Privacy-Preserving Image Sharing Via Sparsifying Layers on Convolutional Groups

Abstract

We propose a practical framework to address the problem of privacy-aware image sharing in large-scale setups. We argue that, while compactness is always desired at scale, this need is more severe when trying to furthermore protect the privacy-sensitive content. We therefore encode images, such that, from one hand, representations are stored in the public domain without paying the huge cost of privacy protection, but ambiguated and hence leaking no discernible content from the images, unless a combinatorially-expensive guessing mechanism is available for the attacker. From the other hand, authorized users are provided with very compact keys that can easily be kept secure. This can be used to disambiguate and reconstruct faithfully the corresponding access-granted images. We achieve this with a convolutional autoencoder of our design, where feature maps are passed independently through sparsifying transformations, providing multiple compact codes, each responsible for reconstructing different attributes of the image. The framework is tested on a large-scale database of images with public implementation available.Index Termsprivacy-preserving image sharing, convolutional autoencoders, sparse representation, learned compression, image obfuscation.

Privacy-Preserving near Neighbor Search via Sparse Coding with Ambiguation

“…Table I provides a quantitative comparison on reconstructed images using normalized MSE and SSIM (Structural Similarity Index) on MNIST [28], Fashion-MNIST [29], and CIFAR-10 [30] databases, where we applied the proposed method on latent representation of a designed convolutional autoencoder in [12], setting L = 128 and considering one code-map for MNIST and Fashion-MNIST databases and four code-maps for CIFAR-10 database. Finally, note that based on these results, our model follows the notion of (β, γ)-recoverable privacy mechanism, which we defined in Section II.…”

“…3e-3f depict the recall measure for the CelebA database. The ground-truth was the pixel domain Euclidean distances and the latent code of the network in [12] is used to measure the approximate distances.…”

“…Concretely, this work brings the following contributions: (1) We consider the methodology of learning to hash for privacy-preserving proximity search which entails minimum information loss for authorized users. The authorized parties can purify the ambiguation noise using the shared secrecy based on the support intersection of sparse codes [12].…”

“…Moreover, we impose no restrictions on the input data, i.e., we assume that as an input we might have raw data, extracted features using any known hand crafted methods, aggregated local descriptors based on BoW, FV, VLAD [17]- [19], etc., or from the last layers of deep nets [20], or the latent space of auto-encoders [21]. We apply our model on the latent representation of a designed network in [12].…”

Multi-Level Reversible Data Anonymization via Compressive Sensing and Data Hiding

“…A privacy-preserving method to which our method has some resemblance was recently described in [20]. In the method of [20], the images are first processed through a parallel group of trained auto-encoders, each generating its own sufficiently diversified sparse code.…”