Loadable crypto for Linux

[PiB]

I have been working on making the Windows loadable crypto code work on Linux. This turned out to be mostly removing the #ifdefs and implementing
some functions in loginserver/Encryption.cpp with the Linux equivalents.

This seems to be working well for me, but one downside is that it requires recompiling EQEmuAuthCrypto as a shared library (.so). I have tried to do
this but the version I have is not compiled with -fPIC:

FAILED: : && /usr/bin/c++  -fPIC -O2 -g -DNDEBUG   -shared -Wl,-soname,libEQEmuAuthCrypto.so -o loginserver/libEQEmuAuthCrypto.so
 loginserver/CMakeFiles/EQEmuAuthCrypto.dir/EQCrypto.cpp.o  ../dependencies/libEQEmuAuthCrypto.a ../dependencies/libcryptopp.a && :

    /usr/bin/ld: ../dependencies/libEQEmuAuthCrypto.a(EQEmuAuthCrypto.o): relocation R_X86_64_32 against `.rodata' can not be used when making
 a shared object; recompile with -fPIC

    ../dependencies/libEQEmuAuthCrypto.a: could not read symbols: Bad value
    collect2: error: ld returned 1 exit status

To test this I have created a similar library that does not use any encryption (except for hashing the password). Using it is similar to how it is done
on Windows, by editing login.ini. One difference is that plugin should be a full path or explicit relative path:

[security]
plugin = ./libEQuilibrePlaintextAuth.so
mode = 5

I have commited these changes to a GitHub fork if anyone is interested: https://github.com/pixelbound/Server...e-crypto-linux.

[KLS]

I've thought about perhaps redoing the crypto system for this at some point.

[Secrets]

Originally Posted by KLS

I've thought about perhaps redoing the crypto system for this at some point.

Honestly, does SOE even use this crypto system anymore? Is there a point in keeping it 'secret' from the general public?

I remember asking about this ages ago and it sure doesn't make sense to me personally that we are 'guarding' this. The only purpose to having it closed source right now is to appease people who don't even contribute to this project anymore, and likely don't care about EverQuest.

It's not like EQLive accounts go for as much as ivory horns anymore...

[KLS]

They don't really, you login via launchpad now which uses TLS.

[lerxst2112]

You can still login with just the client. The ease of switching accounts in the launchpad is cool though.

[PiB]

I have no idea if it is used any more on Live but this 'crypto system' is so weak and broken that they shouldn't be using it in 2013. I do respect keeping it closed, though ultimately it's your decision.

[image]

Originally Posted by PiB

I have no idea if it is used any more on Live but this 'crypto system' is so weak and broken that they shouldn't be using it in 2013. I do respect keeping it closed, though ultimately it's your decision.

+1

Triple DES is quite weak and they do not use it anymore. But I think everyone already knew my stance on this for years Least there is a public login server available now though officially.

[KLS]

So long as you log in via the launcher it's fairly secure. Basically using TLS which is what you often use logging into a website via https.

I might release it one of these days since at this point it's basically superseded by their normal login process. Login is kind of a pain on linux atm and I'd like to see it be easier to build but I'll have to think about it.

[sereal]

Originally Posted by KLS

So long as you log in via the launcher it's fairly secure. Basically using TLS which is what you often use logging into a website via https.

I might release it one of these days since at this point it's basically superseded by their normal login process. Login is kind of a pain on linux atm and I'd like to see it be easier to build but I'll have to think about it.

Can you clear up this whole deal? From what I can understand reading old threads someone cracked the crypto Live used and gave it to the eqemu devs on the condition it not be distributed (because it would endanger live accounts). (in other words we are insecurely authenticating to eqemu?)

Is the reason we cannot authenticate securely using more modern methods due to constraints in the client? (being we can't patch it)

A separate note - someone I talked to mentioned passing around binaries like this may violate crypto export laws?

Excuse my ignorance on the subject. I'm really curious more than anything.

[KLS]

A separate note - someone I talked to mentioned passing around binaries like this may violate crypto export laws?

The laws were relaxed and now have an exception that many consumer products (and to my knowledge EQEmu) fall under.

(in other words we are insecurely authenticating to eqemu?)

The client is quite insecure in how it sends passwords when you don't use the live-launcher.

[sereal]

Originally Posted by KLS

The client is quite insecure in how it sends passwords when you don't use the live-launcher.

By 'live-launcher' do you mean the eq client as of now(ie it's been patched to fix any insecurities) or something else that was present in Titanium and or underfoot?

Could there be a possible work around where we run a insecure login server on the users computer that eq authenticates to, it then communicates securely to the eqemulator official login server.

+==== users-computer =========+ 
|+---------+    +----------+  |              +---------------+
||eqclient |<-->|fake login|<---internet---->| eqemulator    |
|+---------+    +----------+  |              +---------------+
+=============================+

The downside would be that the user needs to install another program. Servers could support either insecure(meaning they run the closed crypto binary), secure or both.

Am I way off base and or missing something here?