The Move to a New Anti-Virus Model

This is the first in a series of articles highlighting reasons why today’s anti-virus security solutions don’t work.

Reason #1: the Basic Model

Anti-virus software vendors still rely on yesterday’s methods for solving today’s problems: they wait for the next virus to wreak havoc and then produce a solution. That worked for a long time when a virus would take years to traverse the world. But in this fast-paced Interet-crazed world we live in today, this type of solution is no longer applicable. Now a virus can traverse the world and infect millions of computers in minutes.

In the good old days a virus traveled by floppy disk. Put a floppy in your computer and save some data to it and the virus would infect the floppy. Then unwittingly put the infected floppy in another computer and presto the new computer would become infected. (I’m skimming over a lot of detail here to make a point). So the virus’ progress was slow and steady. Anti-virus vendors had time on their side. They had the time to get a copy of the virus, dissect it, run it through a series of tests to come up with a signature string (see below for definition), put the string into a database of strings to search for when scanning your hard drive (and floppies) and release the new database to the public. Ten years ago this system worked very well.

But now everyone is connected via the Internet. Now, using email as a transport point, it doesn’t take years to gather momentum, instead it takes a matter of minutes. And here is where the model breaks. Step back and ask yourself the following question: if vendors can catch “known and unknown viruses” as their literature states, how then is it that we continue to have virus problems?

The answer lies in the fact that virus authors have been more creative in coming up with new ways to infect and wreak havoc and the software industry has not responded in kind, preferring to stay embedded in its old fashioned methodologies.

Why don’t the old ways work any more, you might ask? It’s relatively simple. Let’s go through the steps.

A virus author unleashes NewVirus via email. He mass mails his virus to thousands of people. Some, not all, unwittingly open the attachment thinking it’s from a friend or the subject is so enticing that they are fooled into opening it without thinking it’s a problem (cf. nude pictures of Anna Kournikova). The email attachment immediately starts emailing everyone in his contact list and embeds itself into his operating system so that it’s activated every time he turns on his computer.

The folks he emails in turn get fooled into thinking the email is valid and they open the attachment. Very quickly all hell breaks loose. Agencies which monitor Internet traffic see problems arising with the sudden spikes in email traffic and they begin to get calls or emails alerting them to the fact that there’s a new problem. Samples are obtained and sent off to anti-virus vendors. They pass the emails through a series of tests to analyze what exactly the virus does and how it does it. Additionally analysis is performed to extract a unique string of 1’s and 0’s to identify this attachment as none other than NewVirus. This is called the signature string. It’s important that whatever string is arrived at does not exist in any other program or piece of software; otherwise, you will get what is commonly called a false positive.

Quick digression on “false positives”: if a vendor arrives at a unique string that just happens to be embedded in Microsoft Word, then every time a user runs a scan of their hard drive, Microsoft Word will be identified as being infected with NewVirus. Users will uninstall Word and re-install only to learn that they are still infected. There will be complaints; the vendor will be forced to re-assess the signature string and re-release his list of strings and admit the error.

Typically signature strings are matched against a whole boatload of commonplace software just to protect against this occurrence, but it still happens and vendors learn to add new software to their test beds.

OK, so the vendor has arrived at a signature string. Next? Implement the string into their string database so that when their scanners are scanning they will match what’s on your hard drive to what’s in the database. After the database has been updated they release the database to their customers in what’s commonly called a “push” where they send the updates to their primary users.

If you did not buy into this service, you must know enough to log into your anti-virus vendor and update your software so that you stay current.

So where are we? The bad guy –or problem teenager- has unleashed NewVirus. NewVirus has infected thousands of computers; vendors have been alerted; NewVirus continues to infect; solutions are achieved and “pushed” to corporate clients; NewVirus continues to infect hundreds and thousands of computers; corporate clients breathe a sigh of relief and alert their users as to the new threat.

Thousands, if not millions, of computers become infected and need to be cleaned because the best way to solve the virus problem is to wait for each new virus to come along and solve on a case by case basis.

But if you sat back and said: what if? What if you categorized all the things a virus can do (or could do), built a series of computers to allow any email attachment or program to have full rein of a computer (much like it would have on your own computer – such a computer is called “honeypot”) and then analyze that computer for unwelcome behavior?

That would be a true pre-emptive strike against all malicious software. This is the behavior-based model. Such a model would actually protect you unknown viruses, along with all the known 70,000 viruses.

In part 2 we’ll discuss the risks and security failures of having distributed vendor software on your desktop.