Inspiration In response to the growing prevalence of deceptive online practices, I was motivated to develop a solution that empowers users to navigate the digital landscape securely. The increasing sophistication of dark patterns used by websites to manipulate user behavior highlighted the urgent need for an effective detection and mitigation tool. This project aims to address that need by providing a robust browser extension capable of identifying and classifying various dark patterns.
What I Learned Building this project was a comprehensive learning experience. I delved deep into the world of dark patterns, understanding their intricacies and impact on user experience. I gained hands-on experience with machine learning, specifically using the Naive Bayes model, and learned how to train it on a large dataset. Additionally, integrating the model with Gemini AI provided valuable insights into enhancing detection accuracy. The project also taught me about real-time user interaction through SMS notifications, requiring an understanding of APIs and seamless integration.
How I Built the Project The project was built using a combination of technologies and tools:
Data Collection: Sourced a comprehensive dataset from KG Mathur, containing information from over 11,000 websites. Machine Learning Model: Trained a Naive Bayes model to identify and classify 12 distinct types of dark patterns. Integration with Gemini AI: Enhanced detection capabilities by integrating the model with Gemini AI, providing nuanced insights. Browser Extension: Developed a user-friendly browser extension that allows users to detect dark patterns with a single click. Real-time Notifications: Implemented a feature where users receive instant SMS notifications via Twilio API to assess the legitimacy of a website. Challenges Faced The journey was not without challenges:
Data Quality and Preprocessing: Ensuring the dataset was clean and representative was a significant challenge. Handling inconsistencies and noise in the data required meticulous preprocessing. Model Accuracy: Achieving high detection accuracy for all 12 types of dark patterns demanded extensive tuning and validation of the machine learning model. API Integration: Integrating Twilio API for real-time SMS notifications required careful handling of API keys and ensuring secure and reliable communication. User Experience: Designing a seamless user experience for the browser extension while maintaining robust functionality was a delicate balance.
Log in or sign up for Devpost to join the conversation.