鉴定(生物学)
计算机科学
计算机安全
人工智能
脆弱性(计算)
欺骗攻击
钥匙(锁)
数据挖掘
风险评估
噪音(视频)
作者
Lakkakula Ashwitha,Divya Prakash,P. Gowtham,C. R. Dhivyaa
出处
期刊:
[Institute of Electrical and Electronics Engineers]
日期:2026-03-05
卷期号:: 605-610
标识
DOI:10.1109/nqcomp68334.2026.11497676
摘要
The proliferation of artificial intelligence has enabled sophisticated biometric attacks including deepfake generation and facial spoofing, posing critical threats to digital identity verification systems. Ensemble is introduced to robustly detect both deepfake videos and facial spoofing attacks in real-time. The model includes MobileNet for antispoofing and ResNeXt for deepfake detection. In contrast to conventional integration-layer techniques, the ensemble methodology leverages the complementary feature extraction capabilities of two models by combining probabilistic outputs through soft-voting aggregation. The EnsembleGuard model is trained and validated on benchmark datasets including FaceForensics++, Deepfake Detection Challenge (DFDC), ASVspoof, and Celeb-DF, employing comprehensive preprocessing strategies such as multi-region facial feature extraction, temporal sequence alignment, and advanced data augmentation. Extensive experiments demonstrate that EnsembleGuard achieves superior performance with an accuracy of 97.8%, precision of 96.9%, recall of 97.5%, and F1 score of 97.2 on FaceForensics++ dataset and significantly outperforming single-model and integration-layer baselines.
科研通智能强力驱动
Strongly Powered by AbleSci AI