VISHGUARD - A Voice Phishing Detector

VishGuard is a real-time voice phishing (vishing) detection framework which utilizes intelligent speech analysis to identify and mitigate social engineering threats. The system records live or stored audio, converts it to text using Google's Speech-to-Text API and processes the text using a trained logistic regression classifier to determine if the interaction is safe or suspicious. A confidence threshold mechanism prevents false positives while preserving the detection rate. Implemented with Streamlit, VishGuard provides an interactive dashboard to display classification results, voice alerts, real-time analytics like word cloud visualization, detection history and all. Experimental evaluation with varying datasets produced an average overall accuracy of 92% and false-positive rate of less than 4% showing reliability of the model for different accents and background conditions. Operating entirely offline, the system protects user privacy and scales for secure communication applications. VishGuard provides an effective, lightweight mechanism and privacy-aware solution to fight against the voice-based cyberthreats in real-world scenarios.