Is That Call Normal? Is it Even a Human Being?
There are two general ways to detect fraudulent calls: CDR pattern matching and media-stream analysis.
I'm on the train from New York to Boston following two days of meetings with customers whose contact centers have recently been attacked by SIP-generated denial-of-service attacks. We discussed ways to protect their networks against future attacks of this kind. We decided that the first layer of protection should be a session border controller that can protect against packet attacks and can perform enough deep-packet inspection to look for malformed SIP headers. For these particular customers, however, that was not enough. As I explained in my previous post (If it Could Happen to Them...), the inbound calls are “normal” SIP calls.
So how can you detect these fraudulent calls? Well, there are two general ways to do this: CDR pattern matching and media-stream analysis.
CDR pattern matching consists of analyzing the historical call-pattern variables (duration, time of day, caller ID, called number, etc.) to determine a profile that is "baseline normal." As calls come in that do not match the profile based on any of these variables, they can be flagged and appropriate action taken (more on that in a bit).
Media-stream analysis looks at the actual media coming into your voice network; for SIP calls it looks at the RTP stream. It analyzes the RTP stream to make sure it contains a human voice as opposed to a music file or other type of computer-generated media. Again, once a call is flagged as "not human," action can be taken on it. The key thing here is to be able to set threshold values for what "non-human" means. You want to make sure that all the calls you identify are truly non-human.
This brings me to the third capability that must be present--the ability to take action once an intrusion is identified. The action can terminate the call or perhaps reroute the call to specially trained agents. So there has to be some sort of integration with your dial-plan manager. You put all of these capabilities together and you have a voice intrusion prevention system--a device that builds on the "firewall" functionality of a typical SBC to extend protection to the voice application layer.
So the enterprises I mentioned earlier are exploring products that combine this functionality in order to enhance their voice security. As we all know, sometimes it takes an actual security incident to acknowledge that it can happen to anyone--including you. But there are products on the market today that can help prevent these types of attacks.