SSH vs SSL/Telnet
tyr asks: "I am setting up a masquerade server for our church to access the Internet. I want to be able to remotely administer the server, but other people have concerns about security. I have decided to implement some type of encrypted login. I have heard rumors that SSH compresses the stream to cut down on the bloat caused by encryption. Is this really significant for a text only login? Anyway I just wanted to see discussion on the relative technical merits of each of the options like SSH 1.x, SSH 2.x, OpenSSH, SSL/telnet, and any others you would recommend."
Yeah, but the ssh 1.5 protocol (the one implemented by all ssh 1.x clients) has, IIRC, some inherent flaws that allow it to be cracked; further, ssh 1.2.28 doesn't demonstrate the same attention to detail that OpenSSH has demonstrated, like (for instance) the recent RSAREF bug that OpenSSH prevented from being tickled.
:)
OpenSSH, while vulnerable when implementing the 1.5 ssh protocol, also supports an incompatible "1.6" ssh protocol that is immune to the known vulnerabilities of the 1.5 protocol; between OpenSSH clients, this protocol can be spoken instead, improving security.
I'm not great at security, but I try to pay attention to those who are
--Matthew
You won't notice a difference between SSH and Telnet for text login, even over a modem. This means of course you should go with SSH. It's a lot more featureful than telnet, and has been around so much longer than all the other altnernatives you mentioned that it's bound to be more secure. Also, the distinction between SSH and TelnetSSL must be made - tunneling telnet through SSL encrypts your communications, so your session won't get sniffed, but anyone who has your public key can still access the box just like regular telnet (assuming you're not using a firewall). Obviously, this isn't that much of a leap over plain old telnet - they still get a login: prompt at will, which with enough motivation will can be cracked. OTOH SSH can disable password based logins altogether and use RSA for authentication, where you create a private/public keyblock that you must supply a decryption key to login with. The only way someone will be able to login to your box period is if they get ahold of and crack your private key, which at 1024 bits would take several billion years IIRC, or if they install their own public key on the server, which makes this whole discussion a moot point. So I would use SSH.
--
I think there is a world market for maybe five personal web logs.
openssh would seem to be the way to go. the plaintext of most encryption schemes today is compressed, btw. this is not so much to reduce the size of the encrypted traffic (it's usually pretty similar in size), but to help hide patterns in the plaintext. various modes of encryption also help hide these patterns. as other people have said, this compression probably won't increase your throughput that much, though, on a text based ssh session. there is a small delay at the beginning of the session, where the session key is generated and encrypted. if you can live with that second delay, openssh is the way to go.
i'm not a security expert, but i have had ample opportunity to ponder this and related questions. my (admittedly basic) research has led me to these conclusions.
some random notes:
(darren)
one should definately go with OpenSSH, as it is 100% compatible with ssh 1.x. Hardly anyone uses ssh 2.x because of it's licencing issues. There are numerous Free/open clients for the ssh 1.x protocol for many platforms. OpenSSH was based on the last opensource version of ssh 1.x. The OpenBSD team fixed all of its bugs and ran it though a security audit. It is very tight. They are a whole lot quicker about fixing new bugs then the normal ssh guys. The recent buffer overflows and DOS vuenrablities are nonexisitant in OpenSSH.
A wealthy eccentric who marches to the beat of a different drum. But you may call me "Noodle Noggin."
Quando Omni Flunkus Moritati