Search

Learning from imbalanced data is a challenging task in a wide range of applications, which attracts significant research efforts from machine learning and data mining community. As a natural approach to this issue, oversampling balances the training samples through replicating existing samples or synthesizing new samples. In general, synthesization outperforms replication by supplying additional information on the minority class. However, the additional information needs to follow the same normal distribution of the training set, which further constrains the new samples within the predefined range of training set. In this paper, we present the Wiener process oversampling (WPO) technique that brings the physics phenomena into sample synthesization. WPO constructs a robust decision region by expanding the attribute ranges in training set while keeping the same normal distribution. The satisfactory performance of WPO can be achieved with much lower computing complexity. In addition, by integrating WPO with ensemble learning, the WPOBoost algorithm outperformsmany prevalent imbalance learning solutions. [ABSTRACT FROM AUTHOR]

Service oriented architecture has been proposed to support collaborations among distributed wireless sensor network (WSN) applications in an open dynamic environment. However, WSNs are resource constraint, and have limited computation abilities, limited communication bandwidth and especially limited energy. Fortunately, sensor nodes in WSNs are usually deployed redundantly, which brings the opportunity to adopt a sleep schedule for balanced energy consumption to extend the network lifetime. Due to miniaturization and energy efficiency, one sensor node can integrate several sense units and support a variety of services. Traditional sleep schedule considers only the constraints from the sensor nodes, can be categorized to a one-layer (i.e., node layer) issue. The service oriented WSNs should resolve the energy optimization issue considering the two-layer constraints, i.e., the sensor nodes layer and service layer. Then, the one-layer energy optimization scheme in previous work is not applicable for service oriented WSNs. Hence, in this paper we propose a sleep schedule with a service coverage guarantee in WSNs. Firstly, by considering the redundancy degree on both the service level and the node level, we can get an accurate redundancy degree of one sensor node. Then, we can adopt fuzzy logic to integrate the redundancy degree, reliability and energy to get a sleep factor. Based on the sleep factor, we furthermore propose the sleep mechanism. The case study and simulation evaluations illustrate the capability of our proposed approach. [ABSTRACT FROM AUTHOR]

P2P collusive piracy, where paid P2P clients share the content with unpaid clients, has drawn significant concerns in recent years. Study on the follow relationship provides an emerging track of research in capturing the followee (e.g., paid client) for the blocking of piracy spread from all his followers (e.g., unpaid clients). Unfortunately, existing research efforts on the follow relationship in online social network have largely overlooked the time constraint and the content feedback in sequential behavior analysis. Hence, how to consider these two characteristics for effective P2P collusive piracy prevention remains an open problem. In this paper, we proposed a multi-bloom filter circle to facilitate the time-constraint storage and query of P2P sequential behaviors. Then, a probabilistic follow with content feedback model to fast discover and quantify the probabilistic follow relationship is further developed, and then, the corresponding approach to piracy prevention is designed. The extensive experimental analysis demonstrates the capability of the proposed approach. [ABSTRACT FROM AUTHOR]

The rise of mobile technologies in recent years has led to large volumes of location information, which are valuable resources for knowledge discovery such as travel patterns mining and traffic analysis. However, location dataset has been confronted with serious privacy concerns because adversaries may re-identify a user and his/her sensitivity information from these datasets with only a little background knowledge. Recently, several privacy-preserving techniques have been proposed to address the problem, but most of them lack a strict privacy notion and can hardly resist the number of possible attacks. This paper proposes a private release algorithm to randomize location dataset in a strict privacy notion, differential privacy, with the goal of preserving users' identities and sensitive information. The algorithm aims to mask the exact locations of each user as well as the frequency that the user visits the locations with a given privacy budget. It includes three privacy-preserving operations: private location clustering shrinks the randomized domain and cluster weight perturbation hides the weights of locations, while private location selection hides the exact locations of a user. Theoretical analysis on privacy and utility confirms an improved trade-off between privacy and utility of released location data. Extensive experiments have been carried out on four real-world datasets, GeoLife, Flickr, Div400 and Instagram. The experimental results further suggest that this private release algorithm can successfully retain the utility of the datasets while preserving users' privacy. [ABSTRACT FROM AUTHOR]

With the motivation of seamlessly extending wireless sensor networks to the external environment, service-oriented architecture comes up as a promising solution. However, as sensor nodes are failure prone, this consequently renders the whole wireless sensor network to seriously faulty. When a particular node is faulty, the service on it should be migrated into those substitute sensor nodes that are in a normal status. Currently, two kinds of approaches exist to identify the substitute sensor nodes: the most common approach is to prepare redundancy nodes, though the involved tasks such as maintaining redundancy nodes, i.e., relocating the new node, lead to an extra burden on the wireless sensor networks. More recently, other approaches without using redundancy nodes are emerging, and they merely select the substitute nodes in a sensor node's perspective i.e., migrating the service of faulty node to it's nearest sensor node, though usually neglecting the requirements of the application level. Even a few work consider the need of the application level, they perform at packets granularity and don't fit well at service granularity. In this paper, we aim to remove these limitations in the wireless sensor network with the service-oriented architecture. Instead of deploying redundancy nodes, the proposed mechanism replaces the faulty sensor node with consideration of the similarity on the application level, as well as on the sensor level. On the application level, we apply the Bloom Filter for its high efficiency and low space costs. While on the sensor level, we design an objective solution via the coefficient of a variation as an evaluation for choosing the substitute on the sensor level. [ABSTRACT FROM AUTHOR]

Wireless sensor networks (WSNs) are widely used in practice for comprehensively monitoring and gathering physical information via a multitude of sensors. As the development of WSNs, the integration of them with the external Internet is a urgent need. By wrapping the sensor functionality as a WSN service, the Web service is considered as the most promising technology to incorporate WSNs into the Internet. The quest for selecting the service with the best performance promotes service ranking technology. However, due to the dynamic WSN environment, traditional quality of service (QoS) based ranking approaches for general Web services are no longer suitable for the WSN service. In this article, in order to fit the characteristics of the WSN environment, we propose a context-aware WSN service ranking approach by aggregating the user rating and WSN service context. First, the User QoS Assessment(UQA) and Context QoS Assessment(CQA) are proposed, respectively. Then, through the performance influence on the WSN service by the variations in their context, a Fuzzy mechanism is further developed to aggregate the UQA and the CQA. Finally, the experiments are presented to confirm the validity of the proposed approach. [ABSTRACT FROM AUTHOR]

Wireless sensor networks ( WSNs) are attractive for monitoring and gathering physical information (e.g. temperature) via lots of deployed sensors. For the applications in WSNs, Web service is one of the recommended frameworks to publish, invoke, and manage services. However, the standard Web service description language ( WSDL), defines only the service input and output while ignoring the corresponding input-to-output mapping relationships. This presents a serious challenge in distinguishing services with similar input and output interface. In this paper, we address this challenge by embedding the service policy into the traditional WSDL2.0 schema to describe the input-to-output mapping relationships. The service policy is then transformed into a policy binary tree so that the similarity between different Web services can be quantitatively evaluated. Furthermore, a new service redundancy detection approach is proposed based on this similarity. Finally, the case study and experimental analysis illustrate the applicability and capability of the proposed service redundancy detection approach. [ABSTRACT FROM AUTHOR]

Reinforcement learning is a core technology for modern artificial intelligence, and it has become a workhorse for AI applications ranging from Atrai Game to Connected and Automated Vehicle System (CAV). Therefore, a reliable RL system is the foundation for the security critical applications in AI, which has attracted a concern that is more critical than ever. However, recent studies discover that the interesting attack mode adversarial attack also be effective when targeting neural network policies in the context of reinforcement learning, which has inspired innovative researches in this direction. Hence, in this paper, we give the very first attempt to conduct a comprehensive survey on adversarial attacks in reinforcement learning under AI security. Moreover, we give briefly introduction on the most representative defense technologies against existing adversarial attacks. [ABSTRACT FROM AUTHOR]