Slashdot Mirror
Can You Trust Anti-Virus Rankings?
Slatterz writes "It seems nobody can agree on a universal set of tests for rating anti-virus software, with Eugene Kaspersky the latest to weigh in on the topic, criticizing the well-known Virus Bulletin 100. Kaspersky is one of several big anti-virus brands to fall foul of the VB100 tests, reportedly failing to pass a recent test of security software on Windows Server 2008, along with F-Secure and Computer Associates. At Kaspersky, bloggers have pointed out that they don't focus on detecting PoCs, calling it a 'dead end,' and saying their anti-virus database focuses on 'real threats and exploits.' 'I don't want to say it's rubbish,' Kaspersky told PC Authority. 'But the security experts don't pay attention to these tests. It doesn't reflect the real level of protection.'"
Norton is an utter piece of crap, it would be advisable to get rid of it now
Anti-Virus is outsourcing the problem of deciding what is good to execute on your computer to a vendor who works backwards and blind.
... and keeping up with that growth rate is not a game I'd want to play! My list of "good" software doesn't increase on a quadratic growth rate, does yours? If this were any other field of computation, the signature approach would have been laughed off the planet by now.
It's "backwards", in that you don't tell them what is "good". They try to guess what would be on your "bad" list. As everyone here knows, it turns out that the "bad" list is much, much longer than the "good" list. In 2007 alone, F-Secure added more virus sigs to their products than the totality of sigs accumulated from the previous 20 years! And last I heard from them, 2008 was projected to double 2007. That sounds almost like quadratic growth to me
It's "blind" in that they aren't seeing what is actually running on your computer. For privacy (and performance) reasons, nobody provides metrics back to AV vendors about all of the executables that weren't labeled "bad", and rarely do the metrics about what is labeled "OK" actually go back to them. The AV vendors have to take a shot in the dark. They can simulate what they think your computing environment looks like, but it's just a guess. They cannot know if you have custom or proprietary software that matches one of their AV sigs unless they actually test that particular program against their sigs (and you don't let them do that, hence the "blind" remark).
Backwards and Blind is very problematic. Every once in awhile, we hear about fiascos like Symantec deciding an asian language DLL is a virus, killing all of their asian customers' windows installs for a day or two.
The question the benchmark is really trying to answer is: Which vendor's product is best tuned for the least amount of false positives and false negatives? When we should really be asking the question: Do I know what is good to run on my computers? And if the answer to that is "yes", then we should be asking the question: Why can't these vendors make a product that only allows my "good" programs to execute and nothing else?
libertarian: (n) socially liberal, financially conservative; neither left, nor right.
Wow, solid, well supported argument right there.
Indeed, it is. Norton really is a load of crap. It is a resource hog of cpu, memory and hard drive. I believe the only reason it is found on anyone's PC is because Norton pays PC companies to install it by default. Because, frankly, you would have to literally know nothing about AV to choose Norton. As in, you did no research and picked the shiniest box off the shelf. At which point, I have lost sympathy for the user.
My company relies on SOPHOS. In 12 years of working with SOPHOS, never has a virus had a chance to spread...despite the users best efforts.
Bearded Dragon