Overview / Usage

The aim of this PoC-based project is to give an idea of how modern phishing website attacks can be prevented using machine learning. To do this, we are going to use the Phishing Websites' Dataset (https://archive.ics.uci.edu/ml/datasets/phishing+websites). The viewers are requested to take a look at this paper (https://archive.ics.uci.edu/ml/machine-learning-databases/00327/Phishing%20Websites%20Features.docx) by the authors of the dataset. The paper discusses the data generation strategy in details and how the authors were able to come up with the most significant set of features for detecting phishing websites.

We can employ machine learning and other data science techniques to detect phishing websites. As a next step, you can wrap the final model as a REST API endpoint and use it along with a browser add-on to detect phishing websites in real time.

Methodology / Approach

• The dataset was is .arff file format which is not compatible with data analysis and manipulation libraries like Pandas, numpy etc. So converted it to .csv for easier analysis and use.
• Thoroughly investigated the dataset to find out what models might be better for prediction purpose.
• Incorporated random search for tuning the hyperparameter of the initial model.
• Incorporated neural networks to improve performance.
• Reproduced a research paper to further push the performance.

Technologies Used

• Python (Intel)
• scikit-learn, pandas, numpy, keras (Libraries)
• Jupyter Notebook
• Google Cloud

Repository

https://github.com/sayakpaul/Phishing-Websites-Detection