PhishGuard AI

Your first line of defense against digital deception.

Comment

Inspiration

PhishGuard AI was born from personal experience. After being targeted by both an inheritance scam and a tender scam, I realized how easy it is for even informed individuals to fall victim to sophisticated digital deception. These experiences motivated me to create a tool that can proactively protect people from similar situations.

I had attempted to build this solution before, but privacy concerns were a major barrier — all data had to be sent to the cloud for analysis. With the arrival of Gemini Nano, that limitation is finally resolved. Now, phishing detection can run locally on the user’s device, ensuring that sensitive information never leaves the browser unless the user explicitly consents.

Only specific, user-approved data is sent to the cloud (via Gemini Flash 2.5) for deeper or more complex analysis. PhishGuard AI represents the next step in privacy-first AI — built from real-world experience, powered by responsible technology.

What it does

PhishGuard AI detects scams in real time, including: Fake login pages and spoofed domains Suspicious URLs and redirects Deceptive messages on SMS or social media When a threat is found, users get immediate alerts and clear explanations. Cloud processing is optional and only with explicit consent — keeping privacy first.

How we built it

Local AI Detection: Gemini Nano analyzes text, links, and webpage content directly on the device. Consent-Based Cloud Analysis: Non-sensitive snippets are sent for advanced reasoning only if the user opts in. Chrome Extension + Dashboard: Live scanning through the extension; manual testing and uploads via Next.js dashboard. Privacy-First Architecture: No automatic uploads or tracking — everything stays local unless explicitly shared.

Challenges & Wins

Making Gemini Nano work locally in Chrome required careful flag configuration and model management. Balanced speed and accuracy while keeping scans lightweight. Designed prompts to detect scams effectively without false positives. Achievements: Real-time, offline phishing detection with 82% early accuracy. Fully privacy-respecting, user-friendly interface. Seamless integration of AI, design, and security in one browsing experience.

Accomplishments that we're proud of

Successfully integrated Gemini Nano for real-time, offline phishing detection. Built a fully privacy-respecting system — user data never leaves the device without consent. Designed a clean, accessible UI that makes cybersecurity approachable. Achieved strong detection performance with an 82% accuracy rate in early tests. Combined AI, design, and privacy into a single seamless browsing experience.

What we learned

Learned how to utilize on-device AI for real-time phishing and scam detection. Gained deep insight into Chrome’s AI architecture and Gemini Nano integration. Understood the importance of ethical AI design — balancing user safety, privacy, and transparency. Discovered how real-world motivation can drive meaningful, socially impactful AI projects.

What's next for PhishGuard AI

We’re expanding PhishGuard to detect fraud in full documents and images, enabling users to verify contracts, letters, and certificates for manipulation. A Chrome extension will also be added in the future to provide live scanning and instant alerts directly in the browser, making protection seamless.

PhishGuard AI = real-world inspired, privacy-first, and smarter online protection.

Built With

Share this project:

Updates