FRAUDAR: How A New Open Source Algorithm Aims To Kill Social Media Frauds
Short Bytes: The researchers from Carnegie Mellon University have developed a unique algorithm that can detect social media frauds with high accuracy. Named FRAUDAR, this algorithm uses a unique set of metrics that set it apart from the existing fraud detection techniques. Also, this breakthrough algorithm is scalable and open source.
Apart from using bots and other automated tools, review fraudsters also use hijacked social media and email accounts of real world users. This makes their “camouflage” really impressive. With an exponential rise in the number of online marketplaces, the number of spammers and fraudsters is also increasing. Talking in technical terms, fraudsters add many ‘edges’ to their profiles and create a ‘large’ and ‘dense’ region around them to look normal. They make their camouflage more believable by sharing some updates of famous actors, singers, or products on their social media pages.
To detect and neutralize these fake actions, researchers from Carnegie Mellon University have developed an open source algorithm that works to spot fake social media profiles and identifies their fraud actions.
The algorithm, being called FRAUDAR, is being pitched as a novel approach for successfully detecting camouflage. “We provide data-dependent limits on the maximum number of edges a group of fraudulent adversaries can have without being detected, on a wide variety of real-world graphs,” researchers wrote in their paper.
FRAUDAR algorithm is open source and scalable
In the recent years, fraud detection has received a significant amount of attention. There are methods like local clustering, dense subgraph mining, social network-based Sybil defence etc. However, spammers, even without the knowledge of these fraud detection methods, are able to mimic the user behaviour in the closest manner possible.
In their paper, researchers have drawn a comparison between FRAUDAR and other algorithms. This new algorithm starts its work by looking for accounts that it can confidentially label legitimate. Such users may follow few accounts and post only an occasional review. After it removes the legitimate accounts, the camouflage ones are remained to face the fury.
By using a novel family of metrics that “satisfy intuitive “axioms”” and advanced theorems, FRAUDAR is able to perform efficiently. It’s scalable–with near-linear time complexity in the number of edges–to satisfy the needs of different sizes of organizations.
“Furthermore, FRAUDAR offers natural extensibility and can easily incorporate more complex relations available in certain contexts such as review text, IP addresses, etc.,” according to the researchers.
Oh, and also, FRAUDAR is open source and its code is freely available.
Did you find this article interesting? Don’t forget to drop your feedback in the comments section below.
Also Read: Using Facebook At Work May Slowdown Your Nation’s Economy
