Biblio

While cloud-based deep learning benefits for high-accuracy inference, it leads to potential privacy risks when exposing sensitive data to untrusted servers. In this paper, we work on exploring the feasibility of steganography in preserving inference privacy. Specifically, we devise GHOST and GHOST+, two private inference solutions employing steganography to make sensitive images invisible in the inference phase. Motivated by the fact that deep neural networks (DNNs) are inherently vulnerable to adversarial attacks, our main idea is turning this vulnerability into the weapon for data privacy, enabling the DNN to misclassify a stego image into the class of the sensitive image hidden in it. The main difference is that GHOST retrains the DNN into a poisoned network to learn the hidden features of sensitive images, but GHOST+ leverages a generative adversarial network (GAN) to produce adversarial perturbations without altering the DNN. For enhanced privacy and a better computation-communication trade-off, both solutions adopt the edge-cloud collaborative framework. Compared with the previous solutions, this is the first work that successfully integrates steganography and the nature of DNNs to achieve private inference while ensuring high accuracy. Extensive experiments validate that steganography has excellent ability in accuracy-aware privacy protection of deep learning.

Dynamic searchable symmetric encryption (DSSE) that enables a client to perform searches and updates on encrypted data has been intensively studied in cloud computing. Recently, forward privacy and backward privacy has engaged significant attention to protect DSSE from the leakage of updates. However, the research in this field almost focused on keyword-level updates. That is, the client needs to know the keywords of the documents in advance. In this paper, we proposed a document-level update scheme, DBP, which supports immediate deletion while guaranteeing forward privacy and backward privacy. Compared with existing forward and backward private DSSE schemes, our DBP scheme has the following merits: 1) Practicality. It achieves deletion based on document identifiers rather than document/keyword pairs; 2) Efficiency. It utilizes only lightweight primitives to realize backward privacy while supporting immediate deletion. Experimental evaluation on two real datasets demonstrates the practical efficiency of our scheme.