This article is more than 1 year old
Security firms plot revamp to minimise false alarms
Whitelisted addresses to reside in heavenly cloud
Analysis Increased incidents of false positives have encouraged anti-virus firms to re-evaluate their signature update process.
Last week, misfiring updates from Symantec falsely categorised Spotify and Adobe Flash as malicious in two separate incidents. The week before a ropey update from Kaspersky Lab falsely flogged Google AdWords as malign.
False positives are a well known problem with anti-virus scanners that affect all vendors from time to time. Vendors say the problem has grown in frequency recently, as a result of the need to issue more frequent updates to combat an increase in the production of malware from the bad guys.
Luis Corrons, technical director of PandaLabs, explained: "Looking at it from a mathematical point of view, the anti-virus industry has improved a lot. 15 years ago we were detecting just a few virus a day and still had false positives. Nowadays, we're detecting more than 50,000 new malware samples a day and there are less false positives in proportion. However, at the end the total number of false positives is much higher than in the old days."
"The problem nowadays is clear: the number of new threats has grown exponentially, and to deal with it antivirus companies have automatic processes that take care of most of them; these processes are good, but not 100 per cent error-proof," he added.
Incidents of false positives can be minimised by greater use of whitelisting and the adoption of cloud-based technologies, Corrons explained.
"To minimize these problems, the best approach is to have a huge whitelist with all the good files: doing that, the anti-virus would know the file is good and wouldn't detect it. But the number of good files is almost endless, so to apply this is to antivirus products you need a 'cloud' approach."
"In the cloud there are almost no limits in size, and you can have that information without loading it in the local computers. And not only that, even if a false positive comes out, working on the cloud means that a fix can be applied within minutes, minimizing the damages."
Fraser Howard, principal virus researcher at SophosLabs, agreed that signature update processes need to change in order to cope with the estimated 50,000 new samples a day that the labs of anti-virus vendors have to process. This increase in malicious code volumes means that labs have to improve existing automatic processing of malware samples and add "aggressive generic detections" for malware.
"The simplest mitigation against false positives is scanning known clean data prior to release of anti-virus updates," Howard told El Reg. "But even the biggest false positive rig in the world is never going to have all the world's files on it. It's not just executables, you need all file types."
Howard explained that even though such a testing regime is impossible it is still possible to refine techniques. For one thing rapid inclusion of popular, clean files onto the false positive rig is needed. "The days when simply importing MSDN CDs was sufficient are long gone," Howard said.
Greater recognition of clean file characteristics so that safe files are not wrongly thought to be look-a-likes with existing malware strains is also important.
Howard agree with Corrons that better whitelisting within products and cloud lookups can also reduce the risk of a false positive.
Taken collectively, the changes forced on the security industry represent a great deal of work, but are necessary in order to minimise incidents of false alarms that invariably cause user confusion and inconvenience.
"There is a danger that anti-virus companies who do not put enough emphasis on false positive mitigation are more likely to push out an update for new malware which false positives," Howard concluded. ®