Federated Learning (FL) enables collaborative model training across distributed devices while safeguarding data and user privacy. However, FL remains susceptible to privacy threats that can compromise data via direct means. That said, indirectly compromising the confidentiality of the FL model architecture (e.g., a convolutional neural network (CNN) or a recurrent neural network (RNN)) on a client device by an outsider remains unexplored. If leaked, this information can enable next-level attacks tailored to the architecture. This paper proposes a novel side-channel fingerprinting attack, leveraging flow-level and packet-level statistics of encrypted wireless traffic from an FL client to infer its deep learning model architecture. We name it FLARE, a fingerprinting framework based on FL Architecture REconnaissance. Evaluation across various CNN and RNN variants-including pre-trained and custom models trained over IEEE 802.11 Wi-Fi-shows that FLARE achieves over 98% F1-score in closed-world and up to 91% in open-world scenarios. These results reveal that CNN and RNN models leak distinguishable traffic patterns, enabling architecture fingerprinting even under realistic FL settings with hardware, software, and data heterogeneity. To our knowledge, this is the first work to fingerprint FL model architectures by sniffing encrypted wireless traffic, exposing a critical side-channel vulnerability in current FL systems.

This paper investigates covert multi-hop communication in wireless networks where an adversary employs a cyclostationary (cycle) detector to reveal hidden transmissions. The covert route employs direct sequence spread spectrum (DSSS) signaling to ensure either maximum end-to-end covertness maximization or minimum latency minimization-under quality-of-service (QoS) and link budget constraints. Optimal bandwidth, transmit power, and spreading gain for each hop jointly satisfy reliability and either rate or covertness requirements. We show the equivalence between the covertness and the detection SNR gain-based widest-path formulations, and, hence, enabling efficient route computation. Numerical simulations in a realistic 3D environment illustrate that (i) end-to-end latency increases exponentially with the covertness requirement, (ii) the end-to-end latency increase is super-linear with the packet size M, and (iii) cycle and energy detectors impose different latency behavior as a function of the message length and the covertness requirement. The proposed framework provides important insights into resource allocation and routing design for covert networks against advanced detection adversaries.

State-of-the-art covert routing in heterogeneous networks (HetNets) focuses on balancing covertness and throughput, but often overlooks explicit energy optimization. While covert communication inherently limits transmit power, meeting throughput demands without coordinated design can still lead to high energy consumption. To this end, we propose DECOR, Decentralized Energy-efficient COvert Routing framework that jointly optimizes covertness, throughput, and energy efficiency. Unlike traditional methods that use a single wireless technology, DECOR leverages the diversity of available wireless communication technologies in HetNet to enable simultaneous multi-modal routing. The core idea behind DECOR is that optimal simultaneous utilization of multiple modalities improves throughput and overall energy efficiency. It minimizes the end-to-end energy consumption while satisfying stringent constraints on throughput and covertness through two core steps: (1)link-level optimizationusing sequential least squares programming (SLSQP), and (2)network-level optimizationthrough a custom cluster-based routing strategy. DECOR introduces a novel clustering-based strategy that aggregates intra-cluster link information and delegates routing decisions to cluster heads, significantly reducing control overhead and enabling scalable, energy-efficient covert communication. Extensive numerical analysis demonstrates that DECOR significantly outperforms existing approaches in terms of energy-efficiency and data overhead.

Multi-hop Heterogeneous wireless networks (HWNs) with multiple communication technologies have been extensively studied driven by the rising demand for enhanced resilience, coverage, and throughput. However, utilizing relays for wireless communication has the potential to heighten the risk of detection by an adversary. Furthermore, information about the adversary is usually unavailable in practice. Therefore, in this paper, we propose a covert routing using Q-learning for HWNs to maximize the detection error probability (DEP) with only limited knowledge about the adversary's general location, referred to as the threat region. To achieve this goal, we exploit fictitious adversaries that are randomly located inside the threat region to obtain estimated DEP. Then, we propose three different approaches to establish a route between the source and destination. Through simulations, we compare our proposed approaches to the conventional covert routing using Q-learning where the location of the adversary is assumed to be known. Our results show that our methods only experience a 10% reduction in DEP with fictitious adversaries. Furthermore, we investigate how the radius of the threat region affects the performance.

Reinforcement learning (RL) holds great promise for network control problems, yet its deployment in real-world systems remains limited due to the instability and unpredictability of RL policies during training. To address this challenge, we propose a two-phase conservative RL framework that combines domain expertise from classical network optimization with modern deep RL techniques. Our key idea is to initialize the learning process with a stable base policy, derived from expert knowledge, and then apply conservative fine-tuning under a Kullback–Leibler (KL) divergence constraint to safely explore improved behaviors. We apply this framework to the problem of covert multi-hop routing, where the objective is to optimize data throughput while minimizing detectability by adversaries. In Phase I, we construct a reliable base policy by imitating the back-pressure algorithm, which guarantees throughput-optimal behavior and stable queue dynamics. Phase II fine-tunes this policy to improve covert performance, as measured by the Detection Error Probability (DEP), while preserving training-time stability. Empirical evaluations on a grid network show that our method enables more reliable learning than pure RL. While pure RL (e.g., PPO) can sometimes achieve higher covert performance, it frequently suffers from large queues and collapsed throughput during training. In our experiments, our conservative RL framework reduces the worst-case training-time queue length by over 99% while maintaining comparable covert communication performance.

We consider a three party wireless network- Alice is a legitimate transmitter, Bob is a legitimate receiver, and Willie is a malicious adversary. Alice's goal is to communicate with Bob with a certain minimum Quality of Service (QoS) while evading detection from Willie. Alice transmits a Direct Sequence Spread Spectrum (DSSS) signal where the processing gain and transmit power are selected with the aim of evading detection by spreading the bit energy over a larger bandwidth while maintaining required QoS. Concurrently, Willie tries to detect the legitimate transmission by using either an energy or a cycle detector. We study this scenario within an adversarial optimization framework that aims at achieving a robust max-min covertness under link performance constraints. The framework is common to both detectors and hence allows for a comparative performance analysis that identifies common key performance parameters. The adversarial signal-to-noise ratio (SNR) detection gain enables direct trading of the DSSS processing gain for the (squared) channel quality ratio when aiming to improve link covertness, regardless of which detector Willie uses. The DSSS processing gain and SNR gain (and, hence, covertness) are limited by the bit rate and bit error rate link requirements, respectively. While both detectors' performance is limited by the SNR at Willie, the cycle detectors benefit significantly more from longer observation time of the legitimate transmission.

Many traditional feedback-based beamforming techniques can pose risks in environments where adversaries try to jam or detect the beamforming agents. Several low- or no-feedback approaches to beamforming have been proposed to remedy this. Still, problems such as limited transmitter power budgets and inaccurate Channel State Information (CSI) complicate this problem. To that end, this work investigates the use of constrained optimization techniques to allow a team of distributed Unmanned Ground Vehicles (UGVs) to form a beam in the absence of feedback without accurate knowledge of CSI. Specifically, we tune each transmitting agent's signal amplitude and phase offset to selectively build higher-power beams at allied receivers and lower-power nulls at adversarial receivers. To achieve this, we implement a decentralized version of the Frank-Wolfe algorithm, which does not rely on a central coordinator to solve the optimal signal amplitudes and phase offsets without accurate CSI measurements. Numerical results are also provided to validate our approach.

A fundamental challenge in covert routing is that meeting both covertness and throughput requirements often leads to increased transmit power, which can significantly elevate the overall energy consumption of the network. Therefore, it is important to achieve higher throughput and better energy efficiency while maintaining the required covertness. To this end, we propose a novel simultaneous multi-modal Decentralized Energy- Efficient covert Routing approach - DEER for a multi-hop heterogeneous network (HetNet). Unlike the prevailing single-modal approaches, DEER leverages the diversity of the available wireless communication technologies for simultaneous multi-modal routing. DEER aims to minimize the end-to-end total transmit power of the whole route in a decentralized fashion while maintaining the constraints on required throughput and covertness. DEER stems into two main steps: node-level optimization followed by network-level optimization using the proposed custom-tailored Dijkstra's based link state routing protocol to meet the constraints while minimizing the end-to-end total transmit power. We demonstrate by numerical analysis that DEER improves the energy efficiency by$23.5 x$and$2.9 x$times in comparison to the baseline single-modal and naive simultaneous multimodal approaches respectively.

In this paper, we study covert communication in wireless networks consisting of a transmitter, a receiver, an adversary, and multiple randomly distributed interferers. We model the spatial distribution of the interferers as a ß -Ginibre point process which can consider a repulsion among the interferers' locations and simplifies to the Poisson point process for a particular case. Both the average detection error probability (DEP) at the adversary and the communication outage probability (COP) at the receiver are investigated. We derive approximations of the average DEP and COP, and validate that the approximations are tight when compared to the simulated results. In addition, we examine the impact of the repulsion among the interferers on both the average DEP and COP. Finally, we show that the achievable average DEP, which is the maximized DEP with a requirement on the COP, is enhanced when there exists repulsion among the positions of the interferers.

In this paper, we develop an approach to exploit multiple disparate wireless communication technologies simultaneously to enhance covertness of a communication link. Specifically, given two available communication modalities between a pair of friendly nodes (Alice and Bob), the goal is to evade detection by an adversary (Willie) who is equipped with a radiometer covering the frequency bands of both modalities. We propose a joint detection threshold optimization technique from Willie's point of view. We also develop a joint transmit power optimization strategy for Alice to maximize covertness while meeting the throughput requirement at Bob. Through numerical simulations we show that the proposed scheme matches the performance of exhaustive search method while reducing the computational time by 98% and also improves the covertness by 56% compared to a naïve benchmark scheme.

In this paper, we explore covert routing communication in a heterogeneous network where a source sends a confidential message to a destination node with the help of relaying nodes where each node adaptively selects one modality among multiple communication modalities based on the wireless environment. We study three optimization problems: 1) the maximization of the end-to-end detection error probability at an adversary with a requirement on the throughput; 2) the end-to-end throughput maximization under a covertness constraint; and 3) the end-to-end latency minimization with a covertness condition. For the three optimization problems, we develop novel algorithms that identify the routes from a source to a destination and allocate resources, which are communication modality, transmit power, and bandwidth, at all nodes along the route. First, for single-hop communications, we derive a closed-form joint optimal power and bandwidth solution for a given modality, and then provide a modality selection method. For multi-hop communications, we propose the optimal routing strategies for the three problems by modeling the network as graphs and defining edge weights based on the objectives of the problems. From numerical simulations, it is validated that the performance of the network can be enhanced with the proposed optimal joint route and resource allocation techniques by judiciously selecting one of the multiple modalities for each hop in the route based on the wireless environment.

Ultraviolet (UV) communications has been proposed as a promising modality for short-range military communications, as it is often presumed to have low-probability-of-detection characteristics, has desirable non-line-of-sight properties, and resides within an underutilized frequency band. Recent research efforts have sought to formalize the first presumption of the detection of UV communications. This effort seeks to begin the study of the localization of UV communication sources after they are detected. We focus here exclusively on the range estimation problem. Using a phenomenon relating UV received power and range over short-to-medium distances (≪ 1 km), we develop a range estimator using only the received signal strength. The approach does not require information about other system or environmental parameters. We also theoretically study the performance of the estimator using the Cramér-Rao bound, via simulations, and using previously collected data.

In this paper, we explore covert communication in a heterogeneous network where a transmitter sends a confidential message to a receiver by utilizing two different radio frequency (RF) bands, low-very high frequency (low-VHF) and microwave frequency modalities. Since two RF modalities exhibit different characteristics in terms of channels and available bandwidths, it is important to efficiently leverage the modalities based on the wireless environment. Therefore, we develop a new algorithm that optimizes the transmit powers and bandwidths for the modalities and selects one modality with the goal of maximizing the detection error probability at a warden while guaranteeing a quality-of-service requirement of the receiver. We first derive a closed-form joint optimal power and bandwidth solution for a given modality, and then provide a modality selection method. From numerical simulations, it is validated that the proposed scheme achieves the optimal performance and the covertness can be enhanced by judiciously choosing one of the two modalities based on the channel condition.

Easily and flexibly deployable ad-hoc communication networks emerging in tactical military or even civilian contexts, frequently suffer from poor synchronization due to lack of coordinating infrastructure. In addition to synchronization issues, and especially in military settings, security from the aspect of detectability is also of crucial importance. Imperfect synchronization can be dealt with by making use of Quasi-Synchronous Code Division Multiple Access (QS-CDMA), relying on Loosely Synchronous Codes to maintain orthogonality in the presence of limited time delays. Security, in terms of low probability of detection (LPD) from the standpoint of a malicious adversary, can be improved (reduced detection) by employing randomization techniques that disrupt the inherent structure of the transmitted QS-CDMA signals. This is based on the fact that QS-CDMA signals are Cyclostationary, having (almost) periodic Auto-Correlation functions (ACF) due to eminent signal periodicities (such as spreading code repetition). In this paper, we propose techniques to disturb the ACF and equivalently the Spectral Correlation function, and reduce the Degree of Cyclostationarity (DCS), our LPD measure. Specifically, we investigate randomization via 1) random per symbol time dithering and 2) random selection of spreading sequences, as well as a hybrid approach combining time dithering and code randomization. In all proposed techniques knowledge of the randomization pattern is not required at the legitimate receiver. We derive the Spectral Correlation function of the QS-CDMA signal under the proposed randomization schemes and compare it to simulations. We show through analysis and extensive numerical simulations that the proposed technique can reduce the DCS by almost two orders of magnitude. We also show that enhanced LPD can be achieved using the proposed techniques while sacrificing a part of the reduced time synchronization requirement. We further analyze the implications of the friendly receiver not knowing the randomization pattern and present results on the resulting communication performance.

Arrays of closely-spaced antennas with mutual coupling have been considered recently with analogies to the hearing mechanism in small insects that exhibit excellent direction finding capabilities. We develop a model for a two-element array system that includes three distinct noise sources and a 4-port electrical network that couples the antennas to the measurement loads. The optimum coupling network that minimizes the Cramer-Rao bound (CRB) for angle of arrival (AOA) estimation is derived and a design method is presented to synthesize the network. A compact expression for the CRB provides insights about the fundamental value of mutual coupling and the importance of an optimum coupling network, and we show simulation results for an AOA estimator that achieves the CRB.

In this paper, we investigate the problem of distributed coherent beamforming in wireless networks where multiple distributed transmitters adjust the phases of their signals to form a directional and targeted communication link to a client receiver. The quality-of-service (QoS) and security are key components of robust and covert wireless networks. Although the security can be improved by exploiting information about potential adversaries, such information may not be available in practical networks since the adversaries are often passive. Therefore, we introduce transmission strategies which not only send a confidential message by forming a beam towards the client but also broadcast interference with the aim of obfuscating adversaries without having any information about them. Two different client feedback scenarios are considered, two-bit and rich feedback. The proposed algorithms can be performed in a fully distributed manner without any knowledge about potential adversaries. Numerical simulations validate the effectiveness of the proposed schemes.

In this paper, we investigate a wireless network where multiple distributed transmitters adjust the phases of their signals so that they can be constructively added at an intended receiver (client). Unlike conventional beamforming with co-located and phase-synchronized antennas, geographically separated transmitters may have phase offsets induced by individual local carrier oscillators, that pose a challenge for coherent distributed beamforming. This is especially true for transmitters that are far apart, when distributed clock synchronization protocols may be more difficult to implement. There may also be a desired spatial repulsion among the positions of the transmitters in order to mitigate mutual coupling effects and extend the coverage region. In this regard, we analyze the performance of distributed beamforming with phase offsets by modeling the spatial distribution of the transmitters as a $\beta$-Ginibre point process that models the repulsive behavior. We consider two transmission strategies: (i) Transmitter selection in which the client chooses the transmitter providing the highest received power at the client, and (ii) Coherent beamforming in which multiple transmitters simultaneously send their signals to the client. From numerical simulations, we examine the impact of the phase offsets on the performance and confirm the accuracy of our analysis. It is shown that even with significant phase offset errors, employing coherent beamforming can be an effective strategy.

Challenging Army-relevant communications and networking environments require novel methods and systems with sufficient robustness and resilience to enable operation in a covert fashion, especially in the presence of sophisticated adversaries. Recent studies suggest that regions of electromagnetic spectrum, such as deep ultraviolet (UV) spectrum, that are not being utilized in existing systems can provide unique advantages, particularly for more-covert short- range operations. However, the UV channel is not well understood, and techniques for the effective and efficient use of this channel for networking, such as spatial multiplexing, are still being investigated. In this paper, we report on experiments that validate previously hypothesized channel modeling behaviors, where we isolate and accurately measure the line-of-sight (LOS), single-scattering, and multiple-scattering components of measured returns and compare the measurements with theory. These experimental measurements exhibit good agreement with theoretical predictions, and demonstrate, in particular, predicted variations from ideal behavior in the multiple-scattering case. The validated behaviors play a key role in a previously proposed UV networking optimization framework. As such, we conclude this paper with additional development and examples of that framework in order to demonstrate the utility of the validated channel modeling.

This paper presents an experimental investigation of real-time digital video streaming in physically complex Non-Line-Of-Sight (NLoS) channels using a low-power, low-VHF system integrated on a compact robotic platform. Reliable video streaming in NLoS channels over infrastructure-poor ad-hoc radio networks is challenging due to multipath and shadow fading. In this effort, we focus on exploiting the near-ground low-VHF channel which has been shown to have improved penetration, reduced fading, and lower power requirements (which is critical for autonomous agents with limited power) compared to higher frequencies. Specifically, we develop a compact, low-power, low-VHF radio test-bed enabled by recent advances in efficient miniature antennas and off-the-shelf software-defined radios. Our main goal is to carry out an empirical study in realistic environments of how the improved propagation conditions at low-VHF affect the reliability of video-streaming with constraints stemming from the limited available bandwidth with electrically small low-VHF antennas. We show quantitative performance analysis of video streaming from a robotic platform navigating inside a large occupied building received by a node located outdoors: bit error rate (BER) and channel- induced Peak Signal-to-Noise Ratio (PSNR) degradation. The results show channel-effect-free- like video streaming with the low-VHF system in complex NLoS channels.

In this paper, we focus on the problem of distributed coherent beamforming in wireless networks where multiple distributed transmitters adjust the phases of their signals to form a directional and targeted communication link to a client receiver. The quality-of-service (QoS) and security are key aspects of robust and covert wireless networks. Although the security can be enhanced by exploiting information about the locations of adversaries, such information may not be available in practical networks since the adversaries are often passive. Therefore, we propose transmission strategies which divide transmitters into two groups where one group forms a beam towards the client and the other group broadcasts interference in order to obfuscate adversaries. As the interference may degrade the QoS of the client, the latter group steers a null to the client to alleviate the interference at the client. The proposed scheme can be performed in a fully distributed manner with only two bits of feedback information from the client and without any knowledge about the locations of potential adversaries.

Ad-hoc flexibly deployable networks destined for tactical military or civilian applications often suffer from poor synchronization which may lead to unreliable communication. Quasi-Synchronous (QS) CDMA employing Loosely Synchronous (LS) codes allows weak synchronization due to the Zero-Correlation Zone of the code's correlation functions. Apart from reliability, an essential characteristic of military networks is having low probability of detection (LPD) in the presence of adversaries employing sophisticated detection techniques. Such techniques are based on the fact that manmade signals are cyclostationary, meaning that they have some periodic structure (e.g. spreading sequence repetition) that can be exploited for improved detection. Disturbing those recurring patterns can reduce the probability of detection measured in terms of the Degree of Cyclostationarity (DCS). We aim to achieve that by employing some techniques that will perturb the signal structure by randomly selecting spreading sequences, random time dithering or a combination of the two. We study the trade-off between the communication performance and DCS reduction of such a system under these different techniques and compare the results.

Supporting networks with a large number of nodes in infrastructure-poor and complex propagation environments is an important challenge for military and civilian applications. A major problem when using classical approaches, such as code division multiple access (CDMA), is maintaining inter-link coordination while mitigating multi-user interference (MUI). By contrast, loosely synchronous (LS) codes have perfect code orthogonality within a window of inter-link delays at a cost of the number of available spreading codes. Since sporadic communications naturally involves a low probability of transmission, we investigate the potential for LS code reuse to effectively support more users. We study this problem by simulating inter-user channels using a high-fidelity physics-based model. We focus our study of the channel on the low-VHF band, which has improved penetration and channel coherence in complex environments. We perform initial characterization of different levels of code reuse, synchronization and coordination. Of particular interest is a purely random (uncoordinated) spreading code assignment. The results illustrate good performance of a scalable medium access scheme.

The lower VHF band shows potential for reliable communications in low power, short range scenarios among near-ground nodes in both indoor and urban environments. Such scenarios are of great interest, for example, in military and search-and-rescue settings. Most prior work at low VHF focuses on modeling path loss at long range. In this paper, we study indoor/outdoor near-ground scenarios through experiments focusing on both line-of-sight (LoS) and non-LoS (NLoS), at ranges up to 200 meters. By transmitting tones and pulses from various locations in a realistic environment, we acquire channel data via a mobile data collection platform which gathers data at hundreds of different locations. We show that the measured channels have a nearly ideal scalar attenuation and delay transfer function, with minimal phase distortion, and little evidence of multipath propagation. We further confirm the absence of small scale fading by measuring bit error rate (BER) versus received signal-to-noise ratio (SNR) for QPSK transmission in an indoor setting. Using only timing and carrier estimation at the receiver, the resulting BER curves coincide with theoretical additive white Gaussian noise channel BER predictions.

A high-resolution subsurface imaging technique based on distributed near-ground sensor networks that utilize ultrawideband waveforms in the very high frequency (VHF) range is presented. An accurate scattering model for a target buried in realistic subsurface environment, modeled as a vertically stratified medium, is presented first. Then an inversion technique that uses ultrawideband near-field focusing is described. The signal penetration depth as a function of frequency and various subsurface parameters is calculated based on the developed forward model. The imaging resolution as it relates to the accuracy of background retrieval is also analyzed. A semi-empirical soil dielectric model that is originally developed for the ultra high frequency band is modified and validated at the VHF range with measured results available in the literature. For a given soil textural composition and frequency, the model predicts the real and imaginary parts of the dielectric constant as a function of soil moisture content. This soil dielectric model is utilized to make the inversion more efficient. To address the challenge associated with the design of compact and ultrawideband VHF antennas, a scheme utilizing multiple antennas and reduced number of frequency points is proposed. The sensor arrangement both in terms of spatial distribution and polarization of each antenna as it relates to the lateral resolution, as well as minimizing the direct coupling between the Tx and Rx antennas, is analyzed. The proposed subsurface imaging approach is validated based on numerical techniques and a laboratory scale model measurement results.

A high resolution subsurface imaging approach based on near-ground sensor networks operating in the VHF range that utilize ultra-wideband near-field focusing was recently proposed [1]. An accurate scattering model for targets buried in realistic subsurface environment modeled as a vertically stratified medium as well as an efficient inversion technique using an UWB near-field focusing were proposed. Numerical models were used to analyze the signal penetration depth in the VHF range and validate the proposed technique for targets that are buried at various depths [2]. In this paper, in order to investigate the performance of the proposed approach under practical limitations, laboratory-based scale model measurement results are utilized. In addition, various non-uniform sensor arrangements are tested to get an insight into how to best arrange a given number of sensors to cover the largest possible area and obtain the best possible lateral resolution.

The concept of a subsurface imaging technique based on Unattended Ground Sensor Networks operating in the VHF range using ultra-wideband waveforms was recently proposed in. In this approach a forward model for realistic subsurface environment based on the Dyadic Green's function for a stratified medium and an inversion technique using an ultra-wideband near-field focusing were presented. Simulation results showed that very good lateral and depth resolution could be achieved. Before carrying out an experiment to test the proposed technique, three vital aspects of the work which are imposed by practical limitations are investigated and presented in this paper. First, the sensitivity analysis to assess the signal penetration depth in realistic subsurface environments for various frequencies is performed. Analysis of the frequency requirements of the inversion as they relate to depth resolution is also analyzed. A semi-analytic soil dielectric model originally devised for microwave frequencies is extended to VHF and validated using measurement results available in literature.

A realistic forward model based on a near-ground distributed sensor grid for a target buried under soil is devised. The soil medium is modeled as a planar stratified medium with a complex dielectric constant profile. The Dyadic Green's function for multilayer dielectric is used to compute the electric field at the target. A high resolution inversion algorithm based on phase-conjugation approach to detect deeply submerged targets is also presented. The sensor network is setup to enhance the lateral resolution by forming a synthetic aperture. The depth resolution is improved by using bandwidth. So as to improve the efficiency of the search algorithm a soil dielectric model to predict the real and imaginary parts of the dielectric constant from the volumetric soil moisture for a given soil textural composition is extended to the frequency of interest (VHF) resulting in less number of variables.