Abstract
Due to their high computational and memory demand, deep learning applications are mainly restricted to high-performance units, e.g., cloud and edge servers. Particularly, in Internet of Things (IoT) systems, the data acquired by pervasive devices is sent to the computing servers for classification. However, this approach might not be always possible because of the limited bandwidth and the privacy issues. Furthermore, it presents uncertainty in terms of latency because of the unstable remote connectivity. To support resource and delay requirements of such paradigm, joint and real-time deep co-inference framework with IoT synergy was introduced. However, scheduling the distributed, dynamic and real-time Deep Neural Network (DNN) inference requests among resource-constrained devices has not been well explored in the literature. Additionally, the distribution of DNN has drawn the attention to the privacy protection of sensitive data. In this context, various threats have been presented, including white-box attacks, where malicious devices can accurately recover received inputs if the DNN model is fully exposed to participants. In this paper, we introduce a methodology aiming at distributing the DNN tasks onto the resource-constrained devices of the IoT system, while avoiding to reveal the model to participants. We formulate such an approach as an optimization problem, where we establish a trade-off between the latency of co-inference, the privacy of the data, and the limited resources of devices. Next, due to the NP-hardness of the problem, we shape our approach as a reinforcement learning design adequate for real-time applications and highly dynamic systems, namely RL-PDNN. Our system proved its ability to outperform existing static approaches and achieve close results compared to the optimal solution.
Highlights
The deep neural networks represent the core technique for a wide spectrum of applications, including computer vision [2] and natural language processing [1]
To relax the optimization problem, we propose a novel approach based on Reinforcement Learning (RL) for privacy-aware distributed Convolutional Neural Networks (CNNs) networks, namely RL-PDNN
PRIVACY-AWARE DISTRIBUTED CNN FOR Internet of Things (IoT) DEVICES we present our distributed CNN approach for privacy-aware and low decision-latency applications, the system model, followed by our strategy formulated as an optimization problem
Summary
The deep neural networks represent the core technique for a wide spectrum of applications, including computer vision [2] and natural language processing [1]. Researchers have investigated the feasibility of leveraging the resources of IoT devices to jointly allocate different segments of deep neural networks and execute the classification at the vicinity of the source device [5]–[7] These aforementioned works mainly studied the optimal partition strategy that reduces the transmitted data and the dependency between inference participants. To match resource-consuming DNN solutions with the constraints of IoT devices, we exploit the hierarchical design characterizing deep learning models in order to suitably place layers of dynamic incoming classification requests. To relax the optimization problem, we propose a novel approach based on Reinforcement Learning (RL) for privacy-aware distributed CNN networks, namely RL-PDNN This approach learns the allocation policy and takes real-time actions based on the available resources, the dynamics of requests, and the required privacy level.
Sign-in/Register to view highlights & summary Sign-in/Register to access full text options 
Free) 
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call
