Protecting the hardware supply chain

[Alan Karp]

How do you know that your chip was not manufactured with malicious changes?  Previously, you'd physically examine each layer of a chip, scraping off one layer to get to the next one.  Slow and destructive.

The May issue of IEEE Spectrum (Yes, I'm behind.) describes a non-destructive approach called ptychographic X-ray laminography that gives you a full picture of all layers.  It uses powerful X-rays to generate an interference pattern that can be used to construct a complete picture of a layer of the chip.  It currently takes 10s of hours to scan a part of a chip at 20 nm resolution, but stronger X-ray sources can reduce the scan time by a factor of 10,000.  That should make it practical to scan enough samples to reliably detect any deviations from what's in the design file.

--------------
Alan Karp

[Mark S. Miller]

This could be huge!

Link please?

Would it protect against https://web.eecs.umich.edu/~taustin/papers/OAKLAND16-a2attack.pdf ?

--
You received this message because you are subscribed to the Google Groups "friam" group.
To unsubscribe from this group and stop receiving emails from it, send an email to friam+un...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CANpA1Z0NTngbBL0X9p2PHBsV8UbVQb4Ec316LD-yPk9oD3bbQg%40mail.gmail.com.

--

  Cheers,
  --MarkM

[๏̯͡๏ Jasvir Nagra]

https://spectrum.ieee.org/chip-x-ray

-- 

Jasvir Nagra

To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CAK5yZYjPo0TK8a4sgsK2LEaX4VWC4Mq0fhrnUUjzYyCr9Se-PA%40mail.gmail.com.

[Mark S. Miller]

This is astonishing! 

To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CAK1HgMjhKzutWdES6mrr3JakRKku0stf7SY2GvkU2cJw5X_Rrg%40mail.gmail.com.

[Ben Laurie]

Pretty cool, but ... don't forget doping attacks, which this will not show.

I suspect some combination of this kind of thing and careful design + test pads is needed to defeat that.

To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CAK-_AD69BQg9OvtSN27H3A%2BX0SYL7%2Bj5UBc9q7%2BuX_jJ79mr9w%40mail.gmail.com.

[Tony Arcieri]

On Mon, Jan 2, 2023 at 6:06 AM 'Ben Laurie' via friam <fr...@googlegroups.com> wrote:

Pretty cool, but ... don't forget doping attacks, which this will not show.

Where are we at on that cat-and-mouse game?

Last I saw they were detectable via scanning electron microscopy (SEM) or focused ion beam (FIB) imaging

https://link.springer.com/chapter/10.1007/978-3-662-44709-3_7

--

Tony Arcieri

[Alan Karp]

I read the dead tree version, but https://ieeexplore.ieee.org/document/9771357.

According to the abstract of the Analog Malicious Hardware paper, the attack requires a single change to the chip.  The ptychographic mechanism was tested on such a case and detected the change.  I don't know if it would have picked up only a change in the size of a capacitor.

--------------
Alan Karp

To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CAK5yZYjPo0TK8a4sgsK2LEaX4VWC4Mq0fhrnUUjzYyCr9Se-PA%40mail.gmail.com.

[Dale Schumacher]

Amazing progress in terms of observability! I wonder how much such advances will complicate the storage of secrets.

Another approach is to use configurable hardware, so you can control the image you're running. For example, on the Precursor platform (https://www.crowdsupply.com/sutajio-kosagi/precursor).

To view this discussion on the web visit https://groups.google.com/d/msgid/friam/CANpA1Z3Wu%3D%3DY3xQJyRGv5SNAbfBXw4pNg%3DjFO8eB9YW2Eii5cw%40mail.gmail.com.