The in-house reputation system used in Internet Explorer 8 and 9 is markedly superior at blocking social-engineering attacks than the Google equivalent used by Chrome, Firefox, Apple's Safari, an independent test by NSS Labs has found.
Rating the browsers against a sample set of European malware URLs over 19 days in April, IE 8 achieved a mean block rate of 90 percent, leaving Chrome 10, Firefox 4 and Safari 5 in the dust on 13 percent each. Opera, which uses technology from antivirus company AVG, came in last on 5 percent.
When assessing IE 9 with application filtering turned on, the results were even more dramatic, taking that version to a mean blocking rate of 100 percent.
Internet Explorer's positive showing appears to be thanks to two embedded technologies; Smartscreen URL Filter, a cloud-based system that checks URLs against a master database. This is present in both IE 8 and 9 and seems to work more or less identically in both.
In addition, IE 9 has added a second system, SmartScreen Application Reputation which on the basis of this test offers browser users a remarkably effective level of download block protection. Chrome, Firefox and Safari all use a rival URL checking system, Google's Safe Browser Feed, which as previous NSS Labs tests have suggested, is now falling some way behind.
"The significance of Microsoft's new application reputation technology cannot be overstated. Application Reputation is the first attempt by any vendor to create a definitive list of every application on the Internet," the authors conclude.
"Browsers provide a layer of protection against socially-engineered malware, in addition to endpoint protection products; as this report shows, not all are created equal. The overall lower protection offered by Firefox, Safari, and Chrome is concerning."
An extra but important dimension also tested was the 'average response time to block malware', basically the time it took each browser to add a problem site to the block list once it had been fed in to the test by NSS Labs.
Again, IE 9 with Application Reputation enabled gained a perfect score, adding a site without any delay, the only browser to manage such a feat. Interestingly, however, without the Application layer, IE 8 and 9 sank down the table, taking nearly 14 and 16 hours respectively, behind Safari's five hours, Chrome's nearly seven hours, and Firefox's 8 hours.
Block time is worth paying attention to because the longer protection takes to be activated, the longer the window of possible exposure.
The limitation of the report is that it is only measuring one dimension of the threat users face when using browsers, that of attacks where the user can be tricked - 'socially-engineered' in security parlance - into downloading malware. This compares with what are called 'drive-by' attacks that seek to exploit specific vulnerabilities in software and which require no user intervention.
Which is more dangerous is a matter of debate although NSS Labs references a separate study by AVG that found socially-engineered attacks to be the most likely way for malware to find its way on to a user's PC.
A social engineering attack has the advantage that it recruits the user to agree to a download event thereby potentially bypassing Windows controls such as User Access Control (UAC) and even the warnings of antivirus software. A drive-by attack, especially one manipulating a zero-day flaw, can sneak on to the PC without any of these defences being aware but requires more engineering effort to work.
The claim that socially-engineered attacks are the more significant doesn't entirely accord with the admittedly patchy evidence that exists on the subject.
A recent and revealing assessment by Qualys using its Browsercheck tool found that large numbers of browser users routinely run out-of-date plug-ins for interfaces such as Flash Adobe Reader and especially Java. Many of these have significant flaws that can be attacked by drive-by exploits.
It could be that both sides of this coin - social-engineering attacks and drive-by attacks - are equally perilous but in different ways.
A final qualification is that the test was conducted on Firefox 4, since supplanted by the rapid-development replacement, version 5.0, likewise Google Chrome, which has reached version 13. The URL-filtering systems used by these are, however the same as in the previous versions so would be unlikely to make a difference to their blocking performance.