Researchers Sharpen Spear-Phishing With New Tool Leveraging Social Networks
A new tool mixes data mining with natural language processing to help pen testers create more attractive spear-phishing messages
Phishing hooks more than its share of people and organizations. But just like its homophonic counterpart, phishing can always be made easier with the right bait.
At the upcoming Black Hat Regional Summit in Brazil, Trustwave researchers Joaquim Espinhara and Ulisses Albuquerque plan to do exactly that. Using a new tool they call 'µphisher' (read as microphisher), the researchers say they have found a way to gather the digital breadcrumbs users leave on the Internet through social networks, mailing lists, online forums, and beyond.
More Security Insights
- Forrester Study: The Total Economic Impact of VMware View
- Securing Executives and Highly Sensitive Documents of Corporations Globally
- Top Big Data Security Tips and Ultimate Protection for Enterprise Data
- Client Windows Migration: Expert Tips for Application Readiness
With a mix of data mining and natural language processing [NLP], the tool can find patterns in the way a target communicates online and about what, so the information can be used to craft a more enticing attack.
"µphisher builds a database of social network status updates and makes these available for building user profiles," Albuquerque says.
Those profiles, he explains, focus on text provided by a target of interest and allow pen testers to build support data structures for the most commonly used words, as well as the people the target most frequently interacts with on social networks, hashtags, and gelocation information. With that in hand, the tool uses the information to rank how close phony content is to legitimate content produced by the target.
"We check sentence length, if the words are typically used by the target, and if the referenced users and hashtags match those actually used by [them]," Albuquerque says.
"Since different social media networks are used for different purposes ... all [social] networks are possible targets," he says. "Professional content, geolocation, pictures and movies, interacting with friends -- every one of these activities involves a different 'online persona' by the user, and the phrasing, words, and sentence length will vary wildly between content written for each of these purposes. So we don't focus on one particular social network because that would mean focusing on content which might not look legitimate on other social networks."
The tool does not try to interpret the meaning of what the user is talking about; therefore, slang, abbreviations, and other "non-standard" words would end up in its dictionary even though the natural language processing engine might not be able to categorize them properly.
"Since the tool was developed to support quick engagements, we do not want to have the consultant/penetration tester spending too much time trying to analyze and infer intention on the subject of interest," the researcher says. "We just want to help produce content that looks like it was written by the target. Thus, anything which is not proper English will be treated as noise, but will end up in our dictionaries,and will be still checked against when evaluating user-provided content."
The tool uses the official APIs for obtaining data, and in their talk the researchers plan to touch on potential legal implications of using the tool. According to Albuquerque, the user must generate the required tokens with each social network, and the tool itself does not try to be stealthy in its activities. For that reason, it may be subject to restrictions by some social networks.
"We also authenticate against the networks using the actual user identity of the person operating the tool when fetching data -- which should be enough to transfer most of the liability to them when using the tool for not-so-legitimate scenarios," he says. "We certainly do not wish it to be used as an umbrella to hide malicious users against an application-wide identity in order to harvest data from unknowing targets."
The researchers' presentation is scheduled for Nov. 26 at the summit, which will be held at the Transamerica Expo Center in Sao Paulo.
Have a comment on this story? Please click "Add Your Comment" below. If you'd like to contact Dark Reading's editors directly, send us a message.