Digital Watchdog

Inspiration

With the rise of online communication among teens, cyberbullying has become a silent epidemic. We wanted to build a tool that empowers parents to protect their children in real time — without violating their privacy or making them feel watched. Digital Watchdog gives parents peace of mind and equips them to act when their child is being targeted online.

What it does

Digital Watchdog is a Chrome extension that monitors real-time messages on platforms like Instagram, WhatsApp, Telegram, and Discord. When a message is flagged as harmful or needing attention by our AI model, the system alerts the parent via email with the timestamp, platform, and content summary. The parent can then choose how to respond.

How we built it

We built a Chrome extension that uses Selenium-based scraping to pull live messages from web apps. Messages are sent to a backend using Python, where we use a RoBERTa NLP model (via Hugging Face Transformers) to classify them as Safe, Needs Attention, or Harmful. Results are saved in a JSON/CSV format and used to trigger automated email alerts using SMTP. The project is hosted in Firebase and integrated with Google Cloud.

Challenges we ran into

We faced challenges with DOM instability across platforms — selectors for messages often changed or failed in headless mode. Running NLP classification in parallel with the Transformers model also introduced multiprocessing errors. Handling authentication manually (like WhatsApp QR codes) limited full automation, but we found workarounds for testing. Designing a system that’s powerful without being intrusive was also an ethical balancing act. The privacy rules of these apps also restricted us from using the traditional API methods.

Accomplishments that we're proud of

We’re proud that we built a fully working prototype that captures real messages from multiple platforms, classifies them in real time using AI, and alerts parents — all within a single session. It’s fast, accurate, and integrates across systems smoothly. We also maintained a focus on user trust and data security throughout the project.

What we learned

We learned how to work with live browser environments, implement NLP pipelines with real-time data, and manage asynchronous workflows in Python. We also learned to collaborate fast under pressure and break a big idea into testable modules. Most importantly, we realized that thoughtful tech design can empower families without compromising independence.

What's next for Digital Watchdog

We plan to add mobile push notifications and a full parent dashboard for reviewing flagged conversations over time. We also want to allow parents to set custom sensitivity levels and notification preferences. In the future, we plan to explore on-device classification and encryption to keep data private while preserving safety.