Search

Detecting oriented and densely packed objects is a challenging problem considering that the receptive field intersection between objects causes spatial feature aliasing. In this paper, we propose a convex-hull feature adaptation (CFA) approach, with the aim to configure convolutional features in accordance with irregular object layouts. CFA roots in the convex-hull feature representation, which defines a set of dynamically sampled feature points guided by the convex intersection over union (CIoU) to bound object extent. CFA pursues optimal feature assignment by constructing convex-hull sets and iteratively splitting positive or negative convex-hulls. By simultaneously considering overlapping convex-hulls and objects and penalizing convex-hulls shared by multiple objects, CFA defines a systematic way to adapt convolutional features on regular grids to objects of irregular shapes. Experiments on DOTA and SKU110K-R datasets show that CFA achieved new state-of-the-art performance for detecting oriented and densely packed objects. CFA also sets a solid baseline for convex polygon prediction on the MS COCO dataset defined for general object detection. Code is available at https://github.com/SDL-GuoZonghao/BeyondBoundingBox. [ABSTRACT FROM AUTHOR]

As a highly integrated industrial system, human cyber-physical systems (HCPSs) provide accurate and high-quality services for Industry 5.0. In HCPSs, machine learning (ML) prediction provides reliable prediction results for users based on matured models, while security and privacy protection are considerable issues. In this article, based on the modified Okamoto–Uchiyama homomorphic encryption, we propose a verifiable privacy-preserving machine learning prediction scheme for the edge-enhanced HCPSs, which outputs the verifiable prediction results for users without privacy leakage. Specifically, a batch of prediction results can be verified at one time, which improves the efficiency of verification. Security analysis shows that our scheme protects the privacy of inputs, ML model, and prediction results. The experiment results demonstrate that the edge computing architecture remarkably alleviates the computational burden of the cloud server. Furthermore, compared with other related schemes, our scheme shows the best execution efficiency, and batch verification optimizes the performance by about 15% compared with single verification on the same scale. [ABSTRACT FROM AUTHOR]

Modern CNN-based object detectors assign anchors for ground-truth objects under the restriction of object-anchor Intersection-over-Union (IoU). In this study, we propose a learning-to-match (LTM) method to break IoU restriction, allowing objects to match anchors in a flexible manner. LTM updates hand-crafted anchor assignment to “free” anchor matching by formulating detector training in the Maximum Likelihood Estimation (MLE) framework. During the training phase, LTM is implemented by converting the detection likelihood to anchor matching loss functions which are plug-and-play. Minimizing the matching loss functions drives learning and selecting features which best explain a class of objects with respect to both classification and localization. LTM is extended from anchor-based detectors to anchor-free detectors, validating the general applicability of learnable object-feature matching mechanism for visual object detection. Experiments on MS COCO dataset demonstrate that LTM detectors consistently outperform counterpart detectors with significant margins. The last but not the least, LTM requires negligible computational cost in both training and inference phases as it does not involve any additional architecture or parameter. Code has been made publicly available. [ABSTRACT FROM AUTHOR]

The booming Internet of Things makes smart healthcare a reality, while cloud-based medical storage systems solve the problems of large-scale storage and real-time access of medical data. The integrity of medical data outsourced in cloud-based medical storage systems has become crucial since only complete data can make a correct diagnosis, and public auditing protocol is a key technique to solve this problem. To guarantee the integrity of medical data and reduce the burden of the data owner, we propose an efficient privacy-preserving public auditing protocol for the cloud-based medical storage systems, which supports the functions of batch auditing and dynamic update of data. Detailed security analysis shows that our protocol is secure under the defined security model. In addition, we have conducted extensive performance evaluations, and the results indicate that our protocol not only remarkably reduces the computational costs of both the data owner and the third-party auditor (TPA), but also significantly improves the communication efficiency between the TPA and the cloud server. Specifically, compared with other related work, the computational cost of the TPA in our protocol is negligible and the data owner saves more than $2/3$ of computational cost. In addition, as the number of challenged blocks increases, our protocol saves nearly 90% of communication overhead between the TPA and the cloud server. [ABSTRACT FROM AUTHOR]

Encrypted deduplication combines encryption and deduplication in a seamless way to provide confidentiality guarantees for the physical data in deduplicated storage, yet it incurs substantial metadata storage overhead due to the additional storage of keys. We present a new encrypted deduplication storage system called ${\sf Metadedup}$ Metadedup , which suppresses metadata storage by also applying deduplication to metadata. Its idea builds on indirection, which adds another level of metadata chunks that record metadata information. We find that metadata chunks are highly redundant in real-world workloads and hence can be effectively deduplicated. We further extend ${\sf Metadedup}$ Metadedup to incorporate multiple servers via a distributed key management approach, so as to provide both fault-tolerant storage and security guarantees. We extensively evaluate ${\sf Metadedup}$ Metadedup from performance and storage efficiency perspectives. We show that ${\sf Metadedup}$ Metadedup achieves high throughput in writing and restoring files, and saves the metadata storage by up to 93.94 percent for real-world backup workloads. [ABSTRACT FROM AUTHOR]

Blockchain is a publicly distributed ledger used to record transactions in Bitcoin-like cryptocurrencies. In recent years, the successful integrations of Public-Key Cryptographic (PKC) algorithms with cryptocurrencies have driven researchers to pursue the study of PKC. However, it is challenging to technically integrate PKC algorithms with blockchain properly in that the studies of blockchain leverage to broad domains and each existing problem can lead to diverse solutions. For cryptographically-solvable problems, it is important to find a secure and practical integration of PKC algorithm with blockchain. We systematically review three major topics in cryptocurrencies, including security, privacy and scalability. We conduct a case analysis which demonstrates how to integrate PKC with blockchains. As an illustration, we propose mutable blockchain which incorporates multiple PKC schemes and show how to use it to remove double-spending transactions via redaction. We then give a concrete construction. As suggested by our performance evaluation, the adopted PKC algorithms can run scalably and efficiently and avoid bottlenecks in the system. [ABSTRACT FROM AUTHOR]

With the increasing demand to deploy convolutional neural networks (CNNs) on 5G mobile platforms, architecture designs with efficient sparse kernels (SKs) were proposed, which can save more parameters than the standard convolution while maintaining the high accuracy. Despite the great potential, neural network designs with SKs still require a lot of expert knowledge and take ample time. In this paper, we first propose a search scheme that effectively reduces the SK design space based on three aspects: composition, performance, and efficiency. Meanwhile, we completely eliminate the model training from our search scheme. Instead, an easily measurable quantity, the information field, is identified and used to predict the model accuracy in the searching process. Additionally, we provide a detailed efficiency analysis on the final designs found by our scheme. Second, based on the analysis we propose a model transformation scheme to better utilize the SK designs on existing models to either reduce the number of parameters or increase the accuracy. Last, considering the extra programming overhead and the expert knowledge required by the model transformation scheme, we develop a compiler prototype to automate the entire process, given the source code of an existing model. Experimental results show that models composed of the sparse kernel designs searched by our search scheme can beat state-of-the-art networks such as ResNets in terms of the accuracy and the efficiency. Also by using our model transformation scheme we can easily improve the accuracy (the same number of parameters) or the efficiency (the same accuracy) upon existing state-of-the-art models. [ABSTRACT FROM AUTHOR]

Wireless communication is the most important carrier for covert channels because increasing security threats have made covert channels an alternative to transmitting confidential information in untrusted wireless communications. As a solution for next-generation mobile communication, voice over LTE (VoLTE) is developed using a high-speed packet-switched core network to provide high-quality end-to-end communication services. Unlike traditional circuit-switched communication technologies, both inter-packet delays and packet orders in VoLTE traffic are limited by specific rules. Even minor modifications to the overt traffic can be detected, so the existing solutions do not apply to VoLTE. To solve this problem, we build a robust packet-dropout covert timing channel through parity casecade coding. Hash-based inter-codewords verification, cyclic redundancy check based codeword self-verification, and an adaptive mapping matrix are designed to ensure robustness and undetectability. In our scheme, the covert message is modulated into the sequence numbers of the actively dropped packets, which can be retrieved by the receiver. With the help of the verification section, the actual codewords combination can be identified to retrieve the embedded covert message. Several experiments are performed to evaluate the performance of the scheme, and the parameters of the covert timing channel (CTC) are adjusted to estimate robustness and throughput. Finally, the CTC scheme is proved to be feasible since both average bit error rate and throughput are acceptable while the covert message is stealthily transmitted. [ABSTRACT FROM AUTHOR]

The advances of artificial intelligence (AI) propels big data processing and transmission for Internet of Things (IoT), by capturing and structuring big data produced by heterogeneous devices. While applying blockchain to manage IoT devices and associated big data, the blockchain itself suffers from abuse of decentralization from anonymous users. Specifically, it has been utilized to facilitate black market trades and illegal activities. Ateniese et al. proposed using the chameleon hash (CH) to derive redactable blockchain (EuroS&P), which works by embedding a trapdoor in the basic hash function so that block content can be rewritten without causing major hard forks. In short, the redacted block hash remains unchanged. However, there is lacking intelligent design where any mistakes observed in the chain can be corrected universally and automatically. This creates disincentives to use redactable blockchain (RB) for managing big data or any data-driven business mainly due to ineffective chain redaction. To solve this problem, in this article, we propose the notion of the self-redactable blockchain (SRB) to support intelligent execution of chain redaction. Specifically, we propose the first revocable chameleon hash (RCH) to power RB. It enables an ephemeral trapdoor for finding collision without any co-operation. Periodical expiration is applied to committed hash and an ephemeral trapdoor to prevent any abuses of redaction power. We instantiate how to use our RCH to build SRB as an intelligent trust-layer for IoT. We also give a rigorous analysis as well as comprehensive experiments to validate our proposals. The evidence showed that our proposal is secure and acceptably efficient for IoT devices. [ABSTRACT FROM AUTHOR]

Cloud-based eHealth systems enable patients and medical institutions to manage their EHRs in a flexible and efficient way. Since EHRs are the most sensitive data for people and have significant value to adversaries, outsourced EHRs suffer from both internal and external threats. In this article, we first identify the threats toward the security of outsourced EHRs and present system and security criteria that a secure cloud-based eHealth system should achieve. Then we provide a blockchain-assisted eHealth framework for securing outsourced EHRs that meets the identified criteria. We also show a case study where the blockchain technology is applied to ensure the security of outsourced EHRs. Finally, we discuss several promising issues and research directions on blockchain-assisted secure EHRs in cloud-based eHealth systems. [ABSTRACT FROM AUTHOR]