Scams present a dangerous and increasingly common threat in todays world. They are typically sent to convince victims to willingly give up their personal details and their money. The loss is enormous; therefore, more efficient solutions are necessary. Our goal is to develop a system that uses machine learning to filter out scams more effectively than spam filters do. Moreover, we wanted to give feedback to the user about the scam itself, by which the user can learn to protect himself from these types of attacks. First, we collected data and learned about scams in general. Next, we developed the detection algorithm based on semantic analysis for our tool. For that, we studied word embedding and machine learning algorithms. We developed a method using these technologies that can successfully detect scam emails. We also implemented category recognition into our tool, which now can differentiate between romance and investment scams. Lastly, we built our tool into an existing spam filter to present a real-life application for extra support and compatibility with mail servers.
Dévai Henrietta
2022-01-15
Tmogat: QUADRON Kibervdelmi Kft.