SafeClick
SafeClick is an AI-powered threat detection platform and browser extension that helps users assess potential security threats through URLs or uploaded screenshots. Its detection system combines a weighted ensemble of security signals, including a machine learning model trained on 60+ phishing features, to evaluate risk and generate clear, plain-language threat reports that explain what was detected and why it may be suspicious.
Why I Created It
SafeClick began as an idea developed for the Hofstra × Pensar Hackathon, where we built the first MVP around making digital threat detection more accessible and understandable for everyday users. After validating the concept through the competition, we continued developing and shipping the product beyond the hackathon, eventually advancing it through the New York Business Plan Competition (NYBPC). Today, we are continuing to evolve SafeClick toward a market-ready cybersecurity product.
Toggle caption
How I Built It
SafeClick uses a layered threat-detection pipeline rather than relying on a single classifier. URLs are evaluated across multiple signals, including domain and registration information, URL-pattern analysis, Google Safe Browsing checks, reputation and blacklist signals, SSL and security metadata. For screenshot-based assessments, OCR extracts visible text and URLs before they are passed through the same security checks, with LLM analysis used to synthesize the collected evidence into a structured, plain-language threat report. A weighted ensemble ranking system combines the signals to produce the final risk assessment.
Toggle caption
Impact
SafeClick won 1st Place and the “Most Secure” award at the Hofstra × Pensar Hackathon, then went on to win 1st Place in the regional round of the 2026 New York Business Plan Competition. Since its initial MVP, the project has grown into a deployed web platform and browser extension, with its machine learning detection model reaching approximately 92% accuracy during evaluation.
Toggle caption
