Dubai’s Electronic Security Centre (DESC) has announced its plan to launch SARAAB, an artificial intelligence (AI) model that can detect deepfake videos with a 91% accuracy rate.
Announced on the first day of the sixth edition of GISEC Global 2026 at the Dubai Exhibition Centre in Expo City, this groundbreaking digital tool will be the Arab world’s first deepfake detection model designed entirely by a government agency to detect digitally manipulated content to combat sophisticated digital impersonation threats targeting public and commercial sectors.
The model, built by an Emirati team at DESC in six months, will be released by the end of 2023.
Deepfake videos can pose an alarming threat to the trustworthiness of digital content and can mimic any content with advanced sophistication.
It primarily aims to help protect public and private entities from the digital ad fraud. The new AI model detects deepfake videos by scanning video clips and generating heat maps with moving hotspots to reveal whether any frames have been manipulated.
Unlike other deepfake detection tools, which focus on analyzing a single static image, SARAAB compares an entire video from start to finish. This means that even short clips with manipulated frames hidden within genuine footage can be spotted.
Keeping open innovation and open source development in mind, DESC will release the open-source version of SARAAB on Hugging Face and other developer platforms.
This will enable cybersecurity researchers, AI developers and software engineers to examine the source code, analyze performance metrics and develop personalized models.
The open-source model can help secure other technical environments in the region, help restore public trust in digital media and provide an innovative open-source attack-defense tool for identifying evolving fake content.
