Federated Learning with Pareto Optimality for Resource Efficiency and Fast Model Convergence in Mobile Environments.

Federated learning (FL) is an emerging distributed learning technique through which models can be trained using the data collected by user devices in resource-constrained situations while protecting user privacy. However, FL has three main limitations: First, the parameter server (PS), which aggregates the local models that are trained using local user data, is typically far from users. The large distance may burden the path links between the PS and local nodes, thereby increasing the consumption of the network and computing resources. Second, user device resources are limited, but this aspect is not considered in the training of the local model and transmission of the model parameters. Third, the PS-side links tend to become highly loaded as the number of participating clients increases. The links become congested owing to the large size of model parameters. In this study, we propose a resource-efficient FL scheme. We follow the Pareto optimality concept with the biased client selection to limit client participation, thereby ensuring efficient resource consumption and rapid model convergence. In addition, we propose a hierarchical structure with location-based clustering for device-to-device communication using k-means clustering. Simulation results show that with prate at 0.75, the proposed scheme effectively reduced transmitted and received network traffic by 75.89% and 78.77%, respectively, compared to the FedAvg method. It also achieves faster model convergence compared to other FL mechanisms, such as FedAvg and D2D-FedAvg.

The influence of contralateral circulation on computational fluid dynamics of intracranial arteries: simulated versus measured flow velocities.

BACKGROUND: This study aimed to retrospectively evaluate the influence of contralateral anterior circulation on computational fluid dynamics (CFD) of intracranial arteries, by comparing the CFD values of flow velocities in unilateral anterior circulation with the measured values from phase-contrast magnetic resonance angiography (PC-MRA). METHODS: We analyzed 21 unilateral anterior circulation models without proximal stenosis from 15 patients who performed both time-of-flight MRA (TOF-MRA) and PC-MRA. CFD was performed with the inflow boundary condition of a pulsatile flow of the internal carotid artery (ICA) obtained from PC-MRA. The outflow boundary condition was given as atmospheric pressure. Simulated flow velocities of the middle cerebral artery (MCA) and anterior cerebral artery (ACA) from CFD were compared with the measured values from PC-MRA. RESULTS: The velocities of MCA were shown to be more accurately simulated on CFD than those of ACA (Spearman correlation coefficient 0.773 and 0.282, respectively). In four models with severe stenosis or occlusion of the contralateral ICA, the CFD values of ACA velocities were significantly lower (< 50%) than those measured with PC-MRA. ACA velocities were relatively accurately simulated in the models including similar diameters of both ACAs. CONCLUSION: It may be necessary to consider the flow condition of the contralateral anterior circulation in CFD of intracranial arteries, especially in the ACA. RELEVANCE STATEMENT: Incorporating the flow conditions of the contralateral circulation is of clinical importance for an accurate prediction of a rupture risk in Acom aneurysms as the bidirectional flow and accurate velocity of both ACAs can significantly impact the CFD results. KEY POINTS: • CFD simulations using unilateral vascular models were relatively accurate for MCA. • Contralateral ICA steno-occlusion resulted in an underestimation of CFD velocity in ACA. • Contralateral flow may need to be considered in CFD simulations of ACA.

Trust-Based Intelligent Routing Protocol with Q-Learning for Mission-Critical Wireless Sensor Networks.

Mission-critical wireless sensor networks require a trustworthy and punctual routing protocol to ensure the worst-case end-to-end delay and reliability when transmitting mission-critical data collected by various sensors to gateways. In particular, the trustworthiness of mission-critical data must be guaranteed for decision-making and secure communications. However, it is a challenging issue to meet the requirement of both reliability and QoS in sensor networking environments where cyber-attacks may frequently occur and a lot of mission-critical data is generated. This study proposes a trust-based routing protocol that learns the trust elements using Q-learning to detect various attacks and ensure network performance. The proposed mechanism ensures the prompt detection of cyber threats that may occur in a mission-critical wireless sensor network and guarantees the trustworthy transfer of mission-critical sensor data. This paper introduces a distributed transmission technology that prioritizes the trustworthiness of mission-critical data through Q-learning results considering trustworthiness, QoS, and energy factors. It is a technology suitable for mission-critical wireless sensor network operational environments and can reliably operate resource-constrained devices. We implemented and performed a comprehensive evaluation of our scheme using the OPNET simulator. In addition, we measured packet delivery rates, throughput, survivability, and delay considering the characteristics of mission-critical sensor networks. The simulation results show an enhanced performance when compared with other mechanisms.

Study on Effects of Nonlinear Behavior Characteristics of Prepreg Dielectric on Warpage of Substrate under Laminating Process.

To analyze the effects of nonlinear behavior characteristics of prepreg (PPG) among the insulating materials of substrate and the residual stress of laminating process on the warpage of substrate, this study investigated the continuous laminating process using the numerical analysis by finite element method. The analysis results showed that the warpage of the substrate in the laminating process of PPG was very low, but it increased rapidly in the solder resist (SR) laminating process. As the laminating process of PPG continued, the stress inside the substrate increased continuously and it was predicted to decrease in the SR laminating process. These results confirmed that the warpage of the substrate is influenced the most by the SR laminating process, and that the warpage and stress of substrate accumulated in the laminating process of PPG have significant effects on the final warpage.

Trust Based Multipath QoS Routing Protocol for Mission-Critical Data Transmission in Tactical Ad-Hoc Networks.

In tactical ad-hoc networks, the importance of various tactical sensors and mission-critical data is increasing owing to their role in determining a tactical situation and ensuring the viability of soldiers. In particular, the reliability of mission-critical data has to be ensured for accurate situation determination and decision making. However, managing the network and trustworthiness in an environment where malicious nodes exist and a large amount of mission-critical data occur is a challenging issue. To solve these issues, a routing protocol is needed that can effectively detect malicious nodes and ensure the reliability and quality of service (QoS) of mission-critical data. In this paper, we propose a trust-based multipath QoS routing protocol (called MC_TQR) for tactical ad-hoc networks that can detect malicious nodes and satisfy the requirements of mission-critical data. The proposed scheme is verified using an OPNET simulator, and the results confirm the improved network performance when compared with existing schemes.

A Stepwise and Hybrid Trust Evaluation Scheme for Tactical Wireless Sensor Networks.

In tactical wireless sensor networks, tactical sensors are increasingly expected to be exploited for information collection in battlefields or dangerous areas on behalf of soldiers. The main function of these networks is to use sensors to measure radiation, nuclear, and biochemical values for the safety of allies and also to monitor and carry out reconnaissance of enemies. These tactical sensors require a network traffic flow that sends various types of measured information to the gateway, which needs high reliability. To ensure reliability, it must be able to detect malicious nodes that perform packet-dropping attacks to disrupt the network traffic flow, and energy-constrained sensors require energy-efficient methods to detect them. Therefore, in this paper, we propose a stepwise and hybrid trust evaluation scheme for locating malicious nodes that perform packet-dropping attacks in a tree-based network. Sensors send a query to the gateway by observing the traffic patterns of their child nodes. Moreover, depending on the situation, the gateway detects malicious nodes by choosing between gateway-assisted trust evaluation and gateway-independent trust evaluation. We implemented and evaluated the proposed scheme with the OPNET simulator, and the results showed that a higher packet delivery ratio can be achieved with significantly lower energy consumption.

Blockchain-Based Lightweight Trust Management in Mobile Ad-Hoc Networks.

As a trending and interesting research topic, in recent years, researchers have been adopting the blockchain in the wireless ad-hoc environment. Owing to its strong characteristics, such as consensus, immutability, finality, and provenance, the blockchain is utilized not only as a secure data storage for critical data but also as a platform that facilitates the trustless exchange of data between independent parties. However, the main challenge of blockchain application in an ad-hoc network is which kind of nodes should be involved in the validation process and how to adopt the heavy computational complexity of block validation appropriately while maintaining the genuine characteristics of a blockchain. In this paper, we propose the blockchain-based trust management system with a lightweight consensus algorithm in a mobile ad-hoc network (MANET). The proposed scheme provides the distributed trust framework for routing nodes in MANETs that is tamper-proof via blockchain. The optimized link state routing protocol (OLSR) is exploited as a representative protocol to embed the blockchain concept in MANETs. As a securely distributed and trusted platform, blockchain solves most of the security issues in the OLSR, in which every node is performing the security operation individually and in a repetitive manner. Additionally, using predefined principles, the routing nodes in the proposed scheme can collaborate to defend themselves from the attackers in the network. The experimental results show that the proposed consensus algorithm is suitable to be used in the resource-hungry MANET with reduced validation time and less overhead. Meanwhile, the attack detection overhead and time also decrease because the repetitivity of the process is reduced while providing a scalable and distributed trust among the routing nodes.

Conducting Super-Hydrophobic Thin Film for Electric Heating Applications.

Frost presents a serious problem for the human environment, resulting in such phenomena as downed power lines, damaged crops and stalled aircraft. In addition, frost and ice accumulation significantly decrease the performance of ships, wind turbines, and HVAC systems with high failure risk. Super-hydrophobic (SH) surface can be an appropriate solution for frost problems, due to its anti-icing properties that can prevent ice nucleation on the surface. In addition, in the case of conducting SH surface using carbon nanotubes (CNTs) as a filler, it can form an excellent heating unit, owing to the resistive heating effect. The purpose of this study is to produce a large-area conducting SH film that can prevent ice nucleus and remove ice formation rapidly. High aspect ratio carbon nanotubes (CNTs) as a conducting filler and adhesive polymer resin as a binder were used to form coating layer. In addition, silica particles (~7 nm) were used to stabilize nano-size roughness of the SH surface. Wet and dry etching processes were used on the substrate to improve wettability and to produce organic functional groups. To evaluate the de-icing effect, the fabricated SH surface was rapidly heated to 150 °C by applying voltage.

A Design and Simulation of the Opportunistic Computation Offloading with Learning-Based Prediction for Unmanned Aerial Vehicle (UAV) Clustering Networks.

Drones have recently become extremely popular, especially in military and civilian applications. Examples of drone utilization include reconnaissance, surveillance, and packet delivery. As time has passed, drones' tasks have become larger and more complex. As a result, swarms or clusters of drones are preferred, because they offer more coverage, flexibility, and reliability. However, drone systems have limited computing power and energy resources, which means that sometimes it is difficult for drones to finish their tasks on schedule. A solution to this is required so that drone clusters can complete their work faster. One possible solution is an offloading scheme between drone clusters. In this study, we propose an opportunistic computational offloading system, which allows for a drone cluster with a high intensity task to borrow computing resources opportunistically from other nearby drone clusters. We design an artificial neural network-based response time prediction module for deciding whether it is faster to finish tasks by offloading them to other drone clusters. The offloading scheme is conducted only if the predicted offloading response time is smaller than the local computing time. Through simulation results, we show that our proposed scheme can decrease the response time of drone clusters through an opportunistic offloading process.

A Continuous Object Boundary Detection and Tracking Scheme for Failure-Prone Sensor Networks.

In wireless sensor networks, detection and tracking of continuous natured objects is more challenging owing to their unique characteristics such as uneven expansion and contraction. A continuous object is usually spread over a large area, and, therefore, a substantial number of sensor nodes are needed to detect the object. Nodes communicate with each other as well as with the sink to exchange control messages and report their detection status. The sink performs computations on the received data to estimate the object boundary. For accurate boundary estimation, nodes at the phenomenon boundary need to be carefully selected. Failure of one or multiple boundary nodes (BNs) can significantly affect the object detection and boundary estimation accuracy at the sink. We develop an efficient failure-prone object detection approach that not only detects and recovers from BN failures but also reduces the number and size of transmissions without compromising the boundary estimation accuracy. The proposed approach utilizes the spatial and temporal features of sensor nodes to detect object BNs. A Voronoi diagram-based network clustering, and failure detection and recovery scheme is used to increase boundary estimation accuracy. Simulation results show the significance of our approach in terms of energy efficiency, communication overhead, and boundary accuracy.