Phishing is a type of cyber-attack that targets individuals through email, text messages, or other forms of communication, with the intent of tricking the recipient into disclosing sensitive information.
There is a clear need for a tool that can objectively assess the risk that an email is a phishing attempt. The purpose of this project will be to create such a tool: a deep learning model capable of accurately evaluating emails and identifying phishing attacks. Deep learning is a suitable approach for this task, as phishing detection is fundamentally a natural language processing problem.
- Since 2021, phishing attacks have increased by 49%, with 64% of businesses reporting business email compromise (BEC) attacks in 2024
- According to the Verizon 2024 Data Breach Investigations Report, only 20% of employees successfully identify and report phishing attempts during training
Phishing identification in the context of neural networks is done by text classification. As a result, we plan to implement a neural network model designed for natural language processing (NLP). We will use a lightweight transformer model to analyze the content of email messages and identify potential phishing threats.