Why is PPTP considered insecure?

[peter-h]

Everybody says so but then everybody on the internet is copying everybody else

I did some digging and it seems that PPTP authentication is equivalent to single DES. We know that can be cracked but you must have an "interesting job" is somebody will run such an attack, and it needs to be known-plaintext.

There is an issue in PPTP with limited keyspace (not 2^56) but it appears that is the case only if a stupid password is used. If you use stupid passwords then anybody can get in, without needing any crypto expertise

It's a pity because PPTP VPNs work straight out of the box, and the cheap routers implement them as terminators which is very handy (a free VPN to your own house).

[golden_labels]

A list of the gaping holes

[Whales]

Yes pptp is really convenient.  Some routers come with OpenVPN implementations and easy QR codes to get you in, but some vendors have been retracting these features from their firmwares rather than keep up with security updates (I had a client who had the whole features vanish from his router overnight).

What problem are you trying to use pptp to solve in the first place?  Remote access to your home LAN whilst out and about? 


> We know that can be cracked but you must have an "interesting job" is somebody will run such an attack

You are not invincible to fraud by being an average person, you own a credit card.  Automated attacks are a profitable industry, the world's entire IPV4 range is constantly scanned for low hanging fruit and ways to deploy malware.


>  and it needs to be known-plaintext.

Network traffic packets have standard headers & their types can be identified by timing.  Known-plaintext is a given.

[ejeffrey]

You are better off just opening a port on your router.  If you don't need a full VPN you can just open port 22 to a trusted computer or a dedicated raspberry pi and use SSH + port forwarding.  Or you can open a port to use OpenVPN or wireguard.  It's maybe more effort that clicking the "enable PPTP" button on your router, but less likely to be compromised by an automated vulnerability scanner.

[peter-h]

What problem are you trying to use pptp to solve in the first place?  Remote access to your home LAN whilst out and about?

Yes that would be the main use of it. Securing your data stream while sitting in a dodgy 3rd World hotel where the wifi is being snooped on, and they have your full details and a passport photo HTTPS does not help much because they can see the site you are connecting to, and if it is a bank then they have all the tools needed to hit it. Of course a VPN to your house is handy for file transfer, etc. Plus, IME, certain UK banks, being incredibly stupid, will generally terminate your login if they see you login from some unusual IP (I could tell stories) hence doing banking always from your home IP is desirable.

If you don't need a full VPN you can just open port 22 to a trusted computer or a dedicated raspberry pi and use SSH + port forwarding.

Surely open ports are an absolute no-no because they will get port sniffed within hours and attacked. On better routers you can set it up to accept connections only from a given IP, but if you are travelling you will be on any IP. It is true the VPN port will be equally visible but there should be no attack surface unless you can log in.

A list of the gaping holes

I saw that. It needs a lot of expertise.

I use OpenVPN already, and L2TP, and IPSEC, but they take a lot more setting up. They also often do not work if the enroute network doesn't support them. I don't have any details but L2TP especially does not work half the time even between two UK ISPs, IPSEC is OK site-to-site only and only between same-model routers (forget it otherwise, due to the complexity) and while OpenVPN seems to work, it is nontrivial to set up. I also have some SoftEther terminators set up (set up to use 100% port 443, so it should pass through anything) but IME that one works only on a Windows laptop or desktop; it is also a bastard to set up the terminator but I got someone to do it (on Centos machines).

[Whales]

Oh yeah IPSEC is a steaming pile of shit.  I have not configured my own OpenVPN server but I've read it's a complex funscape.  I've had good luck with tinc and wireguard.

Wireguard in particular has worked well in several places I have set it up.  Notes:

* The windows client is actually alright
* Unlike pptp you don't distribute a username+password to a user, instead you give them a .conf file that contains keys.
* If your config is wrong (equivalent to a wrong password) then you won't get told, instead the tunnel will activate but it will go to nowhere.  This "silent" failure is very confusing if you're not used to it.  The motivation is that people can't even detect that you're running a wireguard server via a portscan; they need valid keys for anything to reply.  In practice it can cause some frustration when you're starting out  and not aware of it.
* I get routers to support it by buying ones that can run OpenWRT.
* Requires static IPs in the VPN internal network; does not do layer 2 or DHCP.  On the plus side this means it's really good for people with bad mobile broadband internet reception.
* Encapsulated into standard UDP packets for travel over the internet, so unlike IPSEC & pptp it requires no magic and is seldom blocked.  I've had some medical and education instrutitions block it, but that's because of blanket bans on everything except HTTPS and DNS traffic.
* Bulletproof reliable.  Compatible across years of wireguard releases.  Has never complained about anything.

and if it is a bank then they have all the tools needed to hit it.

They can only snoop on your bank login credentials if either:

(1) they impersonate the bank's site AND you click through your browser's warnings about the HTTPs certs for your bank suddenly being wrong,

or

(2) they run a similarly-spelled site, try to get you to go there and enter your login creds (but this can be done without needing a hotel anyway)

or

(3) The hotel is really a nation-state actor that can pressure a certificate authority to grant fake certs (at which point you are royally fucked in so many ways beyond the hotel).

Otherwise https + your browser's cert chain should be airtight.

[ejeffrey]

Surely open ports are an absolute no-no because they will get port sniffed within hours and attacked.

So what?

On better routers you can set it up to accept connections only from a given IP, but if you are travelling you will be on any IP. It is true the VPN port will be equally visible but there should be no attack surface unless you can log in.

It's the same thing.  You open a service you increase the attack surface.  99% of the time I would say that opening a port to a well secured computer is better than  running any externally accessible service on a typical SOHO router.  "VPN" isn't magic, it's just another service, and mostly SOHO routers are hot garbage.

[Jeroen3]

Let me introduce modern type solution to this problem: Wireguard, Zerotier or Tailscale
They have wide support, many modern routers can use them, eg mikrotik both has wireguard and zerotier,
They do not require certs!

I've installed Zerotier on my router and all my devices, my entire home network is available everywhere. On my iphone I can tick the gateway option to use it as internet vpn. But mostly I use it there as routed option (home network available to iphone, but internet not going via my home)

[peter-h]

They can only snoop on your bank login credentials if either:

Sure. What actually happens is that somebody does a personal visit to a bank branch with forged documents. You would be amazed how well that works. Online, things are tight, but in person they aren't.

It's the same thing.  You open a service you increase the attack surface.  99% of the time I would say that opening a port to a well secured computer is better than  running any externally accessible service on a typical SOHO router.  "VPN" isn't magic, it's just another service, and mostly SOHO routers are hot garbage.

Isn't there a difference that an open port enables the attacker to place packets onto your internal network (and these could be structured to exploit weaknesses) whereas a VPN port does not enable this unless you get past the login.

[Jeroen3]

The VPN obfuscates and adds additional security layer the services you expose.
The sole purpose of the VPN is to be secure. As other applications often flop this, eg: RDP is infamous for being critically insecure.

If you obfuscate what you expose it becomes harder to apply known vulnerabilities.

Changing a port number does not obfuscate what you expose. People with a reasonably fast fiber connection can scan the entire ipv4 in about 10 minutes.
Then there are also indexers, like shodan.io that do this for you to find your next target.

[peter-h]

Yes one should never run RDP on an open port. Same with the old PC/Anywhere. Only via a VPN.

But is there a known vulnerability with PPTP if you discover a router with it enabled? Unless the router software is crap, nothing short of guessing the login credentials should get you anywhere. The PPTP "vulnerabilities" are all to do with monitoring the login process and running an attack on the crypto, which is a subset of the 56 bit DES keyspace.

One has to assume that any VPN terminator is getting plenty of "attention" but are there known attacks?

[Jeroen3]

That's assuming people follow the rules... what if you don't?

https://www.cve.org/CVERecord?id=CVE-2024-5227

Authentication is not required to exploit this vulnerability. However, devices are only vulnerable if configured to use a PPTP VPN with LDAP authentication. The specific flaw exists within the handling of the username parameter provided to the /usr/bin/pppd endpoint.

And more:
https://www.cve.org/CVERecord?id=CVE-2025-38574
https://www.cve.org/CVERecord?id=CVE-2023-51736
https://www.cve.org/CVERecord?id=CVE-2022-41026
https://www.cve.org/CVERecord?id=CVE-2022-26995

No auth required. All vulnerabilities exposed when port is opened.

So aside from being cryptographically obsolete, it also has some poor implementations.
And the cryptographic issue is a major flaw when you are using it on wifi, because anyone on that same wifi can see your packets.

[peter-h]

OK so we have some shitty cheap routers having easy exploits on VPN ports.

I am not surprised. Most of them are no-name junk and I would never use TP-LINK for anything that needs to work The TP-LINK "gigabit" ethernet over mains boxes manage about 1/20 of that speed in reality.

I think that is the bottom line with all VPNs. You are relying on the terminator not having exploits. You hope that a $xxxx Cisco box will be OK but there is no guarantee. How can you ever be sure?

[ejeffrey]

If you aren't going to accept any answer why bother asking.  If you are dead set on using an obsolete, vulnerable, and poorly implemented standard, nobody is going to stop you.  But it's a bad idea.  You would be better off doing almost anything else. 

[peter-h]

I asked a simple question but all answers were circular

As to why use PPTP, a simple speed check:

DOWN UP

direct 75 18
PPTP 20 15
L2TP 5 3
Softether (port 443 only) 27 12

The last one terminates on a virtual server with a gigabit internet connection speed, so while useful if you just want some VPN, it is not useful for file transfer to e.g. one's office, and the client is windows only.

If you aren't going to accept any answer why bother asking

Why discuss anything?

[golden_labels]

I asked a simple question but all answers were circular

If you aren't going to accept any answer why bother asking

Why discuss anything?

That’s your feeling about this thread. Do you know, what is mine and maybe ejeffrey’s? If you don’t want to know, skip to the bold part.

A contention veiled as a question.

It’s a common situation and it works like following. X has some intention regarding statement P. X ask about P. But what is the intention?⁽¹⁾ If it’s knowing about P, then there is no bias: answers agreeing and disagreeing with P are equally accepted. But if the intention is to argue without producing the statement, all disagreeing answers are challenged or rejected. Because the goal is go find confirmation of one’s own belief, not to collect information.

These is the vibes I get from this thread. Against the practices currently accepted, a person believing that PPTP is the right solution, waiting for somebody to finally confirm their point.

The only reason I wrote this post is because years of experience tell me there is one more explanation. So I will give it a shot.
What kind of answers do you expect?

If you wanted advice, this is what everybody delivers. Replies in this thread reflect the current best practices. PPTP is a now abandoned solution, agreed to be best left to history due to both the questionable security and the availability of much better alternatives. It didn’t help that it is rooted in 80s telephone networks mental model, that the encryption layer is an afterthought deperately glued onto PPTP, and who created that patch. To show some specific points, examples were linked.

If you wanted an in-depth analysis, of whether a historical artifact is suitable for a very specific set of conditions, then you asked in a wrong place. Unless some expert here is very bored and wishes to entertain themselves with enganging in that task, you’ll not get the answer. Or maybe some student as a semestral assignment, hopefully under good guidance. Why? Because this isn’t something to be answered right away. It’s a tedious, thankless task, for which normally one takes a hefty sum of money. Studying outdated material and higher than normal risk of giving a bad advice, both making it even less attractive.

⁽¹⁾ Before logic pedants beat me with arrow-shaped sticks: abductively.

[Jeroen3]

As to why use PPTP, a simple speed check:

DOWN UP

direct 75 18
PPTP 20 15
L2TP 5 3
Softether (port 443 only) 27 12

Oof, only 20 Mbit PPTP is just sad. That router must be ancient. Todays stuff can do gigabit IPsec if you want.

I strongly recommend that you update your infrastructure, for full fiber VPN get a RB5009, it also does WireGuard and ZeroTier, or enter the Unifi ecosystem.

Remember PPTP is in such a way abandoned that Apple removed it from all their products back in 2016. (9 years ago)

[paulca]

You are not invincible to fraud by being an average person, you own a credit card.  Automated attacks are a profitable industry, the world's entire IPV4 range is constantly scanned for low hanging fruit and ways to deploy malware.

Exactly, but most people think it's about them as an individual, not just that they were in the "low hanging fruit" category that day.

Also people assume if they get hacked they will know they are, because they think "hacked" means damage or theft to them.  A lot of attacks just want to become a user of your computer.

[Hogwild]

I'll just put out there that you don't even have to buy a newer router to do OpenVPN, Wireguard, Tailscale. All three can be done on FreshTomato. IIRC, all three can be done on OpenWRT. Wireguard is a big leap in security and even more so in speed. With the "client" in your router, you don't have to install and config. separate client software on each client device.

[5U4GB]

What problem are you trying to use pptp to solve in the first place?  Remote access to your home LAN whilst out and about? 

Came here to ask the same thing, what's your threat model?  It has a bunch of known vulnerabilities, as does your front door lock and the padlock you have on your garden shed, but if it's just to keep nuisances out it'll be good enough.

[peter-h]

Only that PPTP is more likely to work "out of the box" than the others.

AIUI, a PPTP VPN terminator port, just sitting there, is no more vulnerable than any other VPN port. The "vulnerability" is the somewhat shorter key length, to someone who intercepts the login sequence.