Search

Deep learning technology has developed rapidly in recent years and has been successfully applied in many fields, including face recognition. Face recognition is used in many scenarios nowadays, including security control systems, access control management, health and safety management, employee attendance monitoring, automatic border control, and face scan payment. However, deep learning models are vulnerable to adversarial attacks conducted by perturbing probe images to generate adversarial examples, or using adversarial patches to generate well-designed perturbations in specific regions of the image. Most previous studies on adversarial attacks assume that the attacker hacks into the system and knows the architecture and parameters behind the deep learning model. In other words, the attacked model is a white box. However, this scenario is unrepresentative of most real-world adversarial attacks. Consequently, the present study assumes the face recognition system to be a black box, over which the attacker has no control. A Generative Adversarial Network method is proposed for generating adversarial patches to carry out dodging and impersonation attacks on the targeted face recognition system. The experimental results show that the proposed method yields a higher attack success rate than previous works. [ABSTRACT FROM AUTHOR]

Narrowband Internet of Things (NB-IoT) is one of the low-power wide-area network (LPWAN) technologies that aim to support enormous connections, featuring wide-area coverage, low power consumption, and low costs. NB-IoT could serve a massive number of IoT devices, but with very limited radio resources. Therefore, how to enable a massive number of IoT devices to transmit messages periodically, and with low latency, according to transmission requirements, has become the most crucial issue of NB-IoT. Moreover, IoT devices are designed to minimize power consumption so that the device battery can last for a long time. Similarly, the NB-IoT system must configure different power-saving mechanisms for different types of devices to prolong their battery lives. In this study, we propose a persistent periodic uplink scheduling algorithm (PPUSA) to assist a plethora of Internet of Things (IoT) devices in reporting their sensing data based on their sensing characteristics. PPUSA explicitly considers the power-saving mode and connection suspend/resume procedures to reduce the IoT device's power consumption and processing overhead. PPUSA allocates uplink resource units to IoT devices systematically so that it can support the periodic–uplink transmission of a plethora of IoT devices while maintaining low transmission latency for bursty data. The simulation results show that PPUSA can support up to 600,000 IoT devices when the NB-IoT uplink utilization is 80%. In addition, it takes only one millisecond for the transmission of the bursty messages. [ABSTRACT FROM AUTHOR]

Modern hackers display increasing sophistication. Intrusion detection systems, both network-based and host-based, now utilize machine learning for improved detection of such advanced attacks. While most of these systems rely on a single data source for training, practical scenarios often involve attack features scattered across multiple sources, posing challenges to the system's effectiveness in detection. This impairs their potential for attack detection. Thus, this study assesses three host-based data sources—network traffic, system logs, and host statistics. It evaluates and compares their combined detection capabilities across diverse attack stages and types. In the proposed framework, network traffic data is handled by a Convolutional Neural Network (CNN) for improved automatic feature selection. System log data are processed using Long Short-Term Memory (LSTM) and an attention model to enhance temporal relationship exploration. Host statistics are processed by a Deep Neural Network (DNN) to improve classification performance. Experimental results show that the F1-scores reach 1.0 for all considered attacks and attack stages when all three data sources are utilized in the detection process. Additionally, employing diverse models based on the data type leads to improved results, a fact exemplified by Lin et al. (2022) which exclusively utilized XGBoost. The host statistics were found to be highly effective in detecting attacks and were thus investigated further for different attack methods and attack stages. The results showed that the disk usage percentage (DSK), minor memory faults (MINFLT), major memory faults (MAJFLT), total virtual memory growth during the last interval (VGROW), and total resident memory growth during the last interval (RGROW) were primarily affected by all the attacks in the initial access and command and control stages. By contrast, in the impact attack stage, the affected system resources varied widely depending on the particular attack. [ABSTRACT FROM AUTHOR]

As the widely adopted IEEE 802.15.4 standard exhibited severe drawbacks when used in multi-hop WSN environment, the IEEE Standards Association Board has released a revised version, IEEE 802.15.4e, which includes Time-Slotted Channel Hopping (TSCH) mechanism as the MAC amendment to IEEE 802.15.4. The key feature of IEEE 802.15.4e TSCH mechanism is to allow the sensors to plan the time-slot schedule for efficient data transmission. However, the standard does not specify how to plan a time-slot schedule so that data collected by the sensors can be received by the base station as soon as possible without being delayed by transmission conflicts. In this paper, we propose a constant-time, distributed time-slot scheduling algorithm to address this problem. Existing works on IEEE 802.15.4e TSCH scheduling depend on one or a few nodes to collect traffic-flow information from their descendants and plan a centralized time-slot schedule for the whole network. In contrast, our algorithm allows every sensor node to compute its time-slot schedule in a simple, truly distributed manner. Simulation results show that our algorithm significantly outperforms a state-of-the-art protocol in terms of the packet transmission delay and yields a smaller duty cycle. [ABSTRACT FROM AUTHOR]

With the rapid development of cloud technology, many services have been transferred from local computers to the cloud-based platform, which decreases the amount of computation done on the former. The local computer could thus be developed in the direction of portability and power saving. Graphics processing, apart from providing user interfaces featuring diversified special effects, is also significant in terms of application programs and play interactions. It is exactly on the basis of the concept of graphics processing that cloud-support rendering is developed, which is aimed to improve the graphics efficiency in mobile devices, via the graphics processing units in the cloud-based platform. The cloud-based platform and the mobile devices are usually connected by the Internet; however, as remote rendering might call for greater network bandwidth, its efficiency will be compromised if the network bandwidth is not stable. Given this limitation, this paper sets out to propose a quality-of-service-aware resource allocation strategy for mobile 3D graphics rendering, which is a hybrid rendering technology combining the client-side graphics processing capabilities with the graphics processing units in the cloud-based platform. When network bandwidth is not stable, the technology is able to assess the current network bandwidth, and dynamically configure the rendered frames on the client side and cloud-based platforms. Even when the client side could not access the network, it would still be possible to carry out the drawing through the graphics processing units on the local computer. Three applications are tested in this research: the technology can increase the frame rate by an average of 44.99% when the bandwidth is 10% greater than the minimum limit, by an average of 44.57% when the bandwidth is less than the minimum limit, by an average of 30.86% when the bandwidth is 10% less than the minimum limit, and by an average of 33.74% when the bandwidth is not stable. [ABSTRACT FROM PUBLISHER]

Internet of things (IoT) has been considered as one of the most promising technologies over the next decade. One of the basic requirements of IoT is the global communication connectivity between smart objects. LTE-A has been considered as the main communication channel for connecting devices. For this reason, the machine-type communication (MTC) has been defined in the 3GPP LTE-A specification. With the rapid growth of the IoT devices, MTC in LTE-A faces many challenges. The primary design of a LTE-A network is to support the human-type communication (HTC). However, MTC and HTC have different characteristics, such as packet size, traffic arrival rate, and delay tolerance. How to accommodate a large amount of MTC traffic without affecting the grade of service of the HTC is challenging. More specifically, before accessing the LTE-A network, a random access (RA) procedure needs to be performed by a device to synchronize in the uplink of an eNB. Since the number of MTC devices is expected to be much greater than the HTC devices, without suitable access control, the success rate of the RA requests from the HTC devices can be significantly degraded. Therefore, this research aims to design an appropriate RA procedure to relieve this problem. The authors propose an Adaptive RACH Resource Allocation (ARRA) which integrates several control schemes, including a Resource Allocation scheme, an Access Class Barring scheme, and a Priority Device Setting scheme. Simulation results show that the proposed ARRA is able to achieve a higher access success rate and a lower latency for HTC devices while providing a different quality of service to different types of MTC devices. [ABSTRACT FROM AUTHOR]

Downlink broadcast in LTE-Advanced-(LTE-A) and WiMAX-based relay networks is a crucial service for multimedia delivery. Currently, most research effort goes into devising efficient resource allocation mechanisms to achieve more efficient resource utilization. However, spatial reuse, which is an important technique in improving transmission capacity, has received little attention in the literature. Thus, in this work, we investigate ways of achieving more efficient resource allocation in wireless relay networks via spatial reuse. We first formulate a joint spatial reuse and resource allocation problem as an integer linear programming (ILP) model. We then consider a grouping mechanism in which relay stations (RSs) are grouped together if they do not interfere with each other's transmission signal. RSs in the same group can thus utilize spatial reuse by using the same set of resources to broadcast data. Because of the high computational complexity of the ILP model, we propose a two-phase heuristic solution. In the first phase, the enhanced-resource diminishing principle approach is employed to determine the number of resources required by the base station and a set of selected RSs. In the second phase, a max-coloring algorithm is employed to organize the selected RSs into broadcast groups and then assign the required resources to each group by exploring the maximum advantage of spatial reuse. The simulation results show that the proposed solution improves system performance up to 61%, as compared with two existing mechanisms. [ABSTRACT FROM AUTHOR]

Virtualization is a core 5G network technology which helps telecom companies significantly reduce capital expenditure and operating expenses by deploying multiple services on the same hardware infrastructure. However, providing QoS-guaranteed services for multi-tenants poses a significant challenge due to multi-tenant service diversity. Network slicing has been proposed as a means of addressing this problem by isolating computing and communication resources for the different tenants of different services. However, optimizing the allocation of the network and computation resources across multiple network slices is a critical but extremely difficult problem. Accordingly, this study proposes two heuristic algorithms, namely Minimum Cost Resource Allocation (MCRA) and Fast Latency Decrease Resource Allocation (FLDRA), to perform dynamic path routing and resource allocation for multi-tenant network slices in a two-tier architecture. The simulation results show that both algorithms significantly outperform the Upper-tier First with Latency-bounded Overprovisioning Prevention (UFLOP) algorithm proposed in previous work. Furthermore, the MCRA algorithm achieves a higher resource utilization than the FLDRA algorithm. [ABSTRACT FROM AUTHOR]

With increasing interest in the concept of 5G wireless networks and the popularization of mobile devices, users gradually watch videos through mobile devices in streaming mode rather than off-line mode. However, the latency and lags of mobile networks will reduce the quality of service. In HTTP live streaming services, when there are multiple sources for the same content stored in the streaming servers, the suited quality content is able to be selected for playing according to the networking bandwidth conditions between servers and user equipment, and the users can obtain appropriate streaming quality automatically. However, the selection policy is difficult to be approved between 5G wireless communication networks and SDNs. In this study a buffer-aware HTTP live streaming approach is proposed, which evaluates the weights of media segments to decide the transmitting priorities based on the current playing time and adjust the appropriate transmission path through the decision making network controller according to the utilization and stability of the routers and switches in SDN. Finally, the experimental results proved that the approach is able to correct the overall streaming source sequence with the buffer status to effectively upgrade overall streaming service quality, as well as maintain a certain level of streaming quality for SDN-enable 5G wireless networks. [ABSTRACT FROM PUBLISHER]

Open source is playing an increasingly important role in the software and communications industries. In particular, open source implementations are widely available for most Internet protocols, from kernel-resident Ethernet drivers, WLAN drivers, and TCP/IP stacks (e.g., in Linux, BSD, and Android operating systems) up to user-space networked applications (e.g., web clients and proxies). Open sourced applications are arguably becoming more prevalent than proprietary closed software. Hundreds of open source networking packages have played important roles in our daily life. The same open source spirit also goes to some medium access control (MAC)/physical (PHY) hardware components and the development platforms for simulation, rapid prototyping, software defined radio, and software defined networking. [ABSTRACT FROM PUBLISHER]

SUMMARY Following recent advances in the performance of ad hoc networks, the limited life of batteries in mobile devices poses a bottleneck in their development. Consequently, how to minimize power consumption in the Medium Access Control (MAC) layer of ad hoc networks is an essential issue. The power-saving mode (PSM) of IEEE 802.11 involves the Timing Synchronization Function to reduce power consumption for single-hop mobile ad hoc networks (MANETs). However, the IEEE 802.11 PSM is known to result in unnecessary energy consumption as well as the problems of overheating and back-off time delay. Hence, this study presents an efficient power-saving MAC protocol, called p-MANET, based on a Multi-hop Time Synchronization Protocol, which involves a hibernation mechanism, a beacon inhibition mechanism, and a low-latency next-hop selection mechanism for general-purpose multi-hop MANETs. The main purposes of the p-MANET protocol are to reduce significantly the power consumption and the transmission latency. In the hibernation mechanism, each p-MANET node needs only to wake up during one out of every N beacon interval, where N is the number of beacon intervals in a cycle. Thus, efficient power consumption is achieved. Furthermore, a beacon inhibition mechanism is proposed to prevent the beacon storm problem that is caused by synchronization and neighbor discovery messages. Finally, the low-latency next-hop selection mechanism is designed to yield low transmission latency. Each p-MANET node is aware of the active beacon intervals of its neighbors by using a hash function, such that it can easily forward packets to a neighbor in active mode or with the least remaining time to wake up. As a consequence, upper-layer routing protocols can cooperate with p-MANET to select the next-hop neighbor with the best forwarding delay. To verify the proposed design and demonstrate the favorable performance of the proposed p-MANET, we present the theoretical analysis related to p-MANET and also perform experimental simulations. The numerical results show that p-MANET reduces power consumption and routing latency and performs well in extending lifetime with a small neighbor discovery time. Copyright © 2011 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

The emergence of Web 2.0 technology provides more opportunities to foster online communication and sharing in an e-learning environment. The purpose of this study was to develop a Web 2.0 annotation system, MyNote, based on the Web 2.0 core concepts which emphasize ease of access and active sharing and then to gain an understanding about people’s perceptions of MyNote from a usability perspective. MyNote was employed on multimedia learning objects in a Learning Management System (LMS), and out of the LMS as well in this study. The evaluation results showed that, with factor analysis, interactivity, usefulness, helpfulness, and willingness for future use were categorized to represent the perceptions of MyNote. It was also found that the factors of interactivity and helpfulness were statistically significant to predict the future use of MyNote. Lastly, the habit of taking notes also affected learners’ perceptions of using MyNote. [Copyright &y& Elsevier]

Abstract: The increased prevalence of digital devices with communication capability heralds the era of ubiquitous computing, as predicted by Mark Weiser. Ubiquitous computing aims to provide users with intelligent human-centric context-aware services at anytime anywhere. Optimal path planning in a ubiquitous network considers the needs of users and the surrounding context. This approach is very different from that applied by existing research on car navigation and mobile robots. This study proposes a context-aware path planning mechanism based on spatial conceptual map (SCM) and genetic algorithm (GA), referred to as UbiPaPaGo. The SCM model is adopted to represent the real map of the surrounding environment. The optimal path is planned using a GA, which is a robust metaheuristic algorithm. UbiPaPaGo attempts to automatically find the best path that satisfies the requirements of an individual user. A prototype of UbiPaPaGo is implemented to demonstrate its feasibility and scalability. Experimental results validate the effectiveness and the efficiency of UbiPaPaGo in finding the optimal path. [ABSTRACT FROM AUTHOR]

Next-generation advanced driver-assistance systems (ADAS) and cooperative adaptive cruise control (CACC) for advanced/autonomous driving are expected to increasingly use wireless connectivity such as V2V and V2I to improve the coverage, particularly in the locations where a vehicle's camera or radar is ineffective. However, using shared sensing data raises grave concerns about the truthfulness of information reported by unreliable stakeholders. For example, a transmitting vehicle may deliberately disseminate false locations to the surrounding receivers. Trusting the data, the automatic control systems in such connected receivers can be trapped to change to a wrong lane or accelerate unexpectedly, and then potentially lead to a crash. This work introduces a novel approach to support a host vehicle in verifying the motion behavior of a target vehicle and then the truthfulness of sharing data in cooperative vehicular communications. Initially, at the host vehicle, the detection system recreates the motion behavior of the target vehicle by extracting the positioning information from the V2V received messages. Furthermore, the next states of that vehicle are predicted based on the unscented Kalman filter. Unlike prior studies, the checkpoints of the predicted trajectory in the update stage are periodically corrected with a new reliable measurement source, namely 5 G V2V multi-array beamforming localization. If there is any inconsistency between the estimated position and the corresponding reported one from V2V, the target vehicle will be classified as an abnormal one. The simulation results demonstrate that our method can achieve accuracy over 0.97 in detecting abnormal reports, including those from collusion and Sybil attacks. [ABSTRACT FROM AUTHOR]

In the past, most of research directly assess the level of creativity from the individual's response test, rather than consider whether the current emotion of the examinee affects their creative performance for individual creativity studying. This study discusses with the emotional reactions of people in creative thinking and the relationship between emotion and creativity. In the meantime, the research will conduct tests after they have introduced the creative thinking method or not, and obtain the relevant emotional data through the physiological sensing instrument to understand the emotional state of the tested person for creative thinking, and then score the creative thinking result. The results show that students can have better imagination when they think about creativity in positive emotions and it can help the teacher to improve as a course in the future. • This study discusses with the emotional reactions of people in creative thinking. • This study uses the Support Vector Machine for the classification of emotions. • The results show that emotions change with creatively thinking. • Emotions are proved to help strengthen their creativity for students. [ABSTRACT FROM AUTHOR]

Distributed denial of service is one of the most critical threats to the availability of Internet services. A botnet with only 0.01 percent of the 50 billion connected devices in the Internet of Things is sufficient to launch a massive DDoS flooding attack that could exhaust resources and interrupt any target. However, the mobility of user equipment and the distinctive characteristics of traffic behavior in mobile networks also limit the detection capabilities of traditional anti-DDoS techniques. In this article, we present a novel collaborative DDoS defense architecture called MECPASS to mitigate the attack traffic from mobile devices. Our design involves two filtering hierarchies. First, filters at edge computing servers (i.e., local nodes) seek to prevent spoofing attacks and anomalous traffic near sources as much as possible. Second, global analyzers located at cloud servers (i.e., central nodes) classify the traffic of the entire monitored network and unveil suspicious behaviors by periodically aggregating data from the local nodes. We have explored the effectiveness of our system on various types of application- layer DDoS attacks in the context of web servers. The simulation results show that MECPASS can effectively defend and clean an Internet service provider core network from the junk traffic of compromised UEs, while maintaining the false-positive rate of its detection engine at less than 1 percent. [ABSTRACT FROM AUTHOR]

As cloud computing and wearable devices technologies mature, relevant services have grown more and more popular in recent years. The healthcare field is one of the popular services for this technology that adopts wearable devices to sense signals of negative physiological events, and to notify users. The development and implementation of long-term healthcare monitoring that can prevent or quickly respond to the occurrence of disease and accidents present an interesting challenge for computing power and energy limits. This study proposed an adaptive sensor data segments selection method for wearable health care services, and considered the sensing frequency of the various signals from human body, as well as the data transmission among the devices. The healthcare service regulates the sensing frequency of devices by considering the overall cloud computing environment and the sensing variations of wearable health care services. The experimental results show that the proposed service can effectively transmit the sensing data and prolong the overall lifetime of health care services. [ABSTRACT FROM AUTHOR]

SUMMARY Cloud computing has emerged as a promising technique to provide storage and computing component on-demand services over a network. In this paper, we present an energy-saving algorithm using the Kalman filter for cloud resource management to predict the workload and to further achieve high resource availability with low service level agreement. Using the proposed algorithm, one can estimate the potential future workload trend then predict the computing component workload utilizations and further retrench energy consumption and achieve load balancing in a cloud system. Experimental results show that the proposed algorithm achieves more than 92.22% accuracy in the computing component workload prediction, improves 55.11% energy in energy consumption, and has 3.71% in power prediction error rate, respectively. Copyright © 2013 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

In this paper, a load balancing mechanism based on cell breathing was proposed for the hybrid EPON-WiMAX network. The hybrid EPON-WiMAX network integrates the emerging Ethernet Passive Optical Network (EPON) and Worldwide Interoperability for Microwave Access (WiMAX) technologies to support high bandwidth, broadband access, mobility, and Quality of Service (QoS) guarantee. Load balancing is required due to uneven distribution of traffic among different WiMAX base stations (BS) as well as mobility of WiMAX Subscriber Station (SS). Cell breathing is a well-known load balancing technique adopted in wireless networks. By changing the transmission powers of the base stations, the cell breathing technique is able to adjust the traffic, both uplink and downlink, load of base stations. In the integrated EPON-WiMAX network, transmission powers of WiMAX base stations can be optimally allocated at the EPON optical networking unit (ONU). In this paper, we formulate the cell breathing-based load balancing problem into a linear programming problem. Our goal is to find the best power adjustment that maximizes system throughput. To alleviate the time complexity of solving the linear programming problem, we propose a Heaviest Load First Algorithm (HLFA) to obtain the near optimal solution. The solution obtained by the HLFA algorithm will not waste any transmission power to make the entire system energy-efficiency. We demonstrate the performance of HLFA via extensive simulations. The simulation results show that the HLFA can provide the best solution to achieve load balancing and enhance the system throughput as compared to existing solutions. Furthermore, HLFA maintains good performance even when the system has multiple overloaded ONU-BSs. [ABSTRACT FROM AUTHOR]

The next generation wireless communication system will likely be heterogeneous networks, as various technologies can be integrated on heterogeneous networks. A mobile multiple-mode device can easily access the Internet through different wireless interfaces. The mobile multiple-mode device thus could switch to different access points to maintain the robustness of the connection when it can acquire more resources from other heterogeneous wireless networks. The mobile multiple-mode device therefore needs to face the handover problem in such environment. This work introduces Session Initiation Protocol (SIP)-based cross-layer scheme to support seamless handover scheme over heterogeneous networks. The proposed scheme consists of a battery lifetime-based handover policy and cross-layer fast handover scheme, called the SIP-based mobile stream control transmission protocol (SmSCTP). This work describes the major idea of the proposed scheme and infrastructure. The proposed scheme has been implemented in Linux system. The simulation and numerical results demonstrate that the proposed SmSCTP scheme yields better signaling cost, hand-off delay time, packet loss and delay jitter than SIP and mSCTP protocols. Copyright © 2010 John Wiley & Sons, Ltd. This work introduces Session Initiation Protocol (SIP)-based cross-layer scheme to support seamless handover scheme over heterogeneous networks. The proposed scheme consists of a battery lifetime-based handover policy and cross-layer fast handover scheme, called the SIP-based mobile stream control transmission protocol (SmSCTP). This work describes the major idea of the proposed scheme and infrastructure. The proposed scheme has been implemented in Linux system. The simulation and numerical results demonstrate that the proposed SmSCTP scheme yields better signaling cost, hand-off delay time, packet loss and delay jitter than SIP and mSCTP protocols. Copyright © 2010 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Recently, most of Voice over Internet applications are based on the Session Initiation Protocol (SIP). This paper addresses issues of allowing a user to continue to communicate with a remote party while changing terminals over multiple devices. Specifically, we propose an SIP extension header to improve the Call Transfer mechanism and hide the changing of the terminal from the remote party. We also propose a mechanism to solve the problem of the user needing to terminate all devices separately when a session is split over multiple devices. Finally, the proposed mechanisms are implemented using Sip-Communicator, an open source of SIP. [Copyright &y& Elsevier]

This work describes a novel location-aware, self-organizing, fault-tolerant peer-to-peer (P2P) overlay network, referred to as Laptop. Network locality-aware considerations are a very important metric for designing a P2P overlay network. Several network proximity schemes have been proposed to enhance the routing efficiency of existing DHT-based overlay networks. However, these schemes have some drawbacks such as high overlay network and routing table maintenance overhead, or not being completely self-organizing. As a result, they may result in poor scalability as the number of nodes in the system grows. Laptop constructs a location-aware overlay network without pre-determined landmarks and adopts a routing cache scheme to avoid maintaining the routing table periodically. In addition, Laptop significantly reduces the overlay maintenance overhead by making each node maintain only the connectivity between parent and itself. Mathematical analysis and simulations are conducted to evaluate the efficiency, scalability, and robustness of Laptop. Our mathematical analysis shows that the routing path length is bounded by logdN, and the joining and leaving overhead is bounded by d logdN, where N is the number of nodes in the system, and d is the maximum degree of each node on the overlay tree. Our simulation results show that the average latency stretch is 1.6 and the average routing path length is only about three in 10 000 Laptop nodes, and the maximum degree of a node is bounded by 32. Copyright © 2006 John Wiley & Sons, Ltd. [ABSTRACT FROM AUTHOR]

Network intrusion detection systems (NIDSs) now adopt machine learning (ML) for detection of wide attack variants. However, ML is also known vulnerable to adversarial attacks, which can degrade the accuracy of ML. A number of defense strategies have been proposed but mostly in image classification areas. In this work, we propose Ensemble Learning with Adversarial Training (ELAT) to combine adversarial training and ensemble learning into a solution. We compare four approaches: single, ensemble, adversarial and ELAT. In the experiments, several models were developed and tested using different approaches to know which method is robust against adversarial attacks for ML-based NIDSs. The average F1 score for the single models was 0.93 within a wide range (0.82-0.99), but dropped to 0.29 when facing adversarial attacks, particularly dropped to 0.07 caused by the strongest attack, Projected Gradient Descent (PGD). With ensemble, adversarial and ELAT, the average scores were recovered to 0.80, 0.88 and 0.91, respectively. In addition, this work involves prediction of the models and approach implemented behind the system using cosine similarity with an accuracy of 99.9%. [ABSTRACT FROM AUTHOR]

This work compares the performance of different combinations of data sources for intrusion detection in depth. To learn and distinguish between normal and malicious behavior, we use machine learning algorithms and train three typical models on three kinds of datasets: system logs, packet flows and host statistics. Unlike other studies, our study captures and monitors the behavior from multiple data sources in order to catch security attacks. Our aim is to figure out how to build the most effective dataset for machine learning with a combination of multiple sources. However, since there are no such datasets which have been generated from multiple sources for given attacks, we show how to build and generate a dataset with three data sources. We then compare the F1 score of the detection by applying machine learning algorithms for various combinations of the data sources. Our evaluation results show that the dataset of host statistics results in better performance (0.91) than traffic flows (0.63) and system logs (0.44) because it has the highest average F1-score in the three stages of attacks, while the other datasets may have poor F1-scores in some of the stages, particularly in the stage of impact. However, in the initial access stage of attacks, the dataset of logs performs the best (0.94), and the packet flows are suitable for detecting network DoS attacks (0.82). Furthermore, running this detection with all three data sources results in minor overheads of at most 2.1% CPU utilization. Finally, we analyze the important features of each model, such as the number of logs generated by apache-access, in.telnetd and postfix in the dataset of logs, SrcBytes and TotBytes in the dataset of flows, and MINFLT, VSTEXT and RSIZE in the dataset of statistics. [ABSTRACT FROM AUTHOR]

Abstract: Timely information refers to information whose ‘most recent’ or ‘latest’ instance is most valuable. In mobile ad hoc networks (MANETs), multiple instances of a piece of timely information may be produced by different nodes at different points in time. The problem is to discover the ‘latest’ instance among all existing instances. Within the context of MANETs, timely information discovery is fundamentally different from the existing resource/service discovery problem whose goal is to discover either any instance or a subset of instances which satisfy a local query constraint that can be specified and evaluated using only local attributes of each individual node. In contrast, the timely information discovery problem imposes the global (timeliness) constraint which should best be evaluated when all the instances are considered to determine the latest instance. The complication of discovering timely information arises from the existence of multiple instances of the information, which are produced at different points in time by different nodes in the network, and the need to collect all these instances to decide the latest instance. For MANETs, the lack of infrastructure supports, frequent topology changes, and potential packet loss in wireless communications further challenge the problem of timely information discovery. This paper describes a self-organizing, peer-to-peer based approach, termed ALADIN, to discovering timely information in MANETs. In ALADIN, nodes that produce instances of the timely information are peers who self-organize an adaptive and distributed ‘search infrastructure’ to facilitate the discovery of the latest instance. A simulation study shows that ALADIN is scalable without incurring network-wide flooding in the case of large-scale networks and popular timely information, and yields a high chance of discovering the latest instance in the presence of mobility. [Copyright &y& Elsevier]