For starters, it won't depend so much on perimeter defenses and anti-virus (although they are still needed):

all these defenses--firewalls and antivirus updates--devouring an organization's time, servers, and technology budget can be useless against even one moderately adept hacker engaging in open-source "net reconnaissance" such as simple Googling; crafty "social engineering" of fake e-mail attachments that trick recipients because they mimic messages from the boss or a client; and leveraging of cyber-break-in "toolkits" readily available online.

Scott Charney of Microsoft is quoted in the same article as saying:

"these activities alone will not make the Internet secure enough and privacy-enhanced enough for many of its potential uses."

Charney has outlined his ideas on trusted computing in this paper. (pdf)

Trust must be based in hardware (yesterday's post not withstanding) and that means a massive replacement of hardware. The federal government isn't giving out details but when the Secretary of Homeland Security says we need a Manhattan Project for security's, that overhaul must be just the starting point.