Phuc Duc Nguyen

Phuc Duc Nguyen, Ryosuke Isogai, Keitarou Kondou et al.

In UAV-to-UAV communication, airborne UAVs need to detect the location and direction of ultra-high-speed millimeter-wave (mmWave) and Terahertz (THz) coverage areas, referred to as ultra-spots. This predictive capability allows UAVs to optimally adjust their flight paths, altitude, and velocity, thereby maximizing the utilization of ultra-spot services. A space-time synchronization technique employing multiple Wireless Two-way Interferometry devices (multi-Wi-Wi) is proposed in this paper to detect mmWave/THz ultra-spot locations during UAV operations. This paper proposes an algorithm that estimates the likelihood of nearby ultra-spots by considering the UAV flight route and ultra-spot direction, and by sharing location and pose information among UAVs in the network via a 920 MHz wireless communication link. For the first time, this work addresses the problem of optimizing UAV flight routes to maximize ultra-spot utilization. To address the inherent challenges of Wi-Wi, such as phase data unreliability, RSSI attenuation, or packet loss caused by obstructions from the UAV's own body, this study proposes the use of multiple Wi-Wi devices equipped with antennas positioned at different positions around the arms of the UAV to leverage spatial diversity effects. The proposed method's effectiveness is confirmed through experimental data derived from real-world UAV-to-UAV communication tests. An error of 37.16 cm was observed experimentally in ultra-spot location estimation, corresponding to 186 ms error in temporal prediction of ultra-spot entry from an in-flight UAV, demonstrating its effectiveness in addressing ultra-spot detection challenges in mmWave communication.

Phuc Duc Nguyen, Kazuhiro Maruyama, Keitarou Kondou et al.

Maximizing data uploading efficiency in a vehicular-to-road data uploading system using millimeter-wave communication is a challenging issue, as the wireless zone is often critically narrow, and vehicles can easily fail to pass through it without the aid of an autonomous guiding system. Variations in driving routes, speeds, approach angles, and distances to the ultra-spot can significantly affect data transmission performance, leading to either efficient or suboptimal results. This study presents a comprehensive analysis based on 75 experimental cases to identify the optimal travel trajectory and conditions that allow the vehicle to pass through the ultra-spot and enhance data transmission effectively. Experimental results show that with an optimal travel trajectory, appropriate movement speed, antenna placement, and prior estimation of the ultra-spot area, the amount of transferred data can be improved by 6 to 8 times.

Phuc Duc Nguyen, Quang Duc Nguyen

Test-time adaptation (TTA) effectively counters distribution shifts but exposes models to adversarial manipulation via the unlabeled test stream. Existing class-wise targeted attacks remain impractical for stealthy exploitation in this setting: since TTA operates on batches, forcing a subset of samples toward a target label unintentionally pulls similar benign samples along, resulting in a conspicuously high frequency of the target label that is easy to detect. To capture a more realistic threat, we introduce a sample-wise targeted attack. Unlike prior approaches, the attacker aims to misclassify only inputs carrying an attacker-chosen trigger, while preserving the global label distribution of benign queries to evade detection. To achieve this, we propose a meta-learning-based attack with a novel priority-aware gradient alignment strategy that explicitly prioritizes attack success. The strategy formulates the gradient update as an ellipsoidal trust-region problem, mitigating the misalignment between attack success and distributional stealth, while providing theoretical guarantees for effective optimization of the attack objective in the presence of gradient misalignment. Extensive experiments on CIFAR-10-C, CIFAR-100-C, and ImageNet-C across TTA protocols demonstrate that our method achieves high targeted success rates while maintaining a label distribution that is consistent with the no-attack baseline, making it difficult to detect in unlabeled TTA deployment scenarios. Furthermore, we demonstrate that our attack shows strong robustness against existing defenses.