Sphynx: ReLU-Efficient Network Design for Private Inference
This work addresses privacy concerns in deep learning by improving efficiency for private inference, though it is incremental as it builds on existing methods with a focus on optimization.
The paper tackles the high latency in private inference due to non-linear operations like ReLUs by introducing Sphynx, a ReLU-efficient network design method, which achieves Pareto dominance on CIFAR-100 and enables private inference on larger datasets such as Tiny-ImageNet and ImageNet.
The emergence of deep learning has been accompanied by privacy concerns surrounding users' data and service providers' models. We focus on private inference (PI), where the goal is to perform inference on a user's data sample using a service provider's model. Existing PI methods for deep networks enable cryptographically secure inference with little drop in functionality; however, they incur severe latency costs, primarily caused by non-linear network operations (such as ReLUs). This paper presents Sphynx, a ReLU-efficient network design method based on micro-search strategies for convolutional cell design. Sphynx achieves Pareto dominance over all existing private inference methods on CIFAR-100. We also design large-scale networks that support cryptographically private inference on Tiny-ImageNet and ImageNet.