Latest news with #DeepfakeDetectionChallengeDataset
Barnama
5 days ago
- Business
- Barnama
FACIA Achieves 100 Pct Accuracy On Deepfake Detection Benchmark
BUSINESS KUALA LUMPUR, July 23 (Bernama) -- FACIA, a global leader in facial biometric technology, has announced that its deepfake detection system achieved 100 per cent accuracy on Meta's Deepfake Detection Challenge Dataset (DFDC), reinforcing its position in combatting synthetic media threats. In a statement, the company said its proprietary algorithm was tested on over 100,000 images and videos across multiple datasets, with an overall detection accuracy of 99.6 per cent. 'This is not just about setting a benchmark. Deepfakes are proliferating rapidly, and scalable detection systems are now critical for public agencies, social platforms, and financial services,' said FACIA Chief Technology Officer, Daniyal Assad Chughtai. Deepfake content tripled in 2023, with manipulated media widely used in fraud, disinformation, and non-consensual content, prompting increased concern and regulatory action globally. The DFDC dataset, considered a benchmark for deepfake testing, includes 2,100 altered videos using eight facial modification techniques. FACIA also tested its algorithm on an internal dataset of 3,430 artificial intelligence-generated images from tools such as Midjourney, Artbreeder, and achieving 89.01 per cent accuracy. Combined testing across four additional open-source datasets contributed to the system's total detection score, which FACIA said was relevant to industries such as finance, defence, and immigration. Unlike conventional frame-by-frame methods, FACIA's detection uses a multi-layered pipeline tailored for today's advanced deepfake techniques, with consistently low false acceptance and rejection rates. The technology's reliability is suited to high-assurance identity verification, offering real-time, scalable implementation for deepfake-vulnerable environments. FACIA plans to further enhance its system with multilingual spoof detection, broader dataset training, and improved application programming interface (API) support to facilitate integration with third-party platforms. The announcement comes amid growing scrutiny of synthetic media on major platforms such as Meta, TikTok, and X (formerly Twitter), which are under pressure to counter misinformation and manipulated content.
Barnama
5 days ago
- Barnama
FACIA Reports 100% Accuracy in Deepfake Detection Across Industry Datasets
LONDON, July 23 (Bernama) -- FACIA, a global leader in facial biometric technology, announced that its deepfake detection system has achieved 100% accuracy on multiple industry datasets, including Meta's Deepfake Detection Challenge Dataset (DFDC). FACIA's proprietary algorithm was tested across more than 100,000 images and videos, reporting an overall detection accuracy of 99.6%. The announcement comes amid growing global concern over synthetic media. In 2023 alone, the number of deepfake videos tripled, with total deepfake content increasing eightfold. These manipulated assets are now widely used in fraud schemes, misinformation campaigns, and non-consensual content creation, prompting regulatory responses worldwide.
