Facial recognition has been increasingly employed in real-world applications, raising serious privacy concerns over mass surveillance and unauthorized tracking. Existing anti-facial recognition methods perturb face images using generative models to protect privacy but often introduce global artifacts, depend on reference image prompts, or require target identity, compromising both visual quality and anonymity. To address the above limitations, we introduce MASQUE, a diffusion-based framework that generates localized adversarial makeup guided by user-defined text prompts. By leveraging precise null-text inversion, targeted cross-attention fusion with masking, and a novel pairwise adversarial guidance mechanism using images of the same individual, MASQUE achieves robust dodging performance without the need for an external target identity. Extensive evaluations on open-source FR models and commercial APIs show that MASQUE significantly enhances privacy protection over all baselines, achieving average protection success rates of 90% for identification and 87% for verification while preserving high perceptual fidelity.