Privacy Aware Offloading of Deep Neural Networks
This addresses privacy concerns for users of resource-constrained devices like IoT when leveraging cloud computing, representing an incremental improvement in secure offloading methods.
The paper tackles the problem of privacy risks when offloading deep neural network computations to the cloud by proposing a data obfuscation technique that maintains high classification accuracy on obfuscated data.
Deep neural networks require large amounts of resources which makes them hard to use on resource constrained devices such as Internet-of-things devices. Offloading the computations to the cloud can circumvent these constraints but introduces a privacy risk since the operator of the cloud is not necessarily trustworthy. We propose a technique that obfuscates the data before sending it to the remote computation node. The obfuscated data is unintelligible for a human eavesdropper but can still be classified with a high accuracy by a neural network trained on unobfuscated images.