Inspiration
As digital connectivity expands through 5G, public WiFi, and IoT, users face invisible privacy risks every day—especially in regions with limited digital literacy or regulatory protection. We were inspired to build a system that empowers anyone, regardless of technical background, to understand and defend their data in real time. With the rise of AI, we saw an opportunity to not just detect threats, but explain them—making privacy actionable, accessible, and equitable.
What it does
AI-powered privacy firewall that runs across 5G, WiFi, and IoT networks. It monitors real-time network traffic, detects anomalies, calculates privacy risks, and uses Claude AI to generate human-readable explanations and recommendations. It operates via a mobile or desktop app, with support for telecom and WiFi edge integration, and uses AWS Bedrock, SageMaker, and edge containers for scalable, low-latency protection.
How we built it
We designed a layered architecture with:
Mobile App (React Native): For real-time risk feedback and privacy control.
Edge Computing (5G/WiFi): MEC and NFV-based traffic analysis with Kubernetes.
AWS Cloud Intelligence: Claude 3 via Bedrock for conversational privacy insights and SageMaker for threat prediction models.
Kafka + Redis + Prometheus: For stream processing, caching, and monitoring.
Terraform + GitHub Actions: For infrastructure-as-code and CI/CD automation.
We also integrated zero-knowledge proof generation and blockchain-based privacy reputation scoring to future-proof trust.
Challenges we ran into
Privacy vs. Performance: Ensuring real-time detection under 100ms while preserving user data privacy was a major architectural challenge.
Claude prompt tuning: Making explanations non-technical, actionable, and emotionally resonant for users required iterative fine-tuning.
Edge deployment complexity: Getting privacy functions to work efficiently across multi-access edge environments (especially 5G network slices) took significant optimization.
Cross-device orchestration: Coordinating traffic from mobile, edge, and cloud components in a seamless UX was technically intensive.
Accomplishments that we're proud of
Built a zero-knowledge proof system that verifies privacy compliance without exposing user data.
Deployed a full-stack edge-cloud hybrid privacy engine that runs across mobile, WiFi, and telecom.
Integrated Claude AI to explain privacy threats in plain English—making compliance human.
Created a blockchain-based privacy reputation system to track malicious networks via user reports.
Achieved <100ms threat detection latency with 95%+ classification accuracy on test data.
What we learned
Users don’t just want to be protected—they want to understand their risks.
Combining AI explanations with real-time security gives users both trust and control.
Edge-first architecture is essential for future network privacy, especially in latency-sensitive environments like 5G.
Interoperability between cloud, edge, and user devices is a non-trivial problem, but a powerful differentiator when done right.
What's next for Privacy Guardian AI
Launch a public beta with telecom partners and educational institutions in underserved regions.
Expand language and cultural context support for global users.
Partner with router and hotspot vendors to integrate privacy by default.
Evolve into a decentralized privacy mesh network, where each user contributes to community protection via AI and zero-knowledge proofs.
Open-source core modules to drive broader adoption and trust in privacy-first infrastructure.
Log in or sign up for Devpost to join the conversation.