Radar serves as the core sensor on the battlefield and constructs collaborative perception networks via wireless communication. Nevertheless, radar networks are vulnerable to low-altitude saturation penetration owing to line-of-sight restrictions, resource bottlenecks, and electromagnetic exposure risks, accompanied by dynamic and time-varying topological characteristics. This paper proposes a greedy-based heuristic Position Occupation Algorithm (POA), which can rapidly construct single-layer strongly closed barrier coverage for time-varying low-altitude radar networks. The framework of this paper models the process of barrier construction as four stages, including frontier establishment, front expansion, residual position occupation, and front contraction. Adaptive divide-and-conquer, tabu search, and local search strategies are integrated in the POA to achieve a polynomial time complexity of O(I2 & sdot; P + N2). Simulation results show that POA slightly outperforms the nondominated sorting genetic algorithm II (NSGA-II), the nondominated sorting fireworks algorithm (NSFWA), and the multiobjective evolutionary algorithm based on decomposition (MOEA/D) in barrier coverage area and number of protected radars, while reducing reconstruction time by over two orders of magnitude, thereby providing an operationally efficient solution for barrier deployment in wireless low-altitude radar networks.
Physical-layer authentication for LoRaWAN sensor nodes increasingly relies on SRAM Physical Unclonable Functions (PUFs) or Radio Frequency (RF) fingerprinting, yet prior work has not reported the joint failure trajectories of both modalities under identical, directly measured stress on the same physical device. This paper reports a device-level finding (referred to here as D4) obtained with a calibrated DS18B20 temperature probe (+/- 0.5 degrees C) bonded adjacent to the SRAM, so that every authentication outcome is paired with a measured rather than estimated temperature. Using a stressor-decoupling protocol on Device C (STM32 B-L072Z-LRWAN1 + Microchip 23K640 SRAM), we show that no single stressor drives the PUF past its 15% authentication threshold: at a fixed 50 cm range the genuine fractional Hamming distance (FHD) rises smoothly and monotonically from 3.0% to only 10.9% across 27.4-56.1 degrees C (0.374%/degrees C), with a False Reject Rate of 0% throughout. Distance alone (+2.2 pp) and elevated temperature alone (+7.0 pp) each leave the PUF below threshold. PUF authentication fails (Genuine Accept Rate 100% -> 0%, FHD 17.7%) only under the compound condition (50 degrees C + 100 cm), where the observed degradation exceeds the sum of the individual stressors, indicating a super-additive thermal-distance interaction. Throughout the same sweep the Carrier Frequency Offset (CFO) fingerprint drifts only-14.9 Hz/degrees C (Spearman = -0.52), matching an AT-cut crystal model and remaining within-device stable (158 Hz) against thousands-of-Hz inter-device separation (ANOVA = 706), so RF discriminability is preserved precisely where the PUF fails. The compound failure reproduces across both MCU platforms (Feather M0 and STM32), while a device sharing the SRAM2 chip with the failing STM32 node but hosted on a Feather remains robust (FHD 7.6%, GAR 100%), indicating platform-dependent rather than chip-intrinsic behaviour. These findings motivate, as a design implication, adaptive hybrid authentication in which RF-only fallback is invoked when PUF reliability degrades. The observations are drawn from a small device set and require validation on a larger population.
Partitioned deep neural network inference exposes intermediate feature representations to potential manipulation during transmission between computational nodes. Adversarial perturbations injected into these boundary tensors can propagate through the remaining network and cause downstream misclassification, while existing defenses either detect anomalies without restoring prediction reliability or apply unconditional recovery at unnecessary computational cost. This paper presents a cost-aware adaptive defense framework for feature-space adversarial attacks in partitioned deep neural networks operating under realistic communication degradation in distributed edge AI systems. A lightweight detector models clean boundary representations using channel-wise feature statistics and identifies anomalous deviations with controlled false-positive behavior. For flagged samples, a repair module projects anomalous statistics toward the clean region and performs moment-matching correction of the received feature map, with optional masked repair to restrict modification to the most anomalous channels. An MD-gated contextual bandit supervisor then selects the recovery action to balance robustness and computational cost. Experiments on partitioned VGG16 and ResNet50 under channel-sparse Gaussian-noise perturbations, together with an additional iterative query-based feature-space attack based on Natural Evolution Strategies (NES), show that the proposed framework maintains near-clean accuracy, substantially recovers attacked performance, and reduces unnecessary recovery compared with unconditional repair. The results demonstrate that lightweight detection combined with selective and adaptive correction provides an effective defense for robust distributed inference.
Performance evaluation of mobile ad hoc network (MANET) routing protocols relies heavily on simulation, yet the choice of propagation model, a critical determinant of link connectivity, is rarely justified or systematically examined. Through a bibliometric audit of 32 comparative routing studies, we demonstrate that urban MANET evaluations systematically employ propagation models unsuited to their claimed deployment contexts, including free-space and two-ray models that ignore building obstruction entirely, and, in a majority of cases, empirical urban models such as ITU-R P.1411 that capture aggregate attenuation but not individual building geometry. We then show that loss model selection has protocol-ranking consequences through controlled ns-3 experiments comparing AODV, OLSR, and DSDV under four propagation models across fixed and stochastic building layouts. Under open-field models, protocol rankings are stable and reproducible; under the recently developed building-aware FOBA model for ns-3, rankings dissolve or reverse depending on layout, density, and mobility. End-to-end delay increases up to tenfold under FOBA in certain configurations. Packet drop analysis reveals that this gap stems from reachability effects rather than physical-layer frame loss. Graph-theoretic path metrics (lifetime, stability and persistency) provide a mechanistic explanation: building obstruction produces position-correlated link breaks absent from statistical models, fundamentally altering network topology dynamics independently of routing behavior. These findings demonstrate that propagation model choice is not a secondary implementation detail but a first-order methodological factor capable of inverting comparative conclusions.
Wireless sensor networks (WSNs) play a critical role in environmental monitoring applications such as pollution detection, forest surveillance, and disaster management. However, conventional centralized monitoring approaches require frequent data transmission to a fusion center (FC), resulting in significant communication overhead, uneven energy consumption, and reduced network lifetime. To address these challenges, this paper proposes a distributed framework for the estimation of convex event regions based on diffusion cooperation and minimum volume ellipsoid (MVE) representation. The proposed Cooperation-Then-Minimum-Volume-Ellipsoid Formation (CTMVEF) strategy enables sensor nodes (SNs) to collaboratively estimate the geometric boundary of a convex event region using only local communication with neighboring nodes, thereby eliminating the need for centralized processing. An adaptive covariance-weighted diffusion mechanism is introduced to improve convergence speed and robustness under noisy measurements. In addition, a boundary-node-based estimation strategy is employed to reduce computational complexity and communication overhead while preserving estimation accuracy. Extensive simulations demonstrate accurate convex event-region estimation and fast convergence while significantly reducing communication cost and energy consumption compared with conventional centralized approaches, confirming that the proposed framework provides an effective and scalable solution for energy-efficient environmental monitoring in large-scale WSNs.
