Integrated sensing and communications (ISAC) is a key enabler for 6G, yet it introduces considerable interference that limits its performance. To minimize interference and enhance spectrum efficiency of the ISAC systems, we propose a user-grouping and full non-orthogonal multiple access (GF-NOMA) approach. Specifically, we first group communication users and sensing targets based on their spatial distribution, so that spatial beam orthogonality can be employed to reduce interference between different groups. Then, by taking sensing targets as virtual communication users, each group is fully multiplexed by NOMA in the power domain for interference cancellation. To optimize the performance of the proposed approach, we formulate a beamformer design problem to maximize the weighted sum of the communication throughput and sensing power. Meanwhile, we propose a double-penalty successive convex approximation (DP-SCA) algorithm to solve this problem by iteratively searching for the optimal solution. The simulation results demonstrate that the proposed approach improves the system performance by about 10% on average compared to the non-grouping approach. When compared to Partial-NOMA, the proposed Full-NOMA approach improves the communication users’ sum rate by about 20%.

Gossip learning (GL), as a decentralized alternative to federated learning (FL), is more suitable for resourceconstrained wireless networks, such as Flying Ad-Hoc Networks (FANETs) that are formed by unmanned aerial vehicles (UAVs). GL can significantly enhance the efficiency and extend the battery life of UAV networks. Despite the advantages, the performance of GL is strongly affected by data distribution, communication speed, and network connectivity. However, how these factors influence the GL convergence is still unclear. Existing work studied the convergence of GL based on a virtual quantity for the sake of convenience, which failed to reflect the real state of the network when some nodes are inaccessible. In this paper, we formulate and investigate the impact of inaccessible nodes to GL under a dynamic network topology. We first decompose the weight divergence by whether the node is accessible or not. Then, we investigate the GL convergence under the dynamic of node accessibility and theoretically provide how the number of inaccessible nodes, data non-i.i.d.-ness, and duration of inaccessibility affect the convergence. Extensive experiments are carried out in practical settings to comprehensively verify the correctness of our theoretical findings.

Estimation of wood moisture content (MC) is a fundamental aspect of woodworking, construction, and various other industries that rely on the versatile properties of wood. In this paper, we present mmMC, a novel system that estimates the MC of wood by a single commercial off-the-shelf (COTS) mmWave radar, which provides accurate end-to-end real-time wood moisture measurement results while being non-invasive, portable, and flexible in deployment. The proposed system uses a novel target reflection feature (TRF) to determine the reflectivity of the wood, thereby correlating it with specific MC levels. To robustly and accurately estimate the TRF of the wood from the mmWave signal, a signal processing pipeline is proposed. Specifically, an object detection and signal extraction module is designed to resolve the range of the object and eliminate multipath effects of the original signal. Then, a multiple chirps/antennas signals processing module is proposed to obtain a stable TRF from the extracted signals by using signals from different antennas and chirps. The TRF is associated with the MC of wood by a regression to enable real-time MC estimation. Through extensive real-world experiments, we have demonstrated that the proposed system has high accuracy and reliability.

Many civil structures are at risk of failure due to bolts loosening under shock or vibration. Early detection of looseness is critical for safety. In this research, we utilize mathematical model with motion equations and optimization technique to estimate parameters that can detect looseness. A novel tightness index is introduced to quantify the extent/degree of looseness. The method was validated on wood and steel structures for demonstrating its effectiveness.

Recent years have witnessed the ever-growing interest and adoption of autonomous vehicles (AVs), thanks to the latest advancement in sensing and artificial intelligence (AI) technologies. The LiDAR sensor is adopted by most AV manufacturers for its high precision and high reliability. Unfortunately, LiDARs are susceptible to malicious spoofing attacks, which can lead to severe safety consequences for AVs. Most current work focuses on protecting LiDAR against spoofing attacks by using perception model-level defense methods, whose effectiveness unfortunately depends on the correctness of the LiDAR’s sensing outcome. A spoofer thus can elude from these methods as long as it fabricates points that maintain the right contextual relationship held by the legitimate points. In this paper, we propose to use the signal’s Doppler frequency shift to verify the sender of the signal and detect potential spoofing attacks. To this end, we first thoroughly analyze the working principle of LiDAR and conduct real-world experiments to deeply understand and reveal the vulnerability of LiDAR sensors. We then prove that the Doppler frequency shifts of legitimate and spoofing signals present different characteristics, which can be used to fundamentally protect the LiDAR sensing outcome. For better demonstration purposes, we consider three attack models, including static attacker, moving attacker, and moving attacker with control of both velocity and signal frequency. For each of the models, we first show how the spoofing attack is performed and then present our countermeasures. We then propose a statistical spoofing detection framework to jointly consider the impact of short-term uncertainty in vehicle velocity, which can provide more accurate spoofing detection results in realistic environments. Extensive numerical results are provided in a wide range of settings and road conditions.

The development of the Internet of Things (IoT) combined with the emergence of federated learning (FL) makes it possible for mobile edge computing (MEC) to gain insight from physically separated data without violating privacy or burdening communication. Due to the distributed nature of MEC devices, researchers have uncovered that the FL is vulnerable to backdoor attacks, which aim at injecting a subtask into the FL without corrupting the performance of the main task. The backdoor attack achieves high accuracy on both the main task and the backdoor subtask when injected at FL model convergence. However, the effectiveness of the backdoor is weak when injected in early training stage. In this article, we strengthen the early-injected backdoor attack by using information leakage. We show that FL convergence can be expedited if the client’s data set mimics the distribution and gradients of the whole population. Based on this observation, we propose a two-phase backdoor attack, which includes a preliminary phase for the subsequent backdoor attack. Taking advantage of the preliminary phase, the later injected backdoor achieves better effectiveness, as the backdoor effect is less likely to be diluted by normal model updates. Extensive experiments are conducted on the MNIST data set under various data heterogeneity settings to evaluate the effectiveness of the proposed backdoor attack. The results show that the proposed backdoor outperforms existing backdoor attacks in both success rate and longevity, even when defense mechanisms are in place.

Reconfigurable intelligent surface (RIS) offers a new way to provide controllable non line-of-sight (NLoS) propagation paths for millimeter-wave (mmWave) directional communication to overcome the performance degradation caused by line-of-sight blockage. However, current coverage models do not consider the impact of path direction difference on path's availability, which is a crucial property of mmWave directional communication network. In the paper, we propose a new coverage model called$(k,\alpha )$-coverage. A receiver is$(k,\alpha )$-covered if it is covered by at least$k$RISs to have$k$different NLoS path directions and the angular separation between any two adjacent path directions is at least$\alpha$. In this case, when the current communication direction is blocked by an obstacle, other RIS created paths are still likely to be available for transmission, which increases the robustness of mmWave directional communication. To tackle the problem of using the least number of RISs to achieve the$(k,\alpha )$-coverage, we formally define the$(k,\alpha )$-coverage models and propose methods to verify if the target area is$(k,\alpha )$-covered by the given set of RISs. Then, we solve the problem under both deterministic and random RIS deployment schemes. For the deterministic deployment scheme, we derive the optimal$k$-sided regular polygon deployment patterns and use it to achieve area$(k,\alpha )$-coverage. An analytical performance bound on the number of RISs needed is also derived. For the random RIS deployment scheme, we derive the$(k,\alpha )$-coverage probability under uniform and spatial-Poisson RIS distributions. Finally, extensive simulation results are provided to validate our analyses.

The development of the Internet of Things (IoT) combined with the emergence of federated learning (FL) makes it possible for mobile edge computing (MEC) to gain insight from physically separated data without violating privacy or burdening communication and the server. Due to the distributed nature of MEC, researchers have uncovered that the FL is vulnerable to backdoor attacks, which aim at injecting a subtask into the FL without corrupting the performance of the main task. However, the single-shot backdoor attack in the early training stage is weak. In this paper, we strengthen the early-injected single-shot backdoor attack by using information leakage. We show that FL convergence can be expedited if the client's dataset mimics the distribution and gradients of the whole population. Based on this observation, we propose a two-phase backdoor attack, which includes a preliminary phase for the subsequent backdoor attack. Benefiting from the preliminary phase, the later injected backdoor achieves better effectiveness, as the backdoor effect is less likely to be diluted by normal model updates. Numerical experiments show that the proposed backdoor outperforms existing backdoor attacks in both success rate and longevity, even when defense mechanisms are in place.

The high directionality of mmWave communication makes its line-of-sight (LoS) path susceptible to blockage when the user is moving. Most existing solutions have very stringent requirements on the antennas of the transmitter and the receiver, which are hardly met by today's consumer-level commercial off-the-shelf (COTS) mmWave products. In reality, a COTS device uses low-resolution wide-beam antennas, and hence cannot support the aforementioned methods for NLoS beam switching in response to the LoS blockage. In this paper, we develop a new method to support high-resolution mmWave multi-path channel resolving based on coarse-grained wide-beam phased array antennas. We design a novel real-time beam-switching algorithm that allows COTS devices to estimate the location and reflection coefficient of the dominant reflectors. Whenever the current LoS is blocked, our algorithm can compute in real-time the best alternative beam direction based on estimated reflectors to establish a strong NLoS link. We implemented the proposed algorithm on a COTS mmWave device and evaluated the system's performance on the physical and transport layer. Our experiments demonstrate the effectiveness of our algorithm on estimating dominant reflectors and calculating strong alternative beam directions, and its efficacy in providing robust connections for COTS mmWave devices.

Differential privacy (DP) has been widely explored in academia recently but less so in industry possibly due to its strong privacy guarantee. This paper makes the first attempt to implement three basic DP architectures in the deployed telecommunication (telco) big data platform for data mining applications. We find that all DP architectures have less than 5% loss of prediction accuracy when the weak privacy guarantee is adopted (e.g., privacy budget parameter ε ≥ 3). However, when the strong privacy guarantee is assumed (e.g., privacy budget parameter ε ≤ 0:1), all DP architectures lead to 15% ~ 30% accuracy loss, which implies that real-word industrial data mining systems cannot work well under such a strong privacy guarantee recommended by previous research works. Among the three basic DP architectures, the Hybridized DM (Data Mining) and DB (Database) architecture performs the best because of its complicated privacy protection design for the specific data mining algorithm. Through extensive experiments on big data, we also observe that the accuracy loss increases by increasing the variety of features, but decreases by increasing the volume of training data. Therefore, to make DP practically usable in large-scale industrial systems, our observations suggest that we may explore three possible research directions in future: (1) Relaxing the privacy guarantee (e.g., increasing privacy budget ε) and studying its effectiveness on specific industrial applications; (2) Designing specific privacy scheme for specific data mining algorithms; and (3) Using large volume of data but with low variety for training the classification models.