Privacy-Preserving Distributed Accelerated Resource Allocation in General Directed Networks

With the proliferation of consumer electronics applications, related devices usually entail allocating system resources to process large-scale distributed data in efficient ways. Effective allocation of system resources is crucial for improving device performance, optimizing energy efficiency, and safeguarding user experience. In this article, we study a distributed constraint-coupled resource allocation problem. Each node in directed networks maintains a private cost function and achieves a solution via only local communication. Problems of this nature considering efficiency and privacy usually emerge in consumer electronics (energy management, intelligent dispatch, energy conservation and utility optimization, etc.), which deserve to be addressed in depth. To tackle this problem, we propose an efficient privacy-preserving distributed accelerated resource allocation algorithm based on state decomposition with added noises. In one aspect, by applying the parametric distributed momentum method to the push-pull distributed resource allocation algorithm, we obtain accelerated allocation in directed networks. Given privacy concerns, our algorithm employs the differential privacy strategy along with a state decomposition scheme to avoid disclosing individual state information in communication. It is proven that our algorithm with a constant stepsize can linearly converge to a neighborhood of the optimal allocation while preserving privacy. The trade-off between convergence accuracy and privacy level is also shown. Extensive simulations further illustrate the efficacy of our algorithm and validate our theoretical discoveries.