Learning from all those spam emails

On average, Onlinespamfilter scans 30 emails per second, with peaks of up to 200 emails per second. And it checks each one on hundreds of characteristics. You can imagine how much data that gives us every day. We use all that data to protect you even better against attacks, even before spammers can think them up.
For most people a spam email is very annoying, but for us such an email contains a lot of valuable information. Think of the senders used, word combinations and sending techniques. With an anonymised database holding the insights from more than two million messages a day, it’s fair to say we are talking about big data here.
Spotting an attack quickly
We analyse all this data so that an attack can be spotted quickly. For example, when we see a certain file being sent where it is not yet entirely clear whether it contains a virus, the scanner ‘parks’ this email. The system immediately looks at what kind of file it is, whether it has been sent more often, and by whom. If we see, for example, that it has been sent from several domains to domains that have no relation to each other, we know we may be dealing with an attack. This way we can stop an attack quickly.
Even more analysis
To discover even more trends and stay ahead of possible attacks, we are continuously improving our system. At the moment this still involves quite a bit of manual work: we tell the system what to look out for in a suspicious situation, but we are working on a new way of analysing that needs less manual work and lets the system spot trends faster. For example, whether a particular spam network becomes active at set times. When we know this, we can anticipate a possible attack sooner.
Learning by itself
We teach the system what to look out for to stop spam attacks, but of course it is best if the system learns that by itself. We are currently working to make that possible, and in the meantime Onlinespamfilter keeps stopping 99.99 per cent of the spam emails that are sent. So you are not bothered by spam emails.
