[kinsomo]

> Their dependence on American silicon may in fact be their greatest security risk. It's clear that the US and its allies will weaponize this advantage (see Stuxnet)

Please elaborate on this point, my understanding is Stuxnet relied purely on software vulnerabilities (to spread and to take control of the centrifuges). I don't think it used any hardware backdoors or anything. If it did, I'd have expected the roar to be enormous, probably several times what we just had with Spectre/Meltdown.

[dnomad]

The actual attack vector relies upon drivers signed by Taiwanese chip makers. (Taiwan happens yup be a country whose continued existence depends entirely upon the US.) If you depend upon say a Realtek card reader you're vulnerable.

[kinsomo]

> The actual attack vector relies upon drivers signed by Taiwanese chip makers.

But that's not a hardware attack. IIRC, Stuxnet used driver signing certificates to bypass some OS-level safeguards. It's quite possible that the needed certificates were stolen from a manufacturer that poorly protected them.

When you said:

>> Their dependence on American silicon may in fact be their greatest security risk. [emphasis mine]

The implication is that American silicon itself is backdoored. That may be true, but I don't think Stuxnet is in any way an example.

[dnomad]

> The implication is that American silicon itself is backdoored.

There was no such implication. Compromising hardware never makes sense unless you can intercept the hardware en route (something the NSA has been known to do [1]).

> It's quite possible that the needed certificates were stolen from a manufacturer that poorly protected them.

It's also possible that unicorns exist. Considering that certificates have been "stolen" from Taiwanese firms multiple times [2] I'd say it's not irrational to consider the possibility that these firms either are directly or via the Taiwanese government cooperating with US cyberattacks.

All of this indicates that yes, relying on such foreign chipsets is huge security threat. China imports an incredible $200 billion a year in such chips so even putting aside the huge technological attack surface associated with such a dependency the dependency constitutes an immediate economic vulnerability.

[1] https://arstechnica.com/tech-policy/2014/05/photos-of-an-nsa...

[2] https://www.wired.com/2015/06/foxconn-hack-kaspersky-duqu-2/

[kinsomo]

> I'd say it's not irrational to consider the possibility that these firms either are directly or via the Taiwanese government cooperating with US cyberattacks.

But to be totally clear: that's no more than speculation.

It's also not irrational to consider the possibility that there's been no cooperation but that the certificates were stolen.

> It's also possible that unicorns exist. Considering that certificates have been "stolen" from Taiwanese firms multiple times [2]

Malware that uses stolen certificates is less unique than once thought. If a group building a bank trojan can steal certs, I'm sure state intelligence agencies can too.

https://arstechnica.com/information-technology/2017/11/evasi...

There is also some reason to think their certificates would be targeted for theft, because code signed by those firms would be some of the least conspicuous. There are a lot of Taiwanese firms that make a lot of low-profile specialized support silicon that's literally everywhere (Sound, USB, Wifi, etc), and a driver signed by one will arouse less suspicion. Inconspicuousness would be a high priority for a nation-state hacker trying to avoid detection.

The possibility that the Stuxnet and Duqu certs were stolen is speculation too, but it's less inflammatory and more likely in my judgement.

It's also worth noting that getting explicit cooperation from a company to use their certificate would be risky for clandestine nation-state operation, since the more organizations that know about aspects of it, the more likely it will fail. If word got out that a particular code signing cert was shared, a rival actor could focus attention on suspicious code signed by that cert and be more likely to detect it.

>> The implication is that American silicon itself is backdoored.

> There was no such implication. Compromising hardware never makes sense unless you can intercept the hardware en route (something the NSA has been known to do [1]).

That's wrong. If the silicon is comprised from the get-go, there's no need for an interception step.