ExtraTorrent | Microsoft Experts Cut Down Skype Spam

[Guest]

Microsoft developers are trying to find a way to cut down on fake Skype accounts and efficiently kill off spammers in the network. They investigated sources like user’s profile, activities and social connections via a so-called supervised machine-learning environment. The latter could automate manual tasks of fraud detection.

Thus far, the results indicated that the framework boosted fraud detection rates for certain account types by almost 70% with a 5% false positive rate. Microsoft reported that there are some types of fraud relevant to Skype. They normally include credit card and other online payment fraud, along with account abuse like spam instant messages.

The idea of Microsoft’s move is to detect the fraudsters who elude the first line of defenses at Skype. They conducted a research across 34,000 Skype members. This number included both legitimate and fraudulent accounts chosen from an initial randomized pool of 200,000 Skype users that hadn’t been blocked for more than 4 months after setting an account.

The research collected and analyzed account habits of the captured Skype members limited to what kinds of communications methods they used and how often. In the meantime, they didn’t record the content of calls, and anonymized Skype usernames. As a result, the researchers managed to identify fraudulent accounts which were inactive after 4 months. However, it turned out hard to find if they remained active for over 10 months after registering an account. By the way, it would appear that major part of the fraudulent accounts operating on Skype were hacked legitimate users.