[cryptography] Another data point on SSL "trusted" root CA reliability (S Korea)

Marsh Ray marsh at extendedsubset.com
Sat Sep 17 22:29:33 EDT 2011


On 09/17/2011 09:03 PM, Arshad Noor wrote:
> On 09/17/2011 06:37 PM, Marsh Ray wrote:
>>
>> It's not entirely clear that a trusted CA cert is being used in
>> this attack, however the article comes to the conclusion that
>> HTTPS application data is being decrypted so it's the most
>> plausible assumption.
>
> Why is it the most plausible assumption?

Did you read the article? It goes on at length about this being a
network-level attack. There is no mention of a compromised endpoint
computer and to the contrary it says they deemed "conventional search
and seizure would be insufficient".

> Isn't it far easier to replace the cryptographic libraries on PCs
> with one that has a "wrapper" that copies all payloads before
> encryption and after decryption, and transmits the payload to the
> snooper?

Sure, if you have an exploit all the target's computers. Investigators
have been known to do this at times. But if that were the case, why
wouldn't the security service have just said "yeah we hacked his PC and
installed a keylogger"?

If you were a security service, wouldn't you prefer to admit to using ye
olde everyday keylogger rather that disclosing your ability to "conduct
packet tapping" on HTTPS connections?

> Why go through the hassle of breaking a cipher when all you have to
> do is replace a few files on the target's PC to get what you want?

I never suggested anyone was breaking a cipher. I said the most
plausible attack was that a trusted CA was being used as the article
seems to describe.

I found the timing particularly interesting as it surfaced at the same 
time as Holtz's detection of the kooky certs in S Korea.

- Marsh



More information about the cryptography mailing list