Right, I think this is referring to a technique described in some of the earlier info releases, where the agency intercepts requests and send a fake response before the real one arrives, and ditching the real response. I'm not clear on the details - it may rely on spoofing the server's IP, falsifying DNS replies and/or manipulating data in transit.

What interests me more, and what the above poster may be asking about is this part: "By concealing its malware within what looks like an ordinary Facebook page, the NSA is able to hack into the targeted computer".

This implies a true drive-by exploit - one not requiring any user interaction. Most of the Windows malware is actually installed by the user - they're tricked into clicking something, thinking it's anti-virus, funny video, "accelerate your internet" or some other innocuous thing. The no-user-action exploits generally have been workable only for plugins, particularly Flash or Java, which the user has allowed to run without any filters.