The problem PhishSight solves

Phishing is a massive problem globally, affecting millions of people every year. According to recent statistics, phishing attacks have increased by over 50% in the past year alone. In fact, it's estimated that nearly 1 in every 100 emails sent is a phishing attempt. This means that billions of phishing emails are sent out annually, targeting individuals, businesses, and organizations alike. Moreover, phishing attacks are responsible for over 80% of reported security incidents, making it the most common form of cyber threat

This predictive model helps people stay safer online by quickly spotting fake websites trying to steal their information, like passwords or credit card numbers. Imagine it like a super-smart guard that watches out for you while you're browsing the web. It makes things easier because instead of worrying about whether a website is trustworthy or not, the model does the hard work for you. So, you can surf the web with more peace of mind, knowing that this guard is on duty, keeping you safe from online scams and frauds.

The future scope of the model entails enhancements in machine learning techniques, multi-language support, and cross-platform integration to bolster its efficacy in detecting phishing attacks. Integrating the model into a user-friendly website interface would allow individuals to effortlessly check the legitimacy of URLs, providing clear feedback on the phishing risk associated with each entered URL. Additionally, offering an API for developers and crowdsourced feedback mechanisms would further refine the model's performance over time, while educational resources alongside the URL checking feature would empower users to better protect themselves against online threats.

Challenges we ran into

Attribute Meaning and Importance: Phishing datasets often contain numerous attributes, each contributing differently to the detection process. Understanding the meaning and significance of these attributes is crucial for developing effective detection models. Attributes such as URL length, domain age, and HTTPS presence carry varying levels of importance in distinguishing legitimate websites from phishing ones.

Noise and Outliers: Noise and outliers in the dataset can significantly impact the performance of detection models. Noise refers to irrelevant or erroneous data, while outliers are data points that deviate significantly from the rest of the dataset. Identifying and mitigating noise and outliers is essential to ensure the robustness and reliability of the detection system.

Dealing with Noise and Outliers: Various techniques can be employed to handle noise and outliers in the dataset. This includes data preprocessing methods such as normalization, standardization, and outlier detection algorithms and using Z Score.

Choosing Suitable Algorithms and Parameter Tuning : Several machine learning algorithms can be applied to phishing detection, including decision trees, random forests, support vector machines (SVM), and neural networks. The choice of algorithm depends on factors such as dataset size, complexity, and interpretability requirements. For instance, decision trees are intuitive and easy to interpret, while neural networks excel at capturing complex patterns.

Evaluation Metrics: Selecting appropriate evaluation metrics is essential for assessing the performance of detection models. Common metrics include accuracy, precision, recall, F1-score, and area under the receiver operating characteristic curve (AUC-ROC). These metrics provide insights into the model's ability to correctly classify phishing and legitimate websites while minimizing false positives and false negatives.