Virus Bulletin lauds first anti-spam tests

Virus Bulletin is to offer a new set of tests to rate the effectiveness of anti-spam products.

Respected malware testing organisation, Virus Bulletin (VB), is to offer a new set of tests to rate the effectiveness of anti-spam products.

The UK-based outfit is known for its VB100 certification, awarded to products that can detect 100 percent of the virus samples from the independently-compiled 'WildList', defined on a rolling basis by the WildList Organisation.

As Virus Bulletin admits, however, setting up similar tests for anti-spam products, was always going to be much more difficult.

The methodology the organisation has come up with is complex. All the products tested will have to receive exactly the same message stream at the same time, in real time, and the messages received will also have to be real rather than simulated examples. Because such a setup can only hope to approximate the Internet spam load at any one period in time, the volume of email will have to be high enough to make the test realistic.

Another hurdle. Anti-spam systems typically block some of the messages they receive at SMTP level based on a number of obvious characteristics such as IP address, meaning that these messages will never get as far as the inbox.

Because testers will need to subjectively define whether a message blocked at inbox level was correctly identified or blocked in error, it will be impossible for them to know whether the filter was right or not.

VB claims it has found a way round these and other technical problems that would otherwise skew the results. The end result will be tests that can, VB says, measure spam-catching rates, false positives, and in future possibly also assess performance too, though it has to decide precisely what a product will have to achieve in order to be 'certified'.

According to VB, commercial anti-spam products will be able to undertake the as-yet unnamed tests for a fee, while free-to-use and open source software will be tested gratis. Preliminary results are expected in February or March. The test spam itself will be gathered using special 'spam traps'.

"We feel strongly that there is a need for a robust and comprehensive anti-spam certification scheme and that our background stands us in good stead to run such a scheme. We are looking forward to publishing the first set of results," said Helen Martin, Editor of Virus Bulletin.

Getting access to the test results will mean having a subscription to Virus Bulletin magazine, although most users will probably enounter the test sticker in the marketing materials of vendors that have met the test criteria.

Tricky as they have no doubt been to develop, the new tests are good news for VB, troubled as it has been over doubts about the meaningfulness of its famous VB100 tests. Critics have for some time complained that despite the integrity of VB, the WildList from which the tests are drawn is no longer a sensible measure of the malware being encountered by users.