Software-Defined Networking (SDN) requires adaptive policy generation to ensure satisfactory Quality of Service (QoS) and Quality of Experience (QoE) expectations under dynamic network conditions. While generative AI can potentially automate the optimization of network configuration, there is a lack of methods for AI-driven policy automation and enforcement, particularly in translating high-level network intent into suitable service function chains using P4 switch configurations without misconfigurations. In this paper, we present a novel framework, viz., NetPrompt, that uses Large Language Models (LLMs) for automated and intent-driven policy generation in SDN in the context of a video streaming application. By integrating prompt engineering and structured model refinement, pre-trained NetPrompt adaptively selects the appropriate LLM configuration to generate suitable P4 scripts that align with user requirements, such as dynamic QoS adaptation. We validate NetPrompt in network emulators and advanced compute/network testbed environments, including Mininet, Chameleon Cloud, and FABRIC, to construct practical network topologies for evaluation against key performance metrics such as latency reduction, throughput improvement, and error rate minimization. Our experimental results demonstrate that NetPrompt reduces misconfigurations significantly, showcasing its potential in dynamic policy management of programmable networks.

Frequency-domain watermarking is highly effective in copyright authentication for various forms of multimedia data, including but not limited to images, videos, and audios. However, it faces vulnerability to vertex attacks, such as map cropping and vertex deletion, when applied to vector maps. To enhance the robustness of frequency-domain watermarking for vector maps, this study proposes a new scheme based on virtual vertices and frequency coefficients ratios. Prior to watermark embedding, virtual vertices are inserted between each pair of adjacent vertices. Next, the discrete wavelet transform is employed to decompose x-and y-coordinates sequences, which include both actual and virtual coordinates. The ratios between high frequency coefficients of the x-and y-coordinates sequences are computed and selected as the embedding domain. To minimize distortion from watermark embedding, the ratios are amplified and modulated to conceal the watermark information. Furthermore, a watermark correction method is devised to address virtual vertices inconsistency, further enhancing robustness. Experimental results demonstrate that the proposed watermark correction method improves robustness performance by up to 35% compared to the uncorrected baseline. The proposed scheme also exhibits remarkable robustness against vertex deletion, vertex addition, and map cropping, achieving bit error rate ranging from 0% to 9.65%. Furthermore, the watermark’s visibility is well controlled, with error statistics approaching 0, thereby ensuring its imperceptibility.

Effective training in Machine Learning and Deep Learning models necessitates datasets that provide sufficient patterns and contextual information, particularly crucial in IoT networks. Imbalanced datasets, however, significantly challenge the performance of Autonomous Intrusion Detection Systems (IDS), leading to suboptimal detection rates for minority classes. In this paper, we address this issue by utilizing various Generative Adversarial Network (GAN) models, including WGANGP, CGAN, CTGAN, and CWGANGP, to generate synthetic data that balance these imbalanced datasets. We evaluate the performance of IDS models trained on GAN-augmented datasets against those trained on unbalanced datasets, considering metrics such as fitting duration, generation duration, accuracy, precision, recall, and F1 score. Our findings reveal substantial improvements in IDS performance with the application of GANs across binary, general, and specific attack classifications. Additionally, we compare the effectiveness of GANs with classical sampling algorithms, such as SMOTE and Random Oversampling. This comprehensive evaluation underscores the potential of GANs as a sophisticated solution for improving IDS accuracy and reliability in handling complex and highly imbalanced datasets.

Unmanned aerial vehicles (also known as drones) equipped with high-resolution video cameras have become increasingly popular for applications such as public safety and smart farming. However, inefficient configurations in drone video analytics due to misconfigured edge networks can lead to degraded video quality and inefficient resource utilization. In this paper, we propose a novel scheme for network edge orchestration that utilizes both offline and online learning-based approaches to achieve pertinent selections of network protocols and video properties in multi-drone-based video analytics. Our approach utilizes both supervised and unsupervised machine learning algorithms to make decisions regarding network protocols and video properties during the pre-takeoff stage of the drones (i.e., offline stage). Additionally, our approach incorporates a reinforcement learning-based multi-agent deep Q-network algorithm for drone trajectory optimization during flights (i.e., online stage) and a memory-to-memory multi-hop data forwarding strategy for drone swarm video transmission. Our evaluation results demonstrate that our offline orchestration approach can suitably choose network protocols (i.e., among TCP/HTTP, UDP/RTP, QUIC), while our unsupervised learning approach outperforms existing methods and achieves efficient offloading while improving network performance (i.e., throughput and round-trip time) by at least 25%, with satisfactory video quality. Furthermore, we demonstrate through trace-based and real-field experiment testbeds how our online orchestration in terms of decision-making and data forwarding strategies achieves 91% of the oracle baseline network throughput performance with comparable video quality. Overall, our approach offers a promising solution for optimizing drone video analytics and enhancing the overall performance of drone-swarm-based applications.

For next-generation logistics management, robotic vehicles such as autonomous ground robots and aerial drones can alleviate the strain on last-mile distribution. They can help avoid on-road congestion, navigate hard-to-reach locations, and parallelize delivery operations. However, as the robotic vehicles move in a given delivery area, environmental barriers e.g., trees or buildings, affect air-to-air (A2A), air-to-ground (A2G), ground-to-ground (G2G) network communications on a hybrid truck-drone-robot system. In this paper, we present an environmentally-aware cooperative network routing computation scheme to avoid obstacle blockage in A2A/A2G/G2G network communications for addressing large-scale coordinated operations of the hybrid truck-drone-robot system. Specifically, we propose an offline policy-based routing algorithm and two online extensions (i.e., heuristics and learning-based) to solve the hybrid last-mile delivery vehicles communication problem in order to trade-off between end-to-end communication (i.e., increase network throughput) and delivery efficiencies (i.e., lower parcel delivery time consumption). We evaluate our scheme using state-of-the-art network routing algorithms in a trace-based simulator that integrates both the vehicles and networking sides. Performance evaluation results from our simulations show that: (i) our offline approach is Pareto-optimal among non-learning supported algorithms in a pre-delivery scenario, and (ii) our RL-based online algorithm achieves between 85–96 % of the Oracle strategy performance during delivery procedures.

Collaborative Drone systems (CDS) have the potential to benefit a variety of application areas such as agriculture, military operations, surveillance, and disaster response. At the same time, CDS can pose challenges due to their limited flight time impacted by battery capacities, and constrained edge computation capabilities on-board the drones. Furthermore, an understudied subject relates to when drones in a CDS trust each other to accomplish a task, resulting in new vulnerabilities that can be exploited via cyber attacks. In this paper, we propose a novel trust quantification methodology in a CDS with intelligence-driven edge routing, which can help detect malicious nodes in a CDS that compromise communication and disrupt the functionality of packet forwarding. Our approach for trust quantification is guided by a CDS vulnerability analysis that characterizes impact due to the presence of two malicious threat agents viz., flooder node and faker node. Detection of these threat agents in a CDS is aided by trust quantification in the form of trust scores obtained by using a Bayesian Network model that allows for decision-making on CDS nodes’ trust levels. We validate our trust quantification methodology in ns-3 based simulation experiments and show how we can categorize nodes based on different thresholds of trust scores with varying sensitivities, which helps in the detection of CDS threat agents.

Unmanned aerial systems (UAS) consist of a swarm of unmanned aerial vehicles (UAVs) with edge resources and collaboration with ground-control-servers (GCS) are useful for heavy computation use cases e.g., traffic management, public safety, and disaster response management. Inefficient setups and collaboration decisions, often stemming from edge/cloud network misconfigurations, can lead to suboptimal resource utilization and delayed response times. In this paper, we present a novel scheme for (soft) real-time learning-based UAS-Edge-Server collaboration and orchestration strategies to achieve pertinent allocations of both computation resources and communication strategies. Our approach includes i) policy-based pre-application collaboration and benchmark analysis as well as ii) learning-based multi-agent deep Q-network (DQN) algorithm that optimizes UAV swarm trajectories during application. Evaluation results demonstrate that our policy-based approach Pareto-optimally trade-off performance (e.g., accuracy, streaming) and disaster response time. In addition, our DQN approach significantly enhances edge-cloud resource cooperation, improving network performance metrics like throughput and round-trip time by a minimum of 12% compared to state-of-the-art edge-internet-of-things (EIoT) collaboration algorithms. Furthermore, through real-world emulations, we illustrate how our orchestration attains 87% of the Oracle baseline network throughput performance while maintaining a comparable disaster response time for various video analytics-based disaster scenarios.

In a disaster response management (DRM) scenario, communication and coordination are limited, and absence of related infrastructure hinders situational awareness. Unmanned aerial vehicles (UAVs) or drones provide new capabilities for DRM to address these barriers. However, there is a dearth of works that address multiple heterogeneous drones collaboratively working together to form a flying ad-hoc network (FANET) with air-to-air and air-to-ground links that are impacted by: (i) environmental obstacles, (ii) wind, and (iii) limited battery capacities. In this paper, we present a novel environmentally-aware and energy-efficient multi-drone coordination and networking scheme that features a Reinforcement Learning (RL) based location prediction algorithm coupled with a packet forwarding algorithm for drone-to-ground network establishment. We specifically present two novel drone location-based solutions (i.e., heuristic greedy, and learning-based) in our packet forwarding approach to support application requirements. These requirements involve improving connectivity (i.e., optimize packet delivery ratio and end-to-end delay) despite environmental obstacles, and improving efficiency (i.e., by lower energy use and time consumption) despite energy constraints. We evaluate our scheme with state-of-the-art networking algorithms in a trace-based DRM FANET simulation testbed featuring rural and metropolitan areas. Results show that our strategy overcomes obstacles and can achieve 81-to-90% of network connectivity performance observed under no obstacle conditions. In the presence of obstacles, our scheme improves the network connectivity performance by 14-to-38% while also providing 23-to-54% of energy savings in rural areas; the same in metropolitan areas achieved an average of 25% gain when compared with baseline obstacle awareness approaches with 15-to-76% of energy savings.

In this paper, we describe how we extended the Pegasus Workflow Management System to support edge-to-cloud workflows in an automated fashion. We discuss how Pegasus and HTCondor (its job scheduler) work together to enable this automation. We use HTCondor to form heterogeneous pools of compute resources and Pegasus to plan the workflow onto these resources and manage containers and data movement for executing workflows in hybrid edge-cloud environments. We then show how Pegasus can be used to evaluate the execution of workflows running on edge only, cloud only, and edge-cloud hybrid environments. Using the Chameleon Cloud testbed to set up and configure an edge-cloud environment, we use Pegasus to benchmark the executions of one synthetic workflow and two production workflows: CASA-Wind and the Ocean Observatories Initiative Orcasound workflow, all of which derive their data from edge devices. We present the performance impact on workflow runs of job and data placement strategies employed by Pegasus when configured to run in the above three execution environments. Results show that the synthetic workflow performs best in an edge only environment, while the CASA - Wind and Orcasound workflows see significant improvements in overall makespan when run in a cloud only environment. The results demonstrate that Pegasus can be used to automate edge-to-cloud science workflows and the workflow provenance data collection capabilities of the Pegasus monitoring daemon enable computer scientists to conduct edge-to-cloud research.

Smart agriculture benefits from unmanned aerial vehicles (UAV), and in-field sensors to collect data used to make responsible crop management decisions which sustainably increase yields. In addition, smart agriculture relies on machine learning algorithms, creative networking solutions, and edge and cloud computing resources to collect, transfer, and process agricultural data. UAV can carry a wide array of sensors, maneuver rapidly throughout the field, apply treatments for some crop health problems, and can be flown by software. UAV, however, have small batteries and limited carrying capacities which keep missions short. In this paper, we provide an overview of state-of-the-art UAV swarm technology for smart agriculture, and present experiences from real-world agricultural UAV swarm case studies. We describe how quick mapping of large areas such as crop fields necessitates multiple UAV missions, potentially using multiple UAV simultaneously as a swarm. We detail how swarms of UAV have added advantages over a single UAV deployment. They can coordinate to map areas in parallel, leverage multiple sensor types, target areas for close inspection, and diagnose and treat problems rapidly. UAV swarms come with additional implementation difficulties beyond single UAV. We list challenges to implementers in terms of Resource allocation, compute orchestration, multi-agent mission planning and swarm goal definition. We also describe recent advances in edge computing, machine learning, and autonomy in orchestration and resource management techniques for swarm deployments. Finally, we conclude with research opportunities that future work can address to improve swarm performance, scale, and adoption for smart agriculture.

Unmanned aerial vehicles (a.k.a. drones) with high-resolution video cameras are useful for applications in e.g., public safety and smart farming. Inefficient configurations in drone video analytics applications due to edge network miscon-figurations can result in degraded video quality and inefficient resource utilization. In this paper, we present a novel scheme for offline/online learning-based network edge orchestration to achieve pertinent selection of both network protocols and video properties in multi-drone based video analytics. Our approach features both supervised and unsupervised machine learning algorithms to enable decision making for selection of both network protocols and video properties in the drones’ pre-takeoff stage i.e., offline stage. In addition, our approach facilitates drone trajectory optimization during drone flights through an online reinforcement learning-based multi-agent deep Q-network algorithm. Evaluation results show how our offline orchestration can suitably choose network protocols (i.e., amongst TCP/HTTP, UDP/RTP, QUIC). We also demonstrate how our unsupervised learning approach outperforms existing learning approaches, and achieves efficient offloading while also improving the network performance (i.e., throughput and round-trip time) by least 25% with satisfactory video quality. Lastly, we show via trace-based simulations, how our online orchestration achieves 91% of oracle baseline network throughput performance with comparable video quality.

Unmanned aerial vehicles or drones provide new capabilities for disaster response management (DRM). In a DRM scenario, multiple heterogeneous drones collaboratively work together forming a flying ad-hoc network (FANET) instantiated by a ground control station. However, FANET air-to-air and air-to-ground links that serve critical application expectations can be impacted by: (i) environmental obstacles, and (ii) limited battery capacities. In this paper, we present a novel obstacle-aware and energy-efficient multi-drone coordination and networking scheme that features a Reinforcement Learning (RL) based location prediction algorithm coupled with a packet forwarding algorithm for drone-to-ground network establishment. We specifically present two novel drone location-based solutions (i.e., heuristic greedy, and learning-based) in our packet forwarding approach to support heterogeneous drone operation as per application requirements. These requirements involve improving connectivity (i.e., optimize packet delivery ratio and end-to-end delay) despite environmental obstacles, and improving efficiency (i.e., by lower energy use and time consumption) despite energy constraints. We evaluate our scheme by comparing it with state-of-the-art networking algorithms in a trace-based DRM FANET simulation testbed. Results show that our strategy overcomes obstacles and can achieve between 81-90% of network connectivity performance observed under no obstacle conditions. With obstacles, our scheme improves network connectivity performance by 14-38 % while also providing 23-54% of energy savings.

Unmanned aerial vehicles (UAV) or drone systems equipped with cameras are extensively used in different surveillance scenarios and often require real-time control and high-quality video transmission. However, unstable network situations and various transport protocols may result in impairments during video streaming, which in turn negatively impacts user's quality of experience (QoE). In this paper, we propose a dynamic computation offloading and control framework, named DyCOCo, based on image impairment detection under various available network bandwith conditions. Our DyCOCo framework demo features IoT devices in a testbed setup on the GENI infrastructure. Our demo results show that our DyCOCo approach can efficiently choose the suitable networking protocols and orchestrate both the camera control on the drone, and the computation offloading of the video analytics over limited edge computing/networking resources.

Computer vision applications are increasingly used on mobile Internet-of-Things (IoT) devices such as drones. They provide real-time support in disaster/incident response or crowd protest management scenarios by e.g., counting human/vehicles, or recognizing faces/objects. However, deployment of such applications for real-time video analytics at geo-distributed areas presents new challenges in processing intensive media-rich data to meet users' Quality of Experience (QoE) expectations, due to limited computing power on the devices. In this paper, we present a novel policy-based decision computation offloading scheme that not only facilitates trade-offs in performance vs. cost, but also aids in offloading decision to either an Edge, Cloud or Function-Centric Computing resource architecture for real-time video analytics. To evaluate our offloading scheme, we decompose an existing computer vision pipeline for object/motion detection and object classification into a chain of container-based micro-service functions that communicate via a RESTful API. We evaluate the performance of our scheme on a realistic geo-distributed edge/core cloud testbed using different policies and computing architectures. Results show how our scheme utilizes state-of-the-art computation offloading techniques to Pareto-optimally trade-off performance (i.e., frames-per-second) vs. cost factors (using Amazon Web Services Lambda pricing) during real-time drone video analytics, and thus fosters effective environmental situational awareness.