Interpretability has been regarded as an essential component for deploying deep neural networks, in which the saliency-based method is one of the most prevailing interpretable approaches since it can generate individually intuitive heatmaps that highlight parts of the input image that are most important to the decision of the deep networks on a particular classification target. However, heatmaps generated by existing methods either contain little information to represent objects (perturbation-based methods) or cannot effectively locate multi-class objects (activation-based approaches). To address this issue, a two-stage framework for visualizing the interpretability of deep neural networks, called Activation Optimized with Perturbation (AOP), is designed to optimize activation maps generated by general activation-based methods with the help of perturbation-based methods. Finally, in order to obtain better explanations for different types of images, we further present an instance of the AOP framework, Smooth Integrated Gradient-based Class Activation Map (SIGCAM), which proposes a weighted GradCAM by applying the feature map as weight coefficients and employs I-GOS to optimize the base-mask generated by weighted GradCAM. Experimental results on common-used benchmarks, including deletion and insertion tests on ImageNet-1k, and pointing game tests on COCO2017, show that the proposed AOP and SIGCAM outperform the current state-of-the-art methods significantly by generating higher quality image-based saliency maps.

Cloud computing encourages users to outsource their data to cloud storage. Data outsourcing means that users lose physical autonomy on their own data, which makes remote data integrity verification become a critical challenge for potential cloud users. To free user from the burden incurred by frequent integrity verifications, Third Party Auditor (TPA) is introduced to perform verifications on behalf of user for data integrity assurance. However, existing public auditing schemes rely on the assumption that TPA is trusted, thus these schemes cannot be directly extended to support the outsourced auditing model, where TPA might be dishonest and any two of the three involved entities (i.e. user, TPA, and cloud service provider) might be in collusion. In this paper, we propose a dynamic outsourced auditing scheme which cannot only protect against any dishonest entity and collision, but also support verifiable dynamic updates to outsourced data. We present a new approach, based on batch-leaves-authenticated Merkle Hash Tree (MHT), to batch-verify multiple leaf nodes and their own indexes all together, which is more appropriate for the dynamic outsourced auditing system than traditional MHT-based dynamism approaches that can only verify many leaf nodes one by one. Experimental results show that our solution minimizes the costs of initialization for both user and TPA (compared to existing static outsourced auditing scheme), and incurs a lower price of dynamism at user side.

In this paper, we build a generalized field-aware factorization machine (GFFM) based on FFM, which stores different feature embeddings in multiple files separately instead of one single file. By making use of the corresponding location relationship between the field index and feature index, GFFM can significantly reduce the computation time. Also, features in GFFM are subdivided into dynamic ones and static ones. When modeling, the temporal variation of dynamic features such as fashion trends and user behavior preferences are considered to establish an accurate dynamic model based on the time window. We further propose DGFFM which uses the wide & deep framework to jointly train GFFM and DenseNet, aiming to combine the benefits of traditional machine learning methods including their faster learning speed on low-rank features and ability to extract high dimensional features. DGFFM add extra first-order identity features into the FM/FFM related models for the first time by adopting the multi-storage pattern designed in this paper. Experimental results on four real-world datasets demonstrate that GFFM can obtain higher accuracy and computational efficiency compared with the state-of-the-art methods used in recommendation and CTR tasks, and DGFFM can achieve an even higher prediction accuracy than recent DNN models.

Remote data auditing service is important for mobile clients to guarantee the intactness of their outsourced data stored at cloud side. To relieve mobile client from the nonnegligible burden incurred by performing the frequent data auditing, more and more literatures propose that the execution of such data auditing should be migrated from mobile client to third‐party auditor (TPA). However, existing public auditing schemes always assume that TPA is reliable, which is the potential risk for outsourced data security. Although Outsourced Proofs of Retrievability (OPOR) have been proposed to further protect against the malicious TPA and collusion among any two entities, the original OPOR scheme applies only to the static data, which is the limitation that should be solved for enabling data dynamics. In this paper, we design a novel authenticated data structure called bv23Tree, which enables client to batch‐verify the indices and values of any number of appointed leaves all at once for efficiency. By utilizing bv23Tree and a hierarchical storage structure, we present the first solution for Dynamic OPOR (DOPOR), which extends the OPOR model to support dynamic updates of the outsourced data. Extensive security and performance analyses show the reliability and effectiveness of our proposed scheme.

As information technology develops, cloud storage has been widely accepted for keeping volumes of data. Remote data auditing scheme enables cloud user to confirm the integrity of her outsourced file via the auditing against cloud storage, without downloading the file from cloud. In view of the significant computational cost caused by the auditing process, outsourced auditing model is proposed to make user outsource the heavy auditing task to third party auditor (TPA). Although the first outsourced auditing scheme can protect against the malicious TPA, this scheme enables TPA to have read access right over user’s outsourced data, which is a potential risk for user data privacy. In this paper, we introduce the notion of User Focus for outsourced auditing, which emphasizes the idea that lets user dominate her own data. Based on User Focus, our proposed scheme not only can prevent user’s data from leaking to TPA without depending on data encryption but also can avoid the use of additional independent random source that is very difficult to meet in practice. We also describe how to make our scheme support dynamic updates. According to the security analysis and experimental evaluations, our proposed scheme is provably secure and significantly efficient.