In order to solve the problems of decentralized copyright registration, hidden content tampering and difficulty in source tracking in the process of digital transmission of folk music, a blockchain-supported algorithm framework for copyright protection of folk music was proposed. This method coupled robust audio watermarking, convolutional autoencoder and alliance chain certificate storage mechanism, and constructed a technical link covering copyright registration, forgery detection, real-time traceability and online verification. In the implementation, the system completed the identity binding of the work through time-frequency feature extraction and latent space watermark embedding, and then used the reconstruction error of the convolutional autoencoder to identify the forgery behaviors such as segment splicing, pitch sandaling, speed variation and deep imitation singing. The on-chain hash index and smart contract were combined to realize ownership verification and evidence retention. Experimental results show that on 4800 folk music audio samples, the accuracy of copyright confirmation of the proposed method reaches 95.8%, the average detection rate of forged samples reaches 94.0%, the success rate of watermark extraction remains above 93.6%, the accuracy of source attribution reaches 91.1%, and the average delay under 100 concurrent verification requests is 138 ms. The research shows that the collaborative design of blockchain and deep learning can effectively improve the credibility, real-time performance and enforceability of folk music copyright protection, and provide a feasible computer technology path for the standardized circulation and long-term protection of folk music digital resources.
