🛡️ AudioShield AI

Real-Time Audio Deepfake & Scam Detection

Don't just detect a fake voice. Detect the threat behind the voice.

AudioShield AI is an AI-powered voice security platform that analyzes audio conversations to detect deepfake indicators, scam patterns, impersonation, social engineering, and attacker intent.

---

🏆 Hack Fusion 2026

Developed for Hack Fusion 2026, conducted by Chennai Institute of Technology.

👥 Team Tech Titans

• Team Leader: Rithika B
• Team Members: Rithika G, RISHIRAM M, ROHESH R

---

🚨 Problem

AI-generated voices and voice cloning are making scams increasingly difficult to identify. A dangerous call may involve multiple signals such as impersonation, urgency, fear, OTP requests, financial manipulation, and credential harvesting.

Traditional systems often ask:

"Is the voice fake?"

AudioShield AI asks:

"Is the interaction dangerous?"

---

💡 Solution

AudioShield AI combines multiple threat signals to provide an explainable security assessment.

Audio Input
     ↓
Speech & Conversation Analysis
     ↓
Voice + Scam + Social Engineering Detection
     ↓
Attacker Intent Analysis
     ↓
Threat Assessment
     ↓
Threat Score + Explainable Results

---

✨ Key Features

• 🎙️ Audio Analysis — Analyze live or uploaded audio
• 🧬 Deepfake Indicators — Identify potential synthetic/manipulated voice signals
• 🚨 Scam Detection — Detect banking, financial, OTP and impersonation scams
• 🎭 Social Engineering Detection — Identify urgency, fear, authority and manipulation
• 🎯 Attacker Intent Detection — Identify possible objectives such as credential theft or financial fraud
• 📊 Threat Score — Generate an overall risk assessment
• 🔍 Explainable Detection — Show the evidence behind the threat
• ⏱️ Threat Timeline — Visualize suspicious events throughout the conversation
• 🔗 Attack Chain — Show how a scam progresses
• 🧠 AI Security Copilot — Ask questions about the detected threat
• 📋 Threat Intelligence Report — Summarize findings and safety recommendations

---

🏗️ System Architecture

       AUDIO INPUT
            ↓
    AUDIO / SPEECH ANALYSIS
            ↓
   CONVERSATION ANALYSIS
            ↓
 ┌──────────┼───────────┐
 ↓          ↓           ↓
VOICE      SCAM      SOCIAL
SIGNALS   PATTERNS   ENGINEERING
 └──────────┼───────────┘
            ↓
     ATTACKER INTENT
            ↓
     THREAT ASSESSMENT
            ↓
       THREAT SCORE
            ↓
   EXPLAINABLE RESULTS

---

🛠️ Tech Stack

Frontend

• React
• TypeScript
• Vite
• HTML5
• CSS

AI

• Google Gemini API
• Generative AI
• Contextual AI Analysis

Audio

• Browser Audio APIs
• Audio Visualization

Development

• Google AI Studio
• Visual Studio Code
• Git & GitHub

---

📁 Project Structure

AudioShield-AI/
├── src/
├── public/
├── index.html
├── package.json
├── server.ts
├── vite.config.ts
├── .env.example
├── .gitignore
└── README.md

---

⚙️ Run Locally

Prerequisites

• Node.js
• npm
• Git
• Gemini API Key

1. Clone the Repository

git clone https://github.com/Rithika-B-7/AudioShield-AI.git
cd AudioShield-AI

2. Install Dependencies

npm install

3. Configure API Key

Create a .env.local file in the project root:

GEMINI_API_KEY=your_api_key_here

⚠️ Never commit your actual API key to GitHub.

4. Start the Application

npm run dev

Open the local URL displayed in the terminal.

---

🎯 Use Cases

• 🏦 Banking & financial fraud detection
• 📞 Customer-support impersonation
• 👨‍👩‍👧 Personal voice-call safety
• 🔐 OTP and credential protection
• 🏢 Enterprise social-engineering defense

---

🚀 Future Scope

• Real-time phone-call integration
• Advanced voice-cloning detection
• Multilingual scam detection
• Mobile application
• Speaker verification
• Automatic scam-call blocking
• Privacy-preserving on-device analysis
• Enterprise threat intelligence

---

🛡️ AudioShield AI

Hear the Voice. Understand the Intent. Detect the Threat.

Team Tech Titans | Hack Fusion 2026

Chennai Institute of Technology