The InfoSec Blog

The Death of Antivirus Software

Posted by Anton Aylward

The real issue here isn't Ubuntu, or any other form of Linux.
Its that AV software doesn't work.

There are over 50,000 new piece of malware developed and released daily. The very nature of the AV software models that John McAfee foisted on the industry simply can't cope.

This isn't news. Signature-based (and hence subscription based and hence that whole business model) AV is a wrong headed approach. As Rob Rosenberger points out at Vmyths.Com, we are addicted to the update cycle model and its business premise is very like that of drug pushers.

What's that you say? Other types of AV? Like what?

Well, you could have a front-end engine that checks all downloads and all email and all email attachments and all URL responses by emulating what would happen when they run on any PC or in any browser or any other piece of software such as any of the PDF readers you use, or any of the graphical display software you use or any of the word processors you use
or any of the spreadsheet programs you use or any music players you use ... and so on.

Many people in the industry - myself included - have proposed an alternative whereby each machine has a unique cryptographic ID and the legally and properly installed libraries are all signed with that ID, and the program loader/kernel will only load and execute correctly signed code.

Yes, Microsoft tried something similar with ActiveX, but that was signed by the vendor - which can be a good thing, and used PKI, which can also be a good thing. But both can be a problem as well: go google for details. A local signature had advantages and its own problems.

The local signature makes things unique to each machine so there is no "master key" out there. If your private key is compromised then do what you'd do with PGP - cancel the old one, generate a new one and sign all your software with the new one.

The real problem, though, is not in having the key compromised but is the problem that has always existed - its the user. Right now, we have many remote code execution blockers. Your browser might be able to block the execution of Java or JavaScript, but does it? Most people either don't bother setting their defaults to "no execution" or just say "yes" to the pop-up asking them to permit execution.

No technical measure can overcome human frailty in this regard.

Enhanced by Zemanta