Phishing URL and Text Detection

Abstract: Phishing attacks are going up fast and they are a big problem when we are online. People who do these attacks use links to websites and fake messages to get important information from us, like our passwords and bank details and the special codes we get to confirm who we are. A lot of the systems we have can only find fake website links or fake messages not both so they do not work as well as they should. Phishing attacks use ways to trick us like fake website links and phishing messages to get our sensitive information. Previous studies used machine learning and deep learning techniques like Random Forest, Logistic Regression, Naïve Bayes, CNN, LSTM and BERT. These machine learning techniques were mainly used for detecting things in URLs. Text-based methods were also used, such as TF-IDF, Bag of Words and keyword analysis. The thing is, there is no simple system that can detect both URLs and messages at the same time, which is what machine learning and deep learning techniques, like Random Forest, Logistic Regression, Naïve Bayes, CNN, LSTM and BERT are supposed to do for URLs and messages. It uses machine learning and natural language processing to do this. The system looks at things like how long a website address how many dots it has and if it has special characters. It also checks if the website address uses HTTPS and if it is an IP address. The system uses tools like Logistic Regression and Random Forest to analyze all these things. It uses natural language processing to get the messages ready. Then it uses special tools like TF-IDF and Naïve Bayes with Logistic Regression. This helps the system figure out if the messages are trying to trick people into doing something or if they are regular messages. The phishing detection system is made to catch messages that try to scare people into doing something. The system is, about catching phishing and it uses machine learning and natural language processing to make sure it works well. The combined system improves detection accuracy, achieving 94% accuracy for phishing URLs and 92% accuracy for phishing text. The system can be extended for real-time use in browsers, emails, and SMS to protect users from online fraud.

Keywords: Phishing Detection, Phishing URL Detection, Phishing Text Detection, Machine Learning, Natural Language Processing (NLP), Cyber Security, Online Fraud Detection, Feature Extraction, Classification Models.