1 June 2020 in Neuroscience
4 November 2020 in Cognitive Neurodynamics
26 October 2021 in IEEE Access
With the fast advancement in cloud computing, progressively more users store their applications and data on the cloud. Cloud computing has lots of features, e.g. virtualization, multi-user, efficiency, cost savings, and most importantly security. Machine learning approaches based on neural networks are being widely applied in cloud infrastructure when training is performed however this may produce possible privacy and security risk as direct access to raw data is required. To address this problem, we propose a new security design using Artificial Neural Networks (ANN) and encryption to confirm a safe communication system in the cloud environment, by letting the third parties access the data in an encrypted form for processing without disclosing the data of the provider party to secure important information. In this paper, to train neural networks using encrypted data we considered the Matrix Operation-based Randomization and Encipherment (MORE) technique, based on Fully Homomorphic Encryption (FHE). This technique allows the computations to be performed directly on floating-point data within a neural network with a minor computational overhead. We examined the speech and voice recognition problem and the performance of the proposed method has been validated in MATLAB simulation. Results showed that applying neural network training with MORE offers improved accuracy, runtime, and performance. These results highlight the potential of the proposed method to protect privacy and provide high accuracy in a reasonable amount of time when compared to other state-of-the-art techniques.
1 March 2023 in Connection Science
Image super-resolution reconstruction is one of the methods to improve resolution by learning the inherent features and attributes of images. However, the existing super-resolution models have some problems, such as missing details, distorted natural texture, blurred details and too smooth after image reconstruction. To solve the above problems, this paper proposes a Multi-scale Dual-Attention based Residual Dense Generative Adversarial Network (MARDGAN), which uses multi-branch paths to extract image features and obtain multi-scale feature information. This paper also designs the channel and spatial attention block (CSAB), which is combined with the enhanced residual dense block (ERDB) to extract multi-level depth feature information and enhance feature reuse. In addition, the multi-scale feature information extracted under the three-branch path is fused with global features, and sub-pixel convolution is used to restore the high-resolution image. The experimental results show that the objective evaluation index of MARDGAN on multiple benchmark datasets is higher than other methods, and the subjective visual effect is better. This model can effectively use the original image information to restore the super-resolution image with clearer details and stronger authenticity.
22 January 2022 in Medical & Biological Engineering & Computing
27 February 2023 in IEEE Transactions on Services Computing
Location data play an important role in location-based services (LBS) since they can help a service provider analyze users' daily activities and further derive users' behavioral patterns. However, in the meanwhile the use of the location data in LBS can also enable the service provider to track users' journeys which will reveal their personal information, such as house address and places of interest they visited, all of which users are usually unwilling to disclose. Therefore, security and privacy incidents can and do occur often. In this paper, aiming at how to hide location privacy and meanwhile provide an accurate search for LBS, we design a new geometric searchable encryption scheme under a public-key system (a.k.a. asymmetric system) for different types of geometric range queries, which enables the service provider to respond to the range queries from users accurately without learning the information about users' location. Towards this goal, we first exploit a novel encoding, and then develop some sophisticated transformations in combination with a special pairing function, which eventually converts different types of geometric range queries in LBS into different inner product problems. Finally, we present our construction by virtue of the techniques above with the inner product encryption technique. In addition, a comprehensive experimental analysis running on a specific application of privacy-preserving LBS and comparison demonstrate, our proposal is able to guarantee a higher accuracy for users' range queries without any false positive, and meanwhile express richer queries. Furthermore, we show another application of our proposal to privacy-preserving remote medical diagnosis at the end of this paper.