Inspiration
As generative AI tools became widely used, we noticed a major gap: privacy and data security were often overlooked. Many AI tools process sensitive data in ways that users cannot fully control, risking exposure of personal or confidential information. We were inspired to build a solution that empowers users to safely use AI without compromising their privacy. Our goal was to create a system that could operate entirely on-device, giving users full control over their data.
What it does
Privacy Guard AI is an on-device privacy layer for AI applications. It: Automatically anonymizes sensitive user inputs (text, documents, images). Monitors AI outputs for potential privacy leaks. Alerts users if content may contain sensitive information. Supports custom privacy policies and offline operation, ensuring that no raw data leaves the device.
How we built it
We used a modular architecture: Privacy Filter Module - detects and masks sensitive data. On-device AI Engine - processes filtered inputs locally. Privacy Monitor - scans outputs and logs anonymized metadata. The prototype was developed using Python, with integration for text and file inputs, and a lightweight interface for user interaction.
Challenges we ran into
Ensuring real-time privacy monitoring without slowing AI responses. Designing on-device AI processing that works efficiently on modest hardware. Handling multi-modal inputs (text, images) while maintaining strict privacy.
Accomplishments that we're proud of
Built a fully functional on-device prototype. Implemented real-time privacy alerts and anonymization. Achieved a balance between AI functionality and user privacy.
What we learned
On-device processing is crucial for data security. Privacy-aware AI design requires layered approaches: filtering, monitoring, and anonymization. Users value transparency and control over their data.
What's next for Privacy Guard AI
Support for audio and video inputs. Advanced threat detection algorithms. Integration with popular AI platforms. Providing privacy scoring for AI-generated content to help users gauge safety.
Log in or sign up for Devpost to join the conversation.