Deepfakes pose rising threats to digital trust, with increasingly convincing forgeries evading conventional detectors. Existing tools often rely on narrow cues from a single mechanism (e.g., a model or watermark), leaving them prone to false positives, false negatives, and circumvention. The Deepfake Detection System (DDS) [3] improves robustness by combining multiple models and user votes, but incurs high blockchain overhead and lacks voter anonymity, limiting scalability and deployment. We present a scalable, privacy-preserving framework that extends DDS with: (i) a content filtration layer using exact and perceptual hashing to remove redundant verification and cut blockchain load, and (ii) an adaptation of the IncogniSense protocol [4] for anonymous, reputation-weighted voting in a blockchain setting. Experiments show over 90% reduction in blockchain load, low-latency responses under stress, and strong privacy guarantees. These results address the key scalability and privacy barriers in blockchain-based deepfake detection, delivering a deployable, high-throughput framework for real-world platforms.