[Cryptech Tech] [FORGED] News item: Major HSM vulnerabilities impact banks, cloud providers, governments

[Leif Johansson]

On 2019-06-12 06:27, Peter Gutmann wrote:
> Warren Kumari <warren at kumari.net> writes:
> 
>> Major HSM vulnerabilities impact banks, cloud providers, governments
>> https://www.zdnet.com/article/major-hsm-vulnerabilities-impact-banks-cloud-providers-governments/
> 
> From TFA:
> 
>   The duo's research paper is currently available only in French,
> 
> Devilishly clever!  That way Thales and Gemalto can fix their HSMs while the
> non-French-speaking hackers have to wait for Black Hat to find out what the
> vulns are.
> 
> Despite the inexplicable lack of being taught phrases like "couche de resine
> epoxy" while still learning everyday useful things like "le ballon tombe dans
> les fleurs", the gist of the paper is that running externally-updatable
> ancient unpatched Linux (an unstripped, unhardened 2.26 (!!!) kernel) with
> buggy PKCS #11 firmware on your HSM isn't a good idea.
> 
> This isn't really an HSM, it's more an IoT device with a crypto accelerator
> attached.  Once I read to the description of the configuration, my only
> surprise was that it took this long to get pwned.  Not wanting to downplay the
> authors' achievement, but it's a hack of a generic, run-of-the-mill IoT
> device, just one that happens to be advertised as an HSM.
> 
> It's also not surprising that you can attack the PKCS #11 API directly, as the
> authors correctly point out it's very complex and therefore has a very large
> attack surface.  I'm sure many PKCS #11 client-app developers have
> inadvertently "attacked" their PKCS #11 implementation just by passing in
> incorrect parameters while developing code (I have, for several
> implementations).
> 
> In addition, with what they're running as the firmware as an indicator, it's
> also not overly surprising that the crypto code itself is of, uhh, sub-par
> quality.  Sorta confirms the comment I made in my book that "A great many
> security systems in use today are secure only because no-one's ever bothered
> attacking them".
> 
> All in all a nice piece of work, and an interesting read.
> 

My french is a bit rusty but did you also understand that the attack
was based on the ability to do fw upgrade over P11? Needless to say
that seems like a ... unique property to have in your P11 impl but
I am worried I have misread the paper.

	Cehers Leif