Recently a great number of ubiquitous Internet-of-Things (IoT) devices have been connecting to the Internet. With the massive amount of IoT data, the cloud-based intelligent applications have sprang up to support accurate monitoring and decision-making. In practice, however, the intrinsic transport bottleneck of the Internet severely handicaps the real-time performance of the cloud-based intelligence depending on IoT data. In the past few years, researchers have paid attention to the computing paradigm of edge-cloud collaboration; they offload the computing tasks from the cloud to the edge environment, in order to avoid transmitting much data through the Internet to the cloud. To present, it is still an open issue to effectively allocate the deep learning task (i.e., deep neural network computation) over the edge-cloud system to shorten the response time of application. In this paper, we propose the latency-minimum allocation (LMA) problem, aimed at allocating the deep neural network (DNN) layers over the edge-cloud environment while the total latency of processing this DNN can be minimized. First, we formalize the LMA problem in general form, prove its NP-hardness, and present an insightful characteristic of feasible DNN layer allocations. Second, we design an approximate algorithm, called CoEdge, which can handle the LMA problem in polynomial time. By exploiting the communication and computation resources of the edge, CoEdge greedily selects the beneficial edge nodes and allocates the DNN layers to the selected nodes by a recursion-based policy. Finally, we conduct extensive simulation experiments with realistic setups, and the experimental results show the efficacy of CoEdge in reducing the deep learning latency compared to two state-of-the-art schemes.

Mobile crowd sensing (MCS) is a new paradigm for urban-scale monitoring. This paper concentrates on the Cost-Fair Task Allocation (CFA) problem for the MCS scenario where the collaboration of multiple probabilistic mobilephone users is needed to yield more reliable observation. CFA aims to allocate sensing tasks to users so that the sensing costs undertaken by all users are as balancing as possible, while the requirement of the requester for data reliability can be satisfied. CFA is greatly important to MCS campaigns in terms of reliability and sustainability. We design two algorithms to solve the CFA problem in the offline and online cases, respectively. Specifically, we propose a novel penalty-based model to reformulate the offline CFA problem, and based on this model, we design an offline algorithm, which can yield a computation-efficient ε-solution with any small ε > 0. For the online case, we design a polynomial-time approximation algorithm, which struggles to allocate each of the sequentially arriving tasks to users as fairly as possible, and can achieve an upper-bounded competitiveness relative to the optimal CFA solution. Finally, we conduct extensive numeric analyses to validate the performance of our algorithms under diverse experimental setups.

Smart contract is the core of blockchain operation, and contract vulnerability will cause huge economic losses. Therefore, effective smart contract vulnerability detection is of vital importance and attracts more and more attention. In this paper, we propose a vulnerability detection model (VDM-AEI) based on automatic feature extraction and feature interaction. For the first time, this model converts smart contracts into gray images and uses VGG16 and GRU models to automatically extract vulnerability features and filter effective features, respectively. Then, a contract graph and an expert knowledge feature vector are constructed by using commonly used methods as part of feature construction. Next, AutoInt and DCN networks are used to build a dual feature interaction network to obtain more abundant vulnerability feature information, which extracts high-dimensional nonlinear features from the low and sparse features of the contract graph feature vector and the expert knowledge-defined feature vector. Finally, all output features of GRU, AutoInt and DCN networks are integrated to obtain vulnerability classification results through fully connected neural networks. We conducted extensive experiments on the ESC and VSC datasets for reentrancy vulnerabilities, timestamp dependency vulnerabilities, and infinite loop vulnerabilities. The experimental results prove the effectiveness and accuracy of the VDM-AEI model. Compared with the latest vulnerability detection model CGE, the accuracy rates of the 3 types of vulnerability detection are improved by 10.85%, 6.18%, and 12.34%, respectively. In addition, the predicted F1 scores of VDM-AEI are all greater than 95%, and the recall rate is no less than 94%.

Aiming to address the problems of the high bit error rate (BER) of demodulation or low classification accuracy of modulation signals with a low signal-to-noise ratio (SNR), we propose a double-residual denoising autoencoder method with a channel attention mechanism, referred to as DRdA-CA, to improve the SNR of modulation signals. The proposed DRdA-CA consists of an encoding module and a decoding module. A squeeze-and-excitation (SE) ResNet module containing one residual connection is modified and then introduced into the autoencoder as the channel attention mechanism, to better extract the characteristics of the modulation signals and reduce the computational complexity of the model. Moreover, the other residual connection is further added inside the encoding and decoding modules to optimize the network degradation problem, which is beneficial for fully exploiting the multi-level features of modulation signals and improving the reconstruction quality of the signal. The ablation experiments prove that both the improved SE module and dual residual connections in the proposed method play an important role in improving the denoising performance. The subsequent experimental results show that the proposed DRdA-CA significantly improves the SNR values of eight modulation types in the range of −12 dB to 8 dB. Especially for 16QAM and 64QAM, the SNR is improved by 8.38 dB and 8.27 dB on average, respectively. Compared to the DnCNN denoising method, the proposed DRdA-CA makes the average classification accuracy increase by 67.59∼74.94% over the entire SNR range. When it comes to the demodulation, compared with the RLS and the DnCNN denoising algorithms, the proposed denoising method reduces the BER of 16QAM by an average of 63.5% and 40.5%, and reduces the BER of 64QAM by an average of 46.7% and 18.6%. The above results show that the proposed DRdA-CA achieves the optimal noise reduction effect.

Recently, there has been a growing interest in using chargers to extend the operational longevity of UAVs (drones). In this paper, we explore a charger-assisted drone application where a drone observes points of interest while a mobile charger moves to recharge its battery. We focus on the route and charging schedule of the drone and mobile charger to maximize observation utility in the shortest possible time, while ensuring continuous drone operation. In our problem, the drone and mobile charger cooperate to complete a task. Their discrete-continuous hybrid actions pose a major computational challenge. To address this issue, we present a hybrid-action deep reinforcement learning framework, called HaDMC, which uses a typical policy learning algorithm to generate latent continuous actions. We specifically design and train an action decoder. It involves two pipelines to convert the latent continuous actions into the original hybrid actions for the drone and mobile charger to directly interact with environment. We incorporate a mutual learning scheme into model training, emphasizing collaboration over individual actions. By extensive numerical experiments, we evaluate HaDMC and compare it with state-of-the-art approaches. The experimental results demonstrate the effectiveness and efficiency of our solution.