Search

Autonomous vehicles (AVs) generate a massive amount of multi-modal data that once collected and processed through Machine Learning algorithms, enable AI-based services at the Edge. In fact, only a subset of the collected data present informative attributes to be exploited at the Edge. Therefore, extracting such a subset is of utmost importance to limit computation and communication workloads. Doing that in a distributed manner imposes the AVs to cooperate in finding an agreement on which attributes should be sent to the Edge. In this work, we address such a problem by proposing a federated feature selection (FFS) algorithm where the AVs collaborate to filter out, iteratively, the less relevant attributes in a distributed manner, without any exchange of raw data, thought two different components: a Mutual-Information-based feature selection algorithm run by the AVs and a novel aggregation function based on the Bayes theorem executed on the Edge. The FFS algorithm has been tested on two reference datasets: MAV with images and inertial measurements of a monitored vehicle, WESAD with a collection of samples from biophysical sensors to monitor a relative passenger. The numerical results show that the AVs converge to a minimum achievable subset of features with both the datasets, i.e., 24 out of 2166 (99%) in MAV and 4 out of 8 (50%) in WESAD, respectively, preserving the informative content of data. [ABSTRACT FROM AUTHOR]

This study considers the simultaneous optimization problem of the transmit beamforming at the access point and the reflecting beamforming at the intelligent reflecting surface (IRS) in an IRS-assisted multiuser downlink system. This joint optimization problem is nonconvex and challenging due to the intricate coupling of transmit and reflecting beamforming variables and the highly nonconvex constant modulus constraints. All the known solutions apply the alternating optimization framework to decouple the joint optimization problem into two problems. Then, these problems are optimized individually with different optimization methods in an alternating manner. However, such an alternative procedure may cause performance degradation. On the basis of the cross-entropy (CE) framework, which was developed initially for machine learning applications, we propose a machine learning-inspired algorithmic framework for simultaneously optimizing transmit and reflecting beamforming in an IRS-assisted wireless system. Extensive simulation results reveal that the proposed CE-based algorithms significantly perform better than the state-of-the-art test algorithms for various system configurations. [ABSTRACT FROM AUTHOR]

Underwater acoustic sensor networks (UASNs) have been widely applied in underwater scenarios where numerous anchoring or floating sensor nodes collaborate in performing some specific assignments, such as information collection or data transmission. In recent years, the trust model has been known as an important tool in responding to attackers inside the network. Although there are a number of trust models recently been suggested as effective methods for terrestrial networks, it is infeasible to directly apply these trust models in UASNs due to the complex underwater environment and the unreliable underwater acoustic communication. For the purpose of achieving accurate and robust trust evaluation for UASNs, a synergetic trust model based on SVM (STMS) is proposed in this paper. The network is divided into a certain number of interconnected clusters in which cluster heads and cluster members (CMs) are synergetic to perform functions. The STMS is mainly comprised of three parts. In the first part, three kinds of trust evidences, which are refined elaborately to reflect most of the attack results, are generated by CMs. In the second part, the support vector machine (SVM) technology is adopted in training a trust prediction model for evaluating accurate trust value. Furthermore, the mechanism of double cluster heads is presented to improve network security and lifetime in the third part. Simulation results demonstrate that STMS performs better than other related works in the sparse deployment environment, which is reflected in the aspect of detect accuracy of malicious nodes, success rate of communication and network lifetime. [ABSTRACT FROM AUTHOR]

In this paper, we propose a machine learning based handover decision algorithm for high-speed railway systems. In a standard LTE-R network, a mobile relay (MR) on a train connects a DeNB and various onboard devices, including train controllers. The standard LTE-R handover decision scheme assumes the moving node may move back and forth. Thus, it makes the serving DeNB keep the inferior connection with the MR during the predefined Time-To-Triger, even though the MR gets closer to the cell boundary and can make a better connection with the neighbor DeNB. However, a train keeps moving forward and never moves backward during the operation. Thus, it is expected that improved quality of wireless connection can be attained if the serving DeNB starts the handover of the MR earlier than the standard by predicting the cell boundary crossing time. Especially, the Gaussian Bayesian regression model outputs the predictive distribution that contains prediction values as well as information of their uncertainty. Thus, we can decide whether to accept the prediction according to the predictive distribution. Based on the prediction, the serving DeNB decides the handover initiation time to make the MR cross the cell boundary during the handover execution. We evaluate the performance of handover decision algorithm through simulations. The simulation results show that the proposed scheme outperforms the standard handover decision algorithm and the H2-based scheme in terms of communication reliability at cell boundary regardless of the moving speed. [ABSTRACT FROM AUTHOR]

Intelligent transportation is an emerging technology that integrates advanced sensors, network communication, data processing, and automatic control technologies to provide great convenience for our daily lives. With the increasing popularity of intelligent transportation, its security issues have also attracted much attention. SQL injection attack is one of the most common attacks in the intelligent transportation system. It has characteristics of various types, fast mutations, hidden attacks, etc., and leads to great harm. Most of the current SQL detection methods are based on manually defined features. The detection results are heavily dependent on the accuracy of feature extraction, so it cannot cope with the increasingly complex SQL injection attacks in the intelligent transportation system. In order to solve this problem, this paper proposes a long short-term memory based SQL injection attack detection method, which can automatically learn the effective representation of data, and has a strong advantage to confront with complex high-dimensional massive data. In addition, this paper proposes an injection sample generation method based on data transmission channel from the perspective of penetration. This method can formally model SQL injection attack and generate valid positive samples. It can effectively solve the over-fitting problem caused by insufficient positive samples. The experimental results show that the proposed method improves the accuracy of the SQL injection attack detection and reduces the false positive rate, which is better than several related classical machine learning algorithms and commonly used deep learning algorithms. [ABSTRACT FROM AUTHOR]

The performance of orthogonal frequency division multiplexing (OFDM) based wireless vehicular communication systems is faced with the great challenge of impulsive noise (IN), which could limit the application of OFDM in ultra-reliable low-latency communication scenarios. In this paper, the challenge of IN elimination for OFDM-based wireless systems is efficiently overcome by the proposed sparse learning algorithms and probabilistic framework inspired by the emerging machine learning theories. For the first time, the sparse machine learning theory is introduced to IN recovery and elimination. Exploiting the measurement vector of IN observed from the reserved null sub-carriers as the input, a novel sparse machine learning based algorithm of sparse cross-entropy minimization is proposed, in which the probability distribution of the IN support is iteratively updated by minimizing the loss function, i.e. the cross-entropy. The proposed algorithm is able to effectively and efficiently learn the sparse pattern and converge to the accurate distribution of IN support. To facilitate an accelerated and even more efficient learning process, regularization is imposed on the loss function by adding a weighting parameter in favor of the accurate distribution. The computer simulation results confirm that the proposed scheme outperforms conventional methods while utilizing fewer spectrum resources over wireless vehicular channels. [ABSTRACT FROM AUTHOR]

In this paper, a new indoor localization algorithm with received signal strength indicators (RSSI) fingerprints by a kernel-based learning technique is proposed. In the offline phase, after the training set pre-processing by the iterative self-organizing data analysis techniques algorithm (ISODATA), the measurement-label set is utilized for classification learning by the c-support vector classification approach. Moreover, each measurement-position training subset is chosen for regressing learning with hybrid kernel and cross-validation techniques. In the online phase, based on the classification result of the received RSSI measurement, the corresponding regression function is chosen for target position estimation. Compared with existing learning techniques, it can increase the training ability of a localization system dramatically. The field tests show that, without adding any extra computation complexity, the proposed algorithm can obtain more accurate position estimation than other existing learning-based localization approaches achieve. [ABSTRACT FROM AUTHOR]

In this paper, we investigate the problem of multiuser computation offloading for cloudlet-based mobile cloud computing in a multichannel wireless contention environment. The studied system is fully distributed so that each mobile device user can make the offloading decisions based only on its individual information, and without information exchange. We first formulate this multiuser computation offloading decision making problem as a noncooperative game. After analyzing the structural property of the formulated game, we show that it is an exact potential game, and has at least one pure-strategy Nash equilibrium point (NEP). To achieve the NEPs in a fully distributed environment, we propose a fully distributed computation offloading (FDCO) algorithm based on machine learning technology. We then theoretically analyze the performance of the proposed FDCO algorithm in terms of the number of beneficial cloudlet computing mobile devices and the system-wide execution cost. Finally, simulation results validate the effectiveness of our proposed algorithm compared with counterparts. [ABSTRACT FROM PUBLISHER]

In this paper, we analyze the spectrum occupancy in cognitive radio networks (CRNs) using different machine learning techniques. Both supervised techniques [naive Bayesian classifier (NBC), decision trees (DT), support vector machine (SVM), linear regression (LR)] and unsupervised algorithms [hidden Markov model (HMM)] are studied to find the best technique with the highest classification accuracy (CA). A detailed comparison of the supervised and unsupervised algorithms in terms of the computational time and the CA is performed. The classified occupancy status is further utilized to evaluate the blocking probability of secondary user for future time slots, which can be used by system designers to define spectrum-allocation and spectrum-sharing policies. Numerical results show that SVM is the best algorithm among all the supervised and unsupervised classifiers. Based on this, we proposed a new SVM algorithm by combining it with a firefly algorithm (FFA), which is shown to outperform all the other algorithms. [ABSTRACT FROM PUBLISHER]

In this paper, we address the problem of reconstructing coverage maps from path-loss measurements in cellular networks. We propose and evaluate two kernel-based adaptive online algorithms as an alternative to typical offline methods. The proposed algorithms are application-tailored extensions of powerful iterative methods such as the adaptive projected subgradient method (APSM) and a state-of-the-art adaptive multikernel method. Assuming that the moving trajectories of users are available, it is shown how side information can be incorporated in the algorithms to improve their convergence performance and the quality of the estimation. The complexity is significantly reduced by imposing sparsity awareness in the sense that the algorithms exploit the compressibility of the measurement data to reduce the amount of data that is saved and processed. Finally, we present extensive simulations based on realistic data to show that our algorithms provide fast and robust estimates of coverage maps in real-world scenarios. Envisioned applications include path-loss prediction along trajectories of mobile users as a building block for anticipatory buffering or traffic offloading. [ABSTRACT FROM AUTHOR]

In this paper, we present a novel cell outage management (COM) framework for heterogeneous networks with split control and data planes—a candidate architecture for meeting future capacity, quality-of-service, and energy efficiency demands. In such an architecture, the control and data functionalities are not necessarily handled by the same node. The control base stations (BSs) manage the transmission of control information and user equipment (UE) mobility, whereas the data BSs handle UE data. An implication of this split architecture is that an outage to a BS in one plane has to be compensated by other BSs in the same plane. Our COM framework addresses this challenge by incorporating two distinct cell outage detection (COD) algorithms to cope with the idiosyncrasies of both data and control planes. The COD algorithm for control cells leverages the relatively larger number of UEs in the control cell to gather large-scale minimization-of-drive-test report data and detects an outage by applying machine learning and anomaly detection techniques. To improve outage detection accuracy, we also investigate and compare the performance of two anomaly-detecting algorithms, i.e., $k$-nearest-neighbor- and local-outlier-factor-based anomaly detectors, within the control COD. On the other hand, for data cell COD, we propose a heuristic Grey-prediction-based approach, which can work with the small number of UE in the data cell, by exploiting the fact that the control BS manages UE-data BS connectivity and by receiving a periodic update of the received signal reference power statistic between the UEs and data BSs in its coverage. The detection accuracy of the heuristic data COD algorithm is further improved by exploiting the Fourier series of the residual error that is inherent to a Grey prediction model. Our COM framework integrates these two COD algorithms with a cell outage compensation (COC) algorithm that can be applied to both planes. Our COC solution utilizes an actor-critic-based reinforcement learning algorithm, which optimizes the capacity and coverage of the identified outage zone in a plane, by adjusting the antenna gain and transmission power of the surrounding BSs in that plane. The simulation results show that the proposed framework can detect both data and control cell outage and compensate for the detected outage in a reliable manner. [ABSTRACT FROM AUTHOR]

Cognitive radio (CR) has been proposed to improve spectrum efficiency by taking advantage of the vacancies in primary channels. Since the frequency range of operation is very wide in a CR network (CRN) and, usually, CRs cannot scan all the channels simultaneously, one of the fundamental tasks for a CR is the channel selection strategy, which directly impacts its performance. In this paper, we present a distributed polynomial time algorithmic framework for computing channel strategies in a CRN with no assumption on the distribution followed by the primary users' channel occupancy. For a secondary user (SU), the upper bound on the gap between the expected profit obtained at each time slot by using the global optimal strategy and the expected profit by using our algorithm is guaranteed to be arbitrarily small when the time horizon is sufficiently large. We also prove an upper bound on the gap between the expected profit by using any strategy sequence and the expected profit by using our strategy sequence. [ABSTRACT FROM PUBLISHER]