Recently, private inference (PI) has addressed the rising concern over data and model privacy in machine learning inference as a service. However, existing PI frameworks suffer from high computational and communication costs due to the expensive multi-party computation (MPC) protocols. Existing literature has developed lighter MPC protocols to yield more efficient PI schemes. We, in contrast, propose to lighten them by introducing an empirically-defined privacy evaluation. To that end, we reformulate the threat model of PI and use inference data privacy attacks (IDPAs) to evaluate data privacy. We then present an enhanced IDPA, named distillation-based inverse-network attack (DINA), for improved privacy evaluation. Finally, we leverage the findings from DINA and propose C2PI, a two-party PI framework presenting an efficient partitioning of the neural network model and requiring only the initial few layers to be performed with MPC protocols. Based on our experimental evaluations, relaxing the formal data privacy guarantees C2PI can speed up existing PI frameworks, including Delphi [1] and Cheetah [2], up to 2.89× and 3.88× under LAN and WAN settings, respectively, and save up to 2.75× communication costs.

Unlike traditional heuristics, we provide in this paper an optimization framework for the routing and wavelength assignment (RWA) problems with the objective of minimizing the rejection penalty of the connection demands in an all-optical wavelength-division-multiplexing (WDM) network. Our new link-based formulation takes the fairness issue and the limited wavelength conversion into consideration. The framework employs a decomposition approach to decide on the rejection/selection of the route and wavelength assignment for a semilightpath, by appropriately relaxing some of the constraints in the Lagrangean relaxation (LR) method. At the higher level, we update Lagrange multipliers iteratively with the subgradient method. At the lower level, we propose the modified minimum cost semilightpath (MMCSLP) algorithm to solve all the subproblems. A heuristic algorithm is also proposed to generate a feasible RWA scheme based on the solution to the dual problem. When compared with some latest methodology in the literature, we demonstrate that our framework can achieve better performance in terms of the computation time and the number of connection demands rejected. The much shorter computation time is due to the polynomial time complexity of our framework. In addition to achieving a very good (near-optimal) solution, the influence from the change of the number of converters is studied. Finally, we demonstrate that our framework produces fairer routing decisions by adjusting some design parameters in our framework.