Inferring Smartphone Application Types Via Inaudible Charging Sound: A New Acoustic Side-Channel Attack

Inferring Smartphone Application Types Via Inaudible Charging Sound: A New Acoustic Side-Channel Attack IEEE Franklin V. Taylor Memorial Award - Nominees in IEEE SMC 2025, Vienna. Dataset The dataset contains 36 audio recordings captured during smartphone charging sessions across different application scenarios. Recording Specifications - **Sampling Rate**: 192 kHz (high-fidelity acoustic capture)- **Recording**: Continuous full-session audio (no pre-sliced segments; slicing done during preprocessing)- **Format**: WAV (uncompressed)- **Microphone**: Double X0 pickup microphone- **Battery Level**: All recordings at 100% capacity- **Environment**: Controlled ambient conditions for reproducible acoustic feature extraction Device Specifications Smartphones: - **Redmi Note 12T**: Mid-range Android device- **Huawei Nova 6**: Legacy device for cross-generational analysis Chargers: - **Xiaomi 67W Charger**: Fast charging (QC/PD protocols)- **Ugreen Dual-Port 20W**: Baseline charging (17W single-port mode) Application Categories | Category | Applications | Protocol || ----------------------- | -------------------------------------------------- | ------------------------------------- || **Games** | Minecraft, Sky: Children of the Light, PUBG Mobile | 3D scene navigation and gameplay || **Social** | QQ, DingTalk | Text messaging and transmission || **Entertainment** | Douyin (TikTok), Kuaishou (Kwai) | Video playback with content switching || **Shopping** | JD, Pinduoduo | Product browsing and selection | File Structure ``` dataset/├── 01.wav ~ 36.wav # Audio recordings└── labels.json # Metadata with app names and device info``` Each entry in `labels.json` contains: ```json{ "app_name": "Application name", "device": { "phone": "Smartphone model", "charger": "Charger specification" }}```