An attribute-based framework for privacy preserving image querying

Abstract

We are specifically concerned with scenarios in which multimedia data is stored once on the server and the same data is queried by multiple parties. We propose a framework for privacy preserving querying, in which encryption is performed only once, and the ciphertexts are stored on a database server. Rather than using public-key homomorphic cryptosystems, the parties querying the database first derive an "attribute" from their query signal. They can decrypt the server's ciphertext only if their attribute satisfies a specified mathematical condition. This queryspecific decryption capability makes attribute based cryptography a vital addition to the secure signal processor's toolkit. We give an example of a construction for privacy preserving querying, in which a client can privately retrieve an image from the server if attribute vectors extracted from the server's and client's images are close enough in Euclidean distance. IEEE International Conference on Image Processing (ICIP)This work may not be copied or reproduced in whole or in part for any commercial purpose. Permission to copy in whole or in part without payment of fee is granted for nonprofit educational and research purposes provided that all such whole or partial copies include the following: a notice that such copying is by permission of Mitsubishi Electric Research Laboratories, Inc.; an acknowledgment of the authors and individual contributions to the work; and all applicable portions of the copyright notice. Copying, reproduction, or republishing for any other purpose shall require a license with payment of fee to Mitsubishi Electric Research Laboratories, Inc. All rights reserved. ABSTRACTWe are specifically concerned with scenarios in which multimedia data is stored once on the server and the same data is queried by multiple parties. We propose a framework for privacy preserving querying, in which encryption is performed only once, and the ciphertexts are stored on a database server. Rather than using public-key homomorphic cryptosystems, the parties querying the database first derive an "attribute" from their query signal. They can decrypt the server's ciphertext only if their attribute satisfies a specified mathematical condition. This query-specific decryption capability makes attributebased cryptography a vital addition to the secure signal processor's toolkit. We give an example of a construction for privacy preserving querying, in which a client can privately retrieve an image from the server if attribute vectors extracted from the server's and client's images are close enough in Euclidean distance.

Privacy preserving identification using sparse approximation with ambiguization

“…N σ 2 z and considering a ternary sparse encoding, we have: In order to preserve distances, given the variance of the measurement noise σ 2 z and sparsity level S x 3 , we have 3 Sparsity level Sx obtained based on the desired distortion level imposed by sparse approximation. Fig.…”

“…(7) Before discussion on the above inequality, we consider lower and upper bounds of distances, which we will use in Section IV. In the original signal domain, considering query vector y in the general case, the lower and upper bounds of distances between the vector x (m) , m ∈ [M ] in the database and query y can be given as N σ 2 In order to preserve distances, given the variance of the measurement noise σ 2 z and sparsity level S x 3 , we have 3 Sparsity level Sx obtained based on the desired distortion level imposed by sparse approximation. to set…”

“…For this reason, the exact identification is replaced by a search of list of closest items, i.e., one tries to trade-off the accuracy of identification by the search complexity. In recent years, many methods providing efficient ANN solutions for multi-billion entry datasets were proposed and we named some of them without pretending to be exhaustive in our overview [1]- [3].…”

Secure and efficient approximate nearest neighbors search

“…Eq. (7) was a first attempt, but the Server can do much better thanks to dus defined in (8). The idea is simple: if dus(i, j, m) is close to zero, it means that c …”

“…Paper [8] builds a solution using attribute based encryption to avoid the last two drawbacks of 3.1.2. The User is able to decrypt the metadata ti if and only if it knows a vector q such that q − xi 2 ≤ τ (vectors are here elements of Z d and τ ∈ N).…”