DecodeX V2: Real-Time Neural Edge Firewall
This is the core neural engine for the DecodeX audio verification platform, optimized for high-speed edge inference and deepfake voice detection. It achieves the following results on the evaluation set:
- Loss: 0.0141
- Accuracy: 0.9973
Model description
DecodeX V2 is a highly optimized audio classification model designed to detect synthetic, spoofed, and AI-generated voices in real-time. It is built to operate within a dual-model cross-verification architecture, ingesting raw PCM audio streams normalized via Z-Score mathematical optimization to output high-confidence threat vectors.
Intended uses & limitations
Intended for enterprise PBX, SIP trunk integrations, and local edge hardware as a real-time firewall against voice spoofing and deepfake social engineering attacks.
Training and evaluation data
Aggregated high-fidelity authentic and synthetic voice samples, resampled and normalized at 16 kHz.
Training procedure
Training hyperparameters
The following hyperparameters were used during training:
- learning_rate: 3e-05
- train_batch_size: 32
- eval_batch_size: 32
- seed: 42
- gradient_accumulation_steps: 4
- total_train_batch_size: 128
- optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
- lr_scheduler_type: cosine
- lr_scheduler_warmup_ratio: 0.1
- num_epochs: 5
Training results
| Training Loss | Epoch | Step | Validation Loss | Accuracy |
|---|---|---|---|---|
| 0.0503 | 1.0 | 1381 | 0.0514 | 0.9858 |
| 0.0327 | 2.0 | 2762 | 0.0174 | 0.9956 |
| 0.0064 | 3.0 | 4143 | 0.0221 | 0.9950 |
| 0.0003 | 4.0 | 5524 | 0.0174 | 0.9965 |
| 0.0115 | 5.0 | 6905 | 0.0141 | 0.9973 |
Framework versions
- Transformers 4.41.2
- Pytorch 2.1.2
- Datasets 2.19.2
- Tokenizers 0.19.1