Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

doc: inform users that fuses are blown *before* shipping #119

Closed
wants to merge 9 commits into from
Closed

doc: inform users that fuses are blown *before* shipping #119

wants to merge 9 commits into from

Conversation

ghost
Copy link

@ghost ghost commented Apr 26, 2023

The documentation is extremely vague about exactly who blows the fuses on the Tillitis. Based on your pre-release hardware and common sense about user-controlled hardware, I and several others assumed that the customer is the one who does this.

It was quite a rude surprise to discover that this is no longer the case as of ~three weeks ago.

You should be more honest about this, particularly at the time that orders are placed from your shop.

secworks and others added 8 commits April 3, 2023 09:57
Signed-off-by: Joachim Strömbergson <joachim@assured.se>
Signed-off-by: Daniel Lublin <daniel@lublin.se>
Signed-off-by: Joachim Strömbergson <joachim@assured.se>
Signed-off-by: Joachim Strömbergson <joachim@assured.se>
Signed-off-by: Daniel Lublin <daniel@lublin.se>
Signed-off-by: Daniel Lublin <daniel@lublin.se>
Signed-off-by: Joachim Strömbergson <joachim@assured.se>
The documentation is extremely vague about exactly *who* blows the
fuses on the Tillitis.  Based on your pre-release hardware and
common sense about user-controlled hardware, I and several others
assumed that the customer is the one who does this.  It was quite a
rude surprise to discover that this is not the case.

You should be more hoenst about this, particularly at the time that
orders are placed from your shop.
@ghost
Copy link
Author

ghost commented Apr 27, 2023

Also, isn't this a significant security risk for postal interception attacks?

A postal interception attacker can order the same FPGA off of digikey and either swap it onto your boards with a hot air gun or simply have their own boards made (the board files are on github after all). Replacing or replicating the plastic casing isn't hard.

If users don't expect to load the bitstream and blow the fuse themselves, the attack cost is well below $500. If users expect to load the bitstream and blow the fuse themselves, the attacker needs to fab their own chip that's capable of imitating an ICE40 yet still betraying the customer. If you add a simple "exercise the FPGA with a few general-purpose bitstreams which look nothing like a pico-RiscV" step to the setup process the attacker really does need to fab their own backdoored FPGA (with NVRAM!) -- and now we're talking about an attack cost in the ~$1million range.

This kind of "leverage" -- being able to jack up the interdiction attack cost by orders of magnitude -- is a unique feature of user-programmed FPGAs. I can't imagine why Tillitis would suddenly abandon this wonderful feature of the strategy it had until three weeks ago. What's up?

According to Lattice's docs1 the "security fuse" that you blow out not only prevents loading new bitstreams, but also prevents reading back the written bitstream. So users can't even verify that the bitstream on their devices is what they expect it to be.

Footnotes

  1. I'm still coming up to speed here, I've only worked with Xilinx+Altera stuff and that was a while ago

@mchack-work
Copy link
Member

We will update the documentation and the shop to state very clearly
that the one currently for sale is a locked-down version meant for not
necessarily technical end-users.

Empty TKeys and a programmer board similar to the ones we handed out
during Open Source Firmware Conference will be available later for
users who want to provision their TKeys themselves.

That this hasn't been clear is our mistake.

Regarding the postal interception attack: You can verify the TKey you
got through the mail with the tkey-verification program which we point
to in

https://tillitis.se/getstarted/

https://github.com/tillitis/tkey-verification

This won't verify the bitstream but it will verify that the computed
CDI is the same as when we provisioned it (thus proving the presence
of the same UDS in the bitstream) and that the firmware is unchanged.

This means that to succeed an attacker has to somehow get the UDS out
from the locked NVCM. Probably not impossible but quite hard to do.
Either that, or attack our infrastructure for the vendor signatures.
Probably much easier than getting something out from locked NVCM.

On the other hand, if we don't lock the NVCM with the security fuse
it's trivial to read out the UDS since it's part of the bitstream.
This means less protection for an unsuspecting non-technical end-user.

@secworks
Copy link
Contributor

First, thanks amjoseph-nixpkgs for the feedback. I will update the threat model to clarify what locked into NVCM means in the given release.

@secworks
Copy link
Contributor

secworks commented May 2, 2023

I've tried to reword the text for the Bellatrix release to (hopefully) better explain what locked down means and how the UDS and UDI comes from:

6ee091b

@ghost
Copy link
Author

ghost commented May 5, 2023

Either that, or attack our infrastructure for the vendor signatures.

... and the cost of attacking that infrastructure is very considerably smaller than the cost of fabricating a custom NVRAM FPGA that can pass the ICE40 test vectors.

Also -- trying to put things delicately here -- there is of course another attack, and ever since the events on April 18th at your parent company I think perhaps that possibility should not be treated so dismissively?

After giving a lot of thought to my disappointment at how things have turned out, it's become clear that TKey is merely a prototype -- you're now likely working on raising funds to push the RTL through some synthesis tools and buy a maskset. If so, then good luck and godspeed. That's become a very crowded space lately -- everyone and their dog seems to be doing a "secure element" design this year.

Unfortunately the rest of us are now back where we started, without a stable, long-term reliable source for a product with the unique security properties of a user-programmed FPGA-based token.

@secworks
Copy link
Contributor

secworks commented May 5, 2023

Tillitis will start selling unlocked TKey devices, as a complement to the provisioned TKey end user version we are selling today.

@SallSim
Copy link
Contributor

SallSim commented May 8, 2023

There has been an unfortunate misunderstanding. A few persons have already reached out though email and we will send unlocked TKey's to all those persons that emails us and present a valid order number. We will honor orders placed during April and send the same amount of unlocked TKey's, free of charge.

@ghost ghost closed this by deleting the head repository Aug 5, 2023
@ghost
Copy link
Author

ghost commented Sep 9, 2023

Hey, I want to thank you guys for following through on this and offering unlocked TKeys through your web store:

Thank you so much for doing this; I ordered ten of them and will be recommending them to several other people, including a client.

It was gracious of you to offer free unlocked keys to those of us who ordered before the clarification, but being able to order unlocked TKeys -- and being able to tell others where they can buy their own unlocked TKeys -- is what really counts. It makes it worthwhile to invest effort into this ecosystem, which is something I would not be able to justify if there were only a handful of unlocked keys in existence.

Thank you once again for adding this to your webshop!

@dehanj
Copy link
Member

dehanj commented Sep 11, 2023

Thanks for putting a good word out for us, and we are happy to hear that you want to invest time into the TKey ecosystem!

This pull request was closed.
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment
Labels
None yet
Projects
None yet
Development

Successfully merging this pull request may close these issues.

None yet

5 participants