The landscape of computer vision and artificial intelligence is evolving rapidly with the rise of powerful generative models, transforming how data-driven challenges are addressed. Building on the success of the first and second editions of the Synthetic Data for Face and Gesture Analysis (SD-FGA) workshop, SD-FGA 2026 (at FG 2026) will spotlight advances in synthetic data generation-from GANs and VAEs to diffusion models and their growing impact on facial analysis, gesture recognition, and behavioral understanding, where privacy, ethics, and data diversity remain key concerns. This third edition expands the scope to synthetic data for attack generation and detection across both physical and behavioral dimensions, reflecting the urgent need for robust, secure AI in open-world settings. The workshop will bring together researchers and practitioners to examine how synthetic data can enable fair and reliable model development while also supporting the simulation, generation, and detection of adversarial and spoofing attacks, ultimately advancing secure, ethical, and scalable face and gesture analysis systems.