aasist

Real-time voice deepfake detection, synthesis watermarking & vishing defence — XLS-R + AASIST (2.84% EER, v9c) that also catches modern voice clones & premium TTS. Explainable, edge-ready (0.62 MB ONNX), with an IPED forensic add-on. FastAPI + React. IEEE SM2026.

Smart India Hackathon 2026 • Problem Statement 104 - AegisVoice AI: Biometric Anti-Spoofing Operations Center combining AASIST backend with React frontend for real-time deepfake audio detection

Audio deepfake detection with interactive SSL/log-Mel fusion and an AASIST-style graph attention backend

Voice-cloning fraud is a live problem in India, and there is no open, usable tool for a non-experts. So the study is to take a detector trained on clean Indic spoof data, push the test audio through real messaging-app and phone-call chains, and measure how much detection accuracy survives.

谛听 VeriCall：面向老年群体的 AI 拟声电话诈骗拦截系统｜声学伪造检测 + 家庭声纹 + 话术风险三通道融合决策

This repository contains a torch-geometric implementation of the AASIST framework for audio anti-spoofing, featuring optimized GNN modules for significantly faster training and enhanced performance.

First Class Final Year Computer Science BSc dissertation project - A Personalised Audio Deepfake Detection system

PyQt6 - Desktop GUI for spoofed speech detection