The paper introduces a universal distortion function called UNIWARD (UNiversal WAvelet Relative Distortion) for steganography in arbitrary domains. UNIWARD is designed to minimize detectability by embedding changes in regions that are difficult to model in multiple directions, such as textures or noisy areas, while avoiding smooth regions or clean edges. The distortion is computed as a sum of relative changes in wavelet coefficients of the cover image. The paper discusses the design of the distortion function, its implementation in the spatial, JPEG, and side-informed JPEG domains, and empirical evaluations using rich models and targeted attacks. The results show that UNIWARD matches or outperforms current state-of-the-art steganographic methods in terms of security and performance. The authors also explore the impact of the directional filter bank and a stabilizing constant on the security of the scheme.The paper introduces a universal distortion function called UNIWARD (UNiversal WAvelet Relative Distortion) for steganography in arbitrary domains. UNIWARD is designed to minimize detectability by embedding changes in regions that are difficult to model in multiple directions, such as textures or noisy areas, while avoiding smooth regions or clean edges. The distortion is computed as a sum of relative changes in wavelet coefficients of the cover image. The paper discusses the design of the distortion function, its implementation in the spatial, JPEG, and side-informed JPEG domains, and empirical evaluations using rich models and targeted attacks. The results show that UNIWARD matches or outperforms current state-of-the-art steganographic methods in terms of security and performance. The authors also explore the impact of the directional filter bank and a stabilizing constant on the security of the scheme.