A spear phishing tool to automate the creation of phony tweets - complete with malicious URLs – with messages victims are likely to click on will be released at Black Hat by researchers from ZeroFOX.
Called SNAP_R (for social network automated phisher with reconnaissance), the tool runs through a target Twitter account to gather data on what topics seem to interest the subscriber. Then it writes a tweet loaded up with a link to a site containing malware and sends it.
The researchers – John Seymour and Philip Tully – say SNAP_R lets them scale up phishing tries on Twitter accounts. It takes five to 10 minutes to write a single spear phishing email, for example, but it takes a matter of seconds or minutes to generate thousands of spear phishing tweets, depending on how much hardware they throw at the problem.
The manual phishing has a 40% to 45% click-through rate, while their automated method garners about 33%. But because of the speed with which the tweets are generated, the net return is much greater. “It’s slightly less effective but it’s dramatically more efficient,” says Seymour.
Twitter accounts are a good place to try spear phishing because of the combination of language used, APIs to rich data and use of shortened links. Because tweets and short and informal, the language doesn’t have to be perfect and messages are so short that victims might forgive mistakes that might tip them off if they occurred in an email, Tully says.
Twitter APIs let the tool auto-post as well as collect significant data about the victims so it’s easier to write tempting tweets. And the shortened links mask the actual URLs, which might raise red flags about the authenticity of the tweets, he says.
SNAP_R triages users by checking how active their accounts are and seeking clues about what they do for a living. Inactive accounts indicate someone with little prospect for even finding the tweets. Career information can indicate whether the person is the right type of person to receive the tweet and also to determine what subject matter might entice them, Seymour says.
The tweets themselves can be fashioned using a Markov model that ties it to a timeframe. So if a person is interested in the Rio Olympics, it wouldn’t pay to write a tweet about them in December, but they might be an excellent topic for July.
They can also be fashioned using a neural network model that is trained how to compose tweets by having it digest millions of tweets beforehand. All it needs is to find a topic to tweet about. It can also write in any language. The goal is for the tweets to be indistinguishable from tweets written by a person.
Sign up for CIO Asia eNewsletters.