Search

Recent research has revealed that single object tracking (SOT) is susceptible to adversarial examples, with even small perturbations added to video frames leading to tracking failure. However, most of the existing methods are online attack methods. These methods generate video-specific or target-specific perturbation to deceive the trackers through a frame-by-frame online paradigm. This method is inherently inefficient in the production of adversarial examples, rendering it less than ideal for real-time application scenarios. To address this, we propose a novel offline attack method that can achieve high-frequency attack strength by using only one frame of video, called box-spoof attack (BS-Attack). In contrast to prior methods perturb feature maps or confidence maps in a frame-by-frame manner. Our BS-Attack directs its focus towards the proposal selection stage, a relatively late stage within the tracking pipeline. Notably, our method generates perturbations by utilizing the information from the initial frame, which can achieve rapid disturbance generation and high-frequency attack strength. In addition to confidence loss, we incorporate joint intersection-over-union and shrink loss to keep the predicted bounding box away from the ground truth and shrink it. BS-Attack requires only one video to generate a perturbation that can deceive the entire dataset. The UAV123, OTB100, VOT2018, and VOT2019 datasets are used to demonstrate the effectiveness of this method in attacking popular trackers such as SiamRPN, SiamRPN++, and SiamMask. The results show that BS-Attack outperforms state-of-the-art online and offline attack methods. [ABSTRACT FROM AUTHOR]

D2D-assisted mobile edge computing has attracted much attention due to its great potential to reduce the pressure on edge servers and improve the utilization of mobile devices. However, users are usually reluctant to share resources with neighbor devices, considering the limitations on resources and energy, and the concerns about security and privacy. In this paper, we firstly propose a blockchain-based resource-sharing system to solve the trust issues between unfamiliar participants. Under this system, we jointly consider resource trading and computation offloading to reduce energy consumption on the premise of maximizing the utilities of both providers and requesters. Resource trading is modeled as Stackelberg Game, and computation offloading is formalized as a mixed optimization problem. Finally, we propose the corresponding algorithms to efficiently get the optimal trading and offloading strategy. The simulation results show the effectiveness of incentivizing more participants. [ABSTRACT FROM AUTHOR]

Aiming at the low efficiency of traditional methods for the management and classification of massive remote sensing image data, a mass remote sensing image classification method based on distributed storage is proposed. The aim is to obtain near real-time image classification in mobile devices or internet applications. In this paper, we designed two levels of an image processing structure. A distributed file system is taken as the underlying storage architecture to efficiently manage and query massive remote sensing images. The upper layer uses a GPU server to train the remote sensing image classification model to improve the classification accuracy. To improve the classification accuracy, we add two parameters to adjust the data of the current layer in U-Net. The experimental results show that the proposed method based on distributed storage has a high degree of scalability, and it has a short processing time while maintaining a high classification accuracy for remote sensing images. [ABSTRACT FROM AUTHOR]

Dynamic provisioning of computational resources in cloud computing system is a challenging problem. With many VMs being deployed in clouds, managing so much resources to stable work becomes a critical problem. Therefore, how to meet the different quality of service (QoS), we consider the service quality requirements is very important. In this paper, we address this problem. We first establish a QoS scheduling model by incorporating the cloud characteristics, and then, we develop a task scheduling objects to ensure faults can be tolerated during the task exaction. Finally, we proposed a QoS-aware scheduling algorithm, QoS-DPSO, to satisfy the QoS required in cloud computing systems. For the target requirements of QoS requirement, we take the time, reliability and cost as a single object problem. Experimental results show that QoS-DPSO can effectively improve the performance and obtain the high reliability. [ABSTRACT FROM AUTHOR]

Consumers are increasingly influenced by product reviews when purchasing goods or services. At the same time, deceptive reviews usually mislead users. It is inefficient and inaccurate to manually identify deceptive reviews in massive reviews. Therefore, automatically identifying deceptive reviews has become a research trend. Most of existing methods are less effective since they are lack of deeply understanding of reviews. We propose a neural network method with bidirectional long short-term memory (BiLSTM) and feature combination to learn the representation of deceptive reviews. We conduct a large amount of experiments and demonstrate the effectiveness of our proposed method. Specifically, in the mixed-domain detection experiment, the results prove that our model is effective by making comparisons with other neural network-based methods. BiLSTM gives more than 3% improvement in F1 score compared with the most advanced neural network method. Since feature selection plays an important role in this direction, we combine features to improve the performance. Then we get 87.6% F1 value which outperforms the state-of-the-art method. Moreover, in the cross-domain detection experiment, our method achieves 82.4% F1 value which is about 6% higher than the state-of-the-art method on restaurant domain, and it is also robust on doctor domain. [ABSTRACT FROM AUTHOR]

Traffic jam has grown to be a more and more difficult problem to be solved in big cities around the world, and people are getting less chance to take taxis. In order to remit this issue, we propose a recommendation strategy based on taxi traces data for passenger by the roads to make it easier. Considering the scale of taxi traces data, Hadoop is employed to handle the traces data, whose tasks include filtering and cleaning of the data, mapping taxi traces, and computing the average passage time and empty taxi arrival rate on the roads. When a user uploads his position and the time, assisted by weather condition gain from the Internet, we get the very model that corresponds to the date and the weather; the time interval is thought to be the expected waiting time between the moment when user requests the service and the moment when the cumulative number of empty taxi is greater than or equal to 1 after adding the time the taxi spends on the road, and is pushed to the user. The experiment is conducted on the base of a real-world dataset generated by over 12,000 taxis over a period of 3 months in Beijing. Experimental results demonstrate that the processing speed of Hadoop is nine times faster than serial's, which displays the feasibility of Hadoop in the application of massive traces data. In addition, the accuracy of the recommendation strategy reaches up to 88.75 %, and it meets the demand of real-time service. [ABSTRACT FROM AUTHOR]

In view of the problem that existing Media Access Control (MAC) protocols can not provide real-time monitor on network conditions, this paper puts forward a new MAC protocol-Predict and Feedback MAC (PFMAC) which can predict the channel's congestion level reasonably. The dominant idea of the new scheme is to record the channel's busy or idle situation in the backoff stage by sending Sensor Nodes (SNs), and according to the congestion level every SN can change the contention window adaptively when send packets successfully to minimize collisions, saving energy and channel resources. The result of simulation shows that compared with other MAC protocols, the PFMAC protocol can improve network throughput and reduce energy consumption in high speed network environment. [ABSTRACT FROM AUTHOR]