Re: [White Paper] The easiest way to get around SSL

A couple of things to remember: Impersonating a server requires an=20
attacker look and act like the server in question.

To look like the server, the attacker(s) must poison upstream switches,=20
routers, or whatever to show that they are the server. The only other=20
ways I know of are to infect the client with a program that would change=20
local DNS resolves to the server in question or actually hijack the=20
upstream DNS/Proxy server. Otherwise, a complete take over of the actual=20
server is involved which is not a MITM attack.

To act like the server, the attacker(s) must have a certificate=20
generated (for SSL that is) that will pose as if this is the real server=20
-- which is really not that hard for fake certs. But for valid=20
certificates, most 3rd party companies who certify the CSR, have a lot=20
of (many headaches for those of us who have actually gotten SSL=20
certificates) checks and balances in place to prevent this type of=20
thing. This does not prevent, however, if your server has been=20
compromised and someone has stolen your certificate(s) and key(s). But=20
the problem here is making the client (the actual person) believe this=20
is the server when the are prompted to accept the certificate (alot of=20
people just hit OK -- i.e. https://mail.yahoo.com).

This is a good write up of how it is possible. I did not like, however,=20
that the author requires you to install and/or change settings to get a=20
proof of exploit to work. This is trival and anything is possible at=20
that stage. Granted, it is hard to explain that someone could, in fact,=20
pose as you, but the way to prove it was flawed (IMO).

A good test was listed in the paper later on, by changing the HTTPS://=20
to HTTP:// to see if you get the same data returned. Most sites offer=20
SSL as a way to keep private information safe in transit, but must keep=20
the normal transit methods around because some browsers do not support=20
higher level ciphers and/or bits (i.e. 128bit or higher). Also, other=20
countries, especially those on the US black lists cannot do SSL (at=20
least in a meaningful manner). So a lot of sites test positive this test=20
unfortunately (which is bad) but have to because of no other means is=20
available.

Good paper overall. Just some notes to think about while reading.

Mike Duncan
[email protected]
[email protected]
http://www.randomtask.net


Keith W. McCammon wrote:

> How does the application "impersonate" the digital certificate=20
> presented to the client by the server? I fail to understand how a=20
> successful impersonation could take place, given the information=20
> presented in your paper.
>
> Roberto Larcher wrote:
>
>> [Title]
>> The easiest way to get around SSL
>>
>> [Abstract]
>>
>> This paper explains how it is often possible, with the simple
>> substitution of a string, to get around a =93secure=94
>> implementation based on an incorrect use of SSL.
>> Please note that this document does not contain any information
>> about weaknesses of the SSL protocol; it simply shows the easiest
>> way to get around the correct functioning of the SSL protocol.
>> In this document typical =93weakly secure=94 implementation based on
>> the SSL protocol are illustrated.
>> A simple test application is also proposed to check if existing
>> implementations are indeed =93weakly secure=94.
>> This document has an informative purpose.
>>
>> [Links]
>> English version
>> http://webteca.altervista.org/download/download4.php?get=3Dwp_https2ht=
tp_en=20
>>
>>
>> Italian version
>> http://webteca.altervista.org/download/download4.php?get=3Dwp_https2ht=
tp_it=20
>>
>>
>>
>> [Contact Information]
>>
>> ing. Roberto Larcher
>>
>> robertolarcher (at) hotmail.com
>> http://webteca.altervista.org
>>
>> _________________________________________________________________
>> Personalizza MSN Messenger con sfondi e fotografie!=20
>> http://www.ilovemessenger.msn.it/
>