical or unnecessary.
What can you do with telnet that you can’t do with NC?
Why are people still using telnet across the internet in this century? Was this _all_ attack traffic?
(OK, I know one ancient talker that uses it - but on a very non-standard port so a port 23 block wouldn't be relevant)
To watch Star Wars in ASCII.
telnet towel.blinkenlights.nl https://www.youtube.com/watch?v=Mhcf6tc2jeQ
(Remember hearing about this a long time ago (from some searching I think it was in 1999 via Slashdot) and verified some instance of it still exists/works.)
IIRC the blinkenlights telnet movies have been offline for a few years already.
I connected a few times today to the IPv4 one. Have had no problems myself.
It might be the telnet filtering in action. The host responds to ping but I get nothing back on TCP/23, not even a reset.
It's still working over IPv6
Telnet is used in legacy, IoT, embedded, and low-level industrial hardware. It's also intentionally enabled on devices where automation was written for telnet and it wasn't easy to switch to ssh.
If you investigate most commercial uses of ssh, the security is disabled or ignored. Nobody verifies host keys, and with automation where hosts cycle, you basically have to disable verification as there's no easy way around the host keys constantly changing. Without host key verification, there's kinda no point to the rest.
Even assuming the host keys were verified, the popular ssh conventions are to use either long-lived static keys (and almost nobody puts a password on theirs), or a password. Very few people use SSH with 2FA, and almost no-one uses ephemeral keys (OIDC) or certificates (which many people screw up).
So in terms of how people actually use it, SSH is one of the least secure transport methods. You'd be much more secure by using telnet over an HTTPS websocket with OAuth for login.
How do you automated, for example, "HTTPS over websocket with OAuth", without providing some kind of hard-coded, static or otherwise persistent authentication credentials to the calling system in some form (either certificate based auth, OAuth credentials, etc.)?
The problem with IoT and embedded secrets isn't really a solved problem, from what I can tell. I'm not sure that OAuth exactly solves the problem here. Though all your comments about SSH (especially host verification) holds true.
Just honestly trying to understand the possible solution space to the IoT problem and automated (non-human) authorization.
Probably one of the reasons this bug survived so long is that it isn't used much for priveleged access any more, but so you can play a moo or play you an ASCII movie, as people below you are replying.
Hams use it over packet radio sometimes since encryption is forbidden on the amateur bands.
IMHO we need a good telnet replacement that sends signed data. Most people interpret signatures as allowed under FCC rules, just not encryption.
> IMHO we need a good telnet replacement that sends signed data. Most people interpret signatures as allowed under FCC rules, just not encryption.
I know from bitter experience that IPsec is a “now you have two problems” kind of solution, but the Authentication Header is a thing and is supported by most (all?) implementations. Ham radio operators probably don’t have much use for the actual features of telnet compared to plain netcat, do they? (It’s mostly terminal feature negotiation and such.)
TIL that IPsec can be used without encryption. That should work pretty well.
Telnet is mostly used for auth and straightforward terminal/BBS access in my experience. There are some other alternatives like HamSSH but I don’t think it’s that common.
One? All the talkers still use it and all the MUDs/MOOs etc. far out number the talkers.
As I understand it, greynoise is monitoring scanner traffic, so yes this would all be scans or attacks
Aardwolf works well from my work laptop. And I don’t care if someone sees what I’m doing
Do you care if they steal your account though and drop all your inventory?
The problem is the auth is plain text too and you're open to having your credentials stolen.
nethack.alt.org still maintains a telnet server!
Yeah instead we use DH where we "voluntarily agree" (lol) that nonces are ephemeral and that this byzantine insane stack we're using (IAM, JWTs, DH, npm, etc) is somehow secure and evolved whereas in the past, human social trust networks replaced our current security theater.
What an amazing bug. I probably spent my first 10 years on the internet just using telnet. They were wild times. You could log ethernet traffic and see passwords. Towards the end of those we started to have a few more single-user machines, but the vast majority were old school many many user machines, where "root" was thought to be tightly restricted (of course, even then, in practice it wasn't if you were in the know).
Anyway, just wild seeing this:
> telnet -l 'root -f' server.test
or
> USER='-f root' telnet -a server.test
Survive 11 years.
The more I work in software, the more amazed I am that anything works at all. There's likely so much low hanging fruit out there
When I was an intern for some reason they issued me a voip phone for my desk. One day I got bored and figured out I could telnet into it. Nothing interesting but it was still a fun moment for me!
A very very long time ago as an intern I was working on a perl cgi script and I would often test it with telnet. I was used to messing around with hayes commands so manually typing in HTTP commands seemed like a natural extension of that.
I did this too, lol
So eleven years ago someone put a backdoor in the Telnet daemon.
Who?
Where's the commit?
https://codeberg.org/inetutils/inetutils/commit/fa3245ac8c28...
That's crazy. This is core business critical software but they just YOLO critical changes without any automated tests? this PR would be insta-rejected in the small SAAS shop I work at.
If you think you can do better you're welcome to do better. I say this without a hint of sarcasm. This is how open source works. It's a do–ocracy, not a democracy. Whoever makes a telnet server gets to decide how the telnet server works and how much testing it gets before release.
Culture has changed a lot since the 20th century and older projects can have antiquated norms around things like testing. I was just listening to a recent podcast talking about how worrisome it is that OpenSSL has a casual culture about testing[1] and was reminded about how normal that used to be. I think in the case of telnetd you also have the problem that it’s been deprecated for multiple decades so I’d bet that they struggle even more than average to find maintainer time.
1. https://securitycryptographywhatever.com/2026/02/01/python-c...
Even with automated tests you'd need to think of this exploit right? Perhaps fuzzing would have got it. The mailing lists says they proved it successful on
- OpenIndiana
- FreeBSD
- Debian GNU/Linux
So not complete YOLO.
See https://lists.gnu.org/archive/html/bug-inetutils/2015-03/msg...
FWIW, a well known LLM agent, when I asked for a review of the patch, did suggest it was dodgy but didn't pick up the severity of how dodgy it was.
> a well known LLM agent
Which one?
Not GP, but my local Ministral 3 14B and GPT-OSS 20B didn't catch anything unless I gave some hints.
Any business that has a telnet daemon able to be reached by an unauthenticated user is negligent. Just the fact that everything is in the clear is reason enough to never use it outside of protected networks.
unless it doesn’t matter if it’s evesdropped
Traffic could be tampered as well.
There's a famous XKCD about this: https://xkcd.com/2347/
In this case the hero's name is apparently Simon Josefsson (maintainer).
I feel like we should just start saying 2347. Everyone knows what you mean.
https://xkcd.com/2347/
Ah, someone beat me to it!
Telnet's cleartext and always has been. A backdoor seems like overkill.
You still have to know the password or snoop on someone typing the password. But with this vuln, you don't. You can just get root instantly.
> backdoor
Do you mean that it's intentional? Why do you think so?
It wasn't a backdoor, just a very serious security bug. Congrats on jumping straight to conspiracy and paranoia, though.
It's only a conspiracy and paranoia if it's wrong. 11 years ago was 2015.
The scope of this CVE and the response to it are genuinely wild.
It's crazy to think that some dude is singlehandedly responsible for ultimately ending the telnet era in such a definitive way.
One for the history books.
>some dude is singlehandedly responsible
Well, one person to put up the PR and one dude to approve it - back in 2015. It isn't the security researcher's fault.
> Someone upstream of a significant chunk of the internet’s transit infrastructure apparently decided telnet traffic isn’t worth carrying anymore. That’s probably the right call.
Does this impact traffic for MUDs at all? I know several MUDs operate on nonstandard Telnet ports, but many still allow connection on port 23. Does this block end-to-end Telnet traffic, or does it only block attempts to access Telnet services on the backbone relays themselves?
Most MUDs do not use Telnet.
MUDs use plaintext TCP protocols that are accessible to a wide range of clients.
The Telnet protocol is well-defined and not completely plaintext. There are in-band signaling methods and negotiations. Telnet is defined to live on 23/tcp as an IANA well-known, privileged, reserved port.
MUDs do none of this. You can usually connect to a MUD using a Telnet client, but most players hate the experience and often deride this method in favor of a dedicated, programmable client.
The fact that MUDs inhabit higher 4-digit ports is an artifact from their beginnings as unprivileged, user-run servers without a standardized protocol or an assigned “well-known port” presence. If you want your MUD to be particularly inaccessible, you could certainly run on port 23 now!
As a MUD enthusiast of two decades, this is not accurate. Where are you getting this information?
Most MUDs implement RFC 854, and a number of non-standard Telnet option subnegotiation protocols have been adopted for compression (MCCP2), transmission of unrendered data (ATCP, GMCP, ZMP), and even a mechanism for enabling marking up the normal content using XML-style tags (MXP). These telopts build on the subnegotiation facility in standard Telnet, whose designers knew that the base protocol would be insufficient for many needs; there are a great number of IANA-controlled and standardized telopt codes that demonstrate this, and the MUD community has developed extensions using that same mechanism.
> You can usually connect to a MUD using a Telnet client, but most players hate the experience and often deride this method in favor of a dedicated, programmable client.
I think you are confusing "telnet" the program with "telnet" the protocol. I am speaking here of the protocol, defined at base in RFC 854, for which "telnet" the program is but one particularly common implementation. You look at any of those "dedicated, programmable clients" and they will contain an implementation of RFC 854, probably also an implementation of RFC 1143 (which nails down the rules of subnegotiation in order to prevent negotiation loops), and an implementation of the RFCs for several standard telopts as well as non-standardized MUD community telopts. I can speak for the behavior of MUSHclient in especial regard here, though I am also familiar with the underlying Telnet nature of Mudlet, ZMud, and CMUD, not to mention my very own custom-made prototype client for which I very much needed to implement Telnet as described above.
Yes, perhaps we should define “MUD” and your incomplete experience of “most”.
As a MUD enthusiast for 37 years, I learned to program in C and Unix through TinyMUD, MUCK, and MUSH derived servers. From the beginning, none of these codebases implemted Telnet. There was nothing but a raw transparent TCP connection. In fact, I facilitated the introduction of a grand innovation: the "port concentrator" system which multiplexed TCP connections. Unix processes had a hard rlimit of 64 file descriptors, which crimped our style as an emerging MMORPG. The multiplexer increased this to 4096, for the biggest games of the era.
You mention MUSHclient, and I do not know about later revisions of the TinyMUSH server, but I can assure you that every MUSH I found from Larry Foard on, was not implementing Telnet. (I was privileged to help Larry "test" new features as I red-teamed his server with bizarre edge cases!)
Likewise, after I handed off TinyMUCK 2.3 to the furries, it was not doing the Telnet protocol. When we backported stuff to MUCK 1.x, it was not doing Telnet. I wrote a bonkers Perl program to read MUCK databases and sort of implement the game. No Telnet there. I've got to wonder whether the Ubermud or MOO guys had folded it in; they were close collaborators with us, back in the day.
Now as for the Diku, LP, and other “combat” type games, I’ve no idea. Perhaps they did. We never cared. I was aware that some of them had a pesky “prompt” that violated the line-mode assumptions of conventional clients and needed workarounds.
telnet(1), the program, was historically the only program that implemented the protocol. If you use Tinyfugue or Tinywar or tinymud.el, they are not, and no, I am not confused, because I was giving an example of why the Telnet-implementation, the program, the client, was so inadequate for playing on MUD servers.
It wouldn’t have been difficult to retrofit the Telnet RFC 854 into any MUD server, but none of us wizards had any use for it, seeing that our clients were mature and capable of much more processing without it.
If modern MUD servers have mostly implemented Telnet, then that is cool, but what surprises me is that it is mandatory, and your clients don’t seem to interoperate without it? That is a strange reversal!
The modern MUSH forks do generally support telnet, but yes -- as a 29 year old who's been pathologically obsessed with "MUD archeology" off and on, I'll confirm -- historically, most MUDs did not do any sort of Telnet negotiation.
Further, most older clients did not anticipate any kind of Telnet negotiation from the server, and will print garbage to the screen if connecting to modern MUSHes that do. (I've tested tinywar, vt, and that one VMS client...)
MUCKs never, to my knowledge, implemented telnet, though. They barely support ANSI escapes, nevermind Telnet. :-)
> [...] no, I am not confused, because I was giving an example of why the Telnet-implementation, the program, the client, was so inadequate for playing on MUD servers.
Then this is at the heart of our disconnect, because the post of mine that you originally replied to --- as well as, unless I drastically misread, the original article under discussion --- was concerned with traffic on port 23, the Telnet protocol port, and not with any particular implementation communicating on that port. The concern of my original comment was that this might affect MUDs that operate on port 23. Perhaps you can understand my confusion when you reply stating categorically that most MUDs do not use the `telnet` program, when that wasn't really what was at concern (and therefore implied that my question had no basis).
It is a true fact that many MUDs operate on port 23. Many do not, but you can skim a MUD aggregator like MudConnect [0] to see that it is quite common. Aardwolf, Discworld MUD, and the IRE games --- which consistently topped TopMudSites (when that aggregator was still running, anyway) all operate on 23, potentially in addition to an unreserved port.
> what surprises me is that it is mandatory, and your clients don’t seem to interoperate without it? That is a strange reversal!
All telopts are disabled by default, per Telnet RFC; the only things you must absolutely parse under the RFC are the standard complement of NVT commands (such as IAC GA "Go Ahead"), even if they are otherwise implemented as no-ops.
Any 7-bit clean input stream is treated as pure data -- with the incidental exception of bare `\r`, which must always be followed either by `\n` or by `\0`; but Postel's Law has turned that into more of a guideline. So as long as the standard NVT encoding is assumed (which is just 7-bit ASCII) and the NVT core escape sequences are avoided, a modern Telnet-based MUD client can interoperate with a 7-bit clean plaintext MUD server without issue.
Some MUD clients will eagerly send IAC DO / IAC WILL subnegotiations, but general practice is to let the server offer first -- probably precisely to ensure compatibility with MUDs that don't implement Telnet subnegotiations.
> Now as for the Diku, LP, and other “combat” type games, I’ve no idea
Diku-family MUDs are certainly the ones I have the most experience with. I understand LP MUDs also generally have Telnet support; or at least, I recall seeing a patch for them that MUD owners often sought to apply to their games.
[0]: https://www.mudconnect.com/cgi-bin/search.cgi?mode=tmc_bigli...
It wasn’t clear from the article but I assumed they were filtering for the attack specifically.
Since Telnet is totally plain text that would absolutely be easy to do right?
Wouldn't that imply that >80% of all monitored telnet sessions were exploit attempts for the specific CVE in question? Even with the scale of modern botnets, that seems unrealistic for a single vuln that was undisclosed at the time.
Not at interconnect speeds
It seems like they are doing a port based block similar to how residential lines often have their SMTP ports shut off.
That said in this day and age, servers on the public network really ought to use SSH.
Between you and me telnet is not dead. Sometimes I use it to probe a port to verify it is working.
It's nice to not see C being blamed for once! ... Just good old lack of reasoning (which is most C's codebase downfall, agreeably).
It's also not DNS!
Am I the only one who feels like it isn't the responsibility of backbone ISPs to filter traffic like this? In the case of a DDoS situation I could get behind it, but in this case I feel as though it's not Cogent's problem if I want to use telnet from a device on Charter's network to a Vultr VPS, even if it may be ill-advised.
(Of course, the article only speculates that this traffic filtering is what's going on; there isn't any hard proof, but it feels plausible to me.)
I used to telnet into my POP3 account and check email by protocol. Shucks.
Your dial up ever die when you where checking your email? My ISP didn't allow for leave copy on server, I remember times I lost emails to this.
Stranger article. I wasn't able to get the main point of this article. Strangely written, but hey - I'm nob native by any means.
ps.
telnet SDF.org
just works...
it was just ai written thats why.. unexpectedly so from greynoise.
Well, I mean, the first part is a song by Don McLean called American Pie. You might know that, unsure that everyone will pick it out though.
One of the most famous play choices at karaoke bars these days too. I think because the song is a long story, of sorts? But it's a terribly long song and I will leave to take a smoke break anytime it gets chosen. You're going to be there for a good 10 minutes before it concludes.
So maybe the AI prompt was something like, "take CVE-2026-24061 and compose a song lyric in the style of American Pie by Don Mclean". I wonder if you would get similar results with that prompt.
The rest of it seems to be substantially edited by an LLM too, or at least it's composed much like LLM outputs often are these days: “not a gradual decline, not scanner attrition, not a data pipeline problem, but a step function.”
"Not X, not Y, not Z" is a common LLM tic, and there's a few more like it in there.
The design of telnet and ssh where you have a daemon running as root is bad security that as shown here is a liability, a ticking time bomb ready to give attackers root.
Oldschool telnetd didn’t actually run as root; rather, it just set up a PTY for the incoming socket to talk to, and then fork-exec’ed a /bin/login subprocess to live inside that pty. /bin/login is setuid-root, so it’s “where the security lived.”
I think we all collectively decided that that was a bad idea at some point — probably because /bin/login was never designed under the assumption that it would have to deal with arbitrary binary network traffic being thrown at it (it really only expects keyboard input.) So we switched to doing auth directly in our network daemons, since at least then “people who are aware the code is network-facing” would be maintaining it.
What do you think proper architecture would be, given that ssh needs a capability to let root logins?
I suppose it could be via a proper PAM module, which is widely supported.
Too bad the first PAM RFC was published about the same time the first be version of ssh was released.
> ssh needs a capability to let root logins
One can disable root login via SSH in /etc/ssh/sshd_config. sshd also drops root priviledges once it's running IIRC.
I use use sudo or doas as a regular user once logged in.
Does ssh need to allow root logins?
Sshing as a regular user and then sudo to root works 95% of the time…
I think a proper architecture would not even have a root account. The server would just expose an authenticated endpoint that allows for configuration and updates to be pushed for it.
You are thinking 20 years ahead. In 1995 most servers were still pets, not cattle.
Literally how else is a remote login daemon supposed to work though?
1. Start with root to bind the port below 1024.
2. give up root because you don't need it any further.
3. Only accept non-root logins
4. when a user creates a session, if they need root within the session they can obtain it via sudo or su.
Congratulations, you've created a server that lets people have shells running as the user running telnetd.
You presumably want them to run as any (non root) user. The capability you need for that, to impersonate arbitrary (non-root) users on the system, is pretty damn close to being root.
That still needs a way to change users, and OpenSSH already has privilege separation. That hardens the process somewhat to reduce the amount of code running in the process which can change the uid for a session but fundamentally something needs permission to call setuid() or the equivalent.
Yes, but changing users is a function of the shell (or maybe more specifically /usr/bin/login), not the SSH daemon.
I'm not sure that you need root because of the port - I think login itself needs to run as root, otherwise it cant login to anything other than the account its running under.
You still need to have privileges to become the userid of the user logging in. Openssh does do privsep, but you still need a privileged daemon.
The remote daemon has its own account and is given a privilege that allows it to connect a network socket to a pseudo terminal.
Those are already unprivileged operations, but how does it start the initial process in that terminal with the correct privileges for a different user?
The kernel could authenticate the user before starting it.
How does it do that?
Any breach of the daemon will still give access to a system that can approve/deny user logins. Breaching the daemon therefore allows permission escalation, because you can simply jump to an account. Chain with any local vuln of your choice to completely own the box.
It doesn't matter what user it is running as.
If this was so easy to deal with, someone would have done it. Instead, we get endless HN comments about people that act like they can do better but never submit a PR.
Breaching the daemon only allows for the attacker to get access to the login. User accounts should still be secured requiring authentication.
>If this was so easy to deal with, someone would have done it.
Sadly this is not the case. There is a lot of inertia towards solutions like ssh or sudo. It may be easy to delete them, but actually getting such a changed accepted is no trivial task.
> Breaching the daemon only allows for the attacker to get access to the login
Yes, but potentially any login. See the problem? If you compromise the gatekeeper, you are now the keymaster. Or whatever :)
I still used telnet today (had to). Unsure of the patching here. But its definitely locked down to a subset of internal use only.
Embedded? Ancient? What sort of systems are you telnetting into?
Not the parent poster, but I also still use telnet. For me it's "Ancient", I have a few retired SPARC and PA-RISC boxes that run their period appropriate OSes as a hobby. Telnet/rlogin is the more reliable method to get into them remotely (just over the LAN).
They're on a LAN behind a NAT Router/Firewall, and I don't always keep them powered up (I'm not that insane) so I really don't have a concern for them.
Some of the more modern/high-performance examples I have run NetBSD with modern sshd and modern ciphers, but you can tell it's a bit of a workout for them.
telnet isn't just for ... telnet.
.. or you can use SWAKS! For some odd reason telnet is becoming rare as an installed binary.The difference between "telnet" the program and "telnet" the protocol is especially important in this discussion, I think.
A more "proper" tool for that is netcat -- I doubt SMTP supports the Telnet option negotiations subsystem. (I also doubt SMTP servers can interpret the full suite of Network Virtual Terminal (NVT) commands that the Telnet protocol supports.) There's clearly enough similarity between the two protocols that if you're just using it to transfer plaintext it will probably work out fine, but they are distinct protocols.
I used telnet(1) as a generic TCP text client for many years before switching to GNU/BSD netcat. Nowadays, netcat is more prominent then telnet, and telnet had its corner cases with control characters.
Never heard about https://jetmore.org/john/code/swaks/, thanks for the tip.
I discovered swaks recently, god I love that tool
You want nc (usually with -v) or socat. telnet is muscle memory for a lot of people (myself included sometimes) but it's a strictly inferior choice these days for poking arbitrary plaintext services.
As long as it works, it doesn’t really matter for a quick test.
I find myself using curl telnet://server:port too often these days because telnet and nc don’t get installed.
Who actually uses the tectia ssh client instead of openssh?
Your cookie banner is very inconvenient and made me leave your website and not read the article
telnet + shijack = good times
Why would somebody read something that somebody couldn't be bothered to write? This article is AI slop.
What stood out as AI written? It felt like a well-written article by an SME to me.
Not the original commenter, but I noticed it too. I guess it's hard since AI is trained on human content, so presumably humans write like this too, but a few that stood out to me:
> Five entire countries vanished from GreyNoise telnet data: Zimbabwe, Ukraine, Canada, Poland, and Egypt. Not reduced — zero.
> An attacker sends -f root as the username value, and login(1) obediently skips authentication, handing over a root shell. No credentials required. No user interaction.
> The GreyNoise Global Observation Grid recorded a sudden, sustained collapse in global telnet traffic — not a gradual decline, not scanner attrition, not a data pipeline problem, but a step function. One hour, ~74,000 sessions. The next, ~22,000.
> That kind of step function — propagating within a single hour window — reads as a configuration change on routing infrastructure, not behavioral drift in scanning populations.
(and I'm not just pointing these out because of the em dashes)
GPTZero (which is just another AI model that can have similar flaws and is definitely not infallible, but is at least another data point) rates my excerpts as 78% chance AI written, 22% chance of AI-human mix.
To me at least, the article still seems to be majority human-written, though.
The pattern points toward one or more North American Tier 1 transit providers implementing port 23 filtering
Someone attempted to compromise my home router last week using CHARGEN. Can you imagine!
Attempted to compromise, or just port scanned?
This is about Telnetd. Not telnet itself.
IIRC, the only traffic that was monitored and detected here is the scanning. The vulnerability scanners that try and detect, for better or worse, what someone's running on port 23, fingerprint it, and figure out if it's a vulnerability.
Interestingly, filtering port 23 only mitigates the CVE by happenstance. It is merely by convention that telnetd runs on port 23, so that people can use it to log in remotely. There is no constraint that requires port 23. Any other service could usurp 23/tcp for itself if the admin decrees it. So, filtering port 23 is an effective mitigation for the defaults of someone running a vulnerable server on the standard port. But it is not a panacea, and it doesn't prevent anyone from using the telnetd server, or the telnet client, except for port 23.
But it also prevents you from offering any service on port 23/tcp, lest it be filtered. You wouldn't want to run a web server, sshd, a MUD, or anything else, because your connectivity would be negatively impacted for this reason. (The common experience is that a lot of Windows SMB/NetBIOS ports are blocked, and SMTP and port 80, on a lot of consumer ISPs, although this is contrasting the ISP situation to Tier-1 transit carriers now.)
...except that port 23 seems to now be filtered across the internet at large, leading to a huge drop-off in telnet traffic over the course of days if not hours. I think it's safe to say that even if you patch telnetd, being able to use telnet over the internet is not possible in many places (including Canada, according to the data).
I wonder if simply moving to a nearby port would work. I assume only port TCP/23 is filtered instead of filtering the telnet protocol itself.