This article was heavy on FUD and speculation, and light on facts (expected that). The trojan (if it exists) would either have to work with the O/S, in which case it would either work in *nix or Windows, but not both; or it would talk directly to the hardware bypassing the O/S, so encrypting your data stored on your hard drive should protect it. Keyloggers are a different story though and I see where a keylogger could send data by talking directly to the hardware and bypassing the O/S. I wonder if you might get more protection from that by not using your motherboard's Ethernet interface at all and install a separate network card instead? Then the trojan would need to have its own TCP/IP stack and drivers for the NIC, and that's less likely (I hope).
The subject matter of the article is sound, however the execution leaves me skeptical that this isn't just hobbled together FUD for the sake of traffic/audience. The last two paragraphs sealed the deal for me.<p>I just hope our geeks are working just as hard as their geeks.
I see a lot of skepticism in the original article's comments about whether this is possible or not.<p>Yes, it'd be hard to hide a whole TCP/IP stack in hardware (though perhaps it could be in ROM if we're talking about CPUs), but how hard would it be to hide a simple UDP-based key logger?<p>Look at the hundreds of thousands of pwned machines that are part of spam botnets. Clearly, machines can be co-opted.<p>And we have plenty of evidence that China is involved in a (asymmetric) full-out cyberwar with us.