'FakeOff' helps detect fake Facebook accounts

The app on the world's largest social networking platform claims to protects users from scams devised by fake Facebook users, who are mistaken for genuine 'new friends'.

"Recent statistics show that at least 10 per cent of about 1.35 billion Facebook users are not authentic. Besides, there are millions of users who create fake identities and appear as regular users," FakeOff creator Eliran Shachar told PTI.

Fake profiles are divided into several groups including criminals, commercial and psychological, that can risk reputations (identity theft), children (paedophiles and sexual abuse), property (home break-ins) and personal safety, he added.

"FakeOff app uses sophisticated algorithm to investigate the behaviour of suspect 'friends' and rank them according to a 1-10 credibility score. It scans up to 365 days of timeline activity for every suspect Facebook friend and checks for abnormal activity," Shachar said.

The app checks timeline activity of the suspect and tries to locate abnormal activity that indicates a non-normal way of usage. It allows the user to scan the photos of the suspect to find out if it was stolen from someplace online, he added.

Also, FakeOff crosses information from all investigations and calculates results of a user based on other investigations on the same suspect, he said.

FakeOff has been live for two months now and has over 15,000 users so far.

"Twenty-four per cent of investigations conducted in the app return as fake. A fake profile can be very complex and some of the fakes that we help the users find is only for their eyes so we can't know the final result from the photo scan results, but the user easily can," he said.

According to Facebook, about 14.3 crore accounts on the social networking site may be false or duplicate, with a major chunk of them coming from developing markets like India and Turkey.

The firm said it estimates up to 7.9 per cent accounts being duplicate and up to 2.1 per cent and up to 1.2 per cent accounts being user-misclassified and un-desirable, respectively.