Full Text
<article class="scholarly-article">
<h2>Introduction</h2>
<p>The agricultural sector is undergoing a profound transformation, often referred to as Agriculture 4.0, driven by the integration of digital technologies. Innovations such as the Internet of Things (IoT) and sophisticated sensor networks are revolutionizing data collection, enabling unprecedented insights into crop health, environmental conditions, and operational efficiencies [25, 26]. This paradigm shift promises to enhance productivity, optimize resource utilization, and promote sustainability. However, the vast amounts of data generated within these smart farming ecosystems present significant challenges, particularly concerning data privacy and security [24]. Farm-specific data, including details about crop yields, pest infestations, and farming practices, is often highly sensitive and proprietary. Traditional approaches to data analysis, which typically involve aggregating data from multiple sources into a central repository, are often infeasible or undesirable due to privacy concerns, competitive disadvantages, and data ownership issues. Effective pest recognition is crucial for mitigating crop losses and ensuring food security, but collaborative efforts to build robust pest identification models are hindered by these data privacy barriers.</p><p>Federated Learning (FL) has emerged as a promising paradigm to address these challenges. FL enables the training of machine learning models across multiple decentralized data sources without requiring the raw data to be shared or aggregated. Instead, only model updates or parameters are communicated, preserving the confidentiality of sensitive farm-level data [1, 4, 7, 10, 14, 20]. This decentralized approach is particularly well-suited for multi-farm data ecosystems where privacy is paramount. While FL offers a robust solution for privacy-preserving collaborative analysis, its application in smart farming for tasks like pest recognition requires careful consideration of specific challenges. These include the heterogeneity of agricultural data across different farms, the potential for data imbalance, communication overhead in distributed settings, and the need for robust and secure FL frameworks, potentially enhanced with technologies like blockchain for added trust and security [2, 9, 15, 21, 22].</p><p>This paper aims to explore the landscape of federated learning frameworks and their applicability for privacy-preserving pest recognition in multi-farm data ecosystems. We will review existing FL architectures and discuss their suitability for agricultural applications, considering the unique constraints and requirements of smart farming. Furthermore, we will examine the challenges and opportunities associated with deploying FL in such environments, paving the way for more secure, collaborative, and intelligent pest management solutions that contribute to the advancement of resilient and data-secure smart farming systems.</p>
<h2>Literature Review</h2>
<p>The burgeoning field of smart farming is increasingly reliant on data-driven insights to optimize agricultural practices. However, the sensitive nature of farm-specific data, particularly concerning pest infestations, presents significant privacy challenges for collaborative analysis. Federated Learning (FL) has emerged as a promising paradigm to address these concerns by enabling distributed model training without centralizing raw data. Research has extensively explored FL frameworks for privacy-preserving analytics in various distributed data ecosystems [1, 4, 6, 10, 11, 12, 13, 14, 18, 19, 20]. These frameworks allow multiple entities (e.g., farms) to collaboratively train a shared model, contributing only model updates rather than their sensitive local data [7].</p><p>FL architectures can be broadly categorized into horizontal, vertical, and transfer FL, depending on the data partitioning across participants. For image-based recognition tasks like pest identification, where farms might have similar feature spaces but different sample instances (e.g., different pest species or stages), horizontal FL is often the most applicable [14]. Vertical FL is more suitable when participants share the same sample space but have different feature spaces. Transfer FL addresses scenarios with limited data overlap and differing feature spaces, which could be relevant for specialized pest recognition models.</p><p>The integration of FL with other emerging technologies is a significant area of research aimed at enhancing security and trust. Blockchain technology, in particular, has been explored to provide a decentralized, immutable ledger for managing model updates and ensuring transparency in FL systems [9, 15, 21]. Such integration can create robust platforms with multi-layered security, safeguarding against malicious actors and ensuring data privacy [2]. Furthermore, concepts like multi-brain federated learning are being investigated to build collaborative, privacy-preserving models across diverse domains [3].</p><p>Beyond core FL, related areas are also gaining traction. TinyML, focusing on deploying machine learning models on resource-constrained edge devices, is crucial for enabling on-farm data processing and reducing communication overhead, which is particularly relevant for agricultural IoT deployments [27]. The broader landscape of modern computing is also influenced by challenges related to distributed systems, data management, and security [28], which are pertinent to the development of large-scale smart farming data ecosystems. The combination of distributed ledger technologies with AI and IoT is also being explored for applications like fresh food logistics, highlighting the trend towards interconnected smart systems [29].</p><p>While FL offers significant privacy advantages, challenges remain. Data heterogeneity across farms, including variations in sensor data, lighting conditions, and pest prevalence, can impact model performance and robustness [24, 25]. Communication overhead during model aggregation and the need for robust privacy guarantees against sophisticated attacks are also critical considerations [13, 17]. Addressing these challenges is essential for the successful deployment of FL in multi-farm data ecosystems for effective pest recognition and overall smart farming advancement [26].</p>
<h2>Methodology</h2>
<p>This section outlines the conceptual framework for a federated learning (FL) system designed for privacy-preserving pest recognition within a multi-farm data ecosystem. The framework aims to enable collaborative model training without compromising the confidentiality of individual farm data. The proposed architecture is illustrated in <figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/federated-learning-frameworks-for-privacy-preserving-pest-recognition-in-multi-farm-data-ecosystems-j77rw/figure-1-1779808601293.octet-stream" alt="Conceptual Federated Learning Framework Diagram" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 1. Conceptual Federated Learning Framework Diagram</figcaption></figure>.</p>
<h3>Framework Components</h3>
<p>The framework comprises several key components, each addressing specific aspects of the federated learning process:</p>
<h4>Data Partitioning and Distribution</h4>
<p>Raw pest image data (e.g., images of affected crops, pest identification) is inherently distributed across multiple farms. Each farm acts as a data node, holding its local dataset. Data partitioning is thus natural, with each farm responsible for its own data. The heterogeneity of agricultural data across farms, including variations in pest types, imaging conditions, and data volumes, is a critical consideration.</p>
<h4>Local Model Training</h4>
<p>Each farm trains a local pest recognition model on its private dataset. This involves using a machine learning model, typically a deep learning architecture such as a Convolutional Neural Network (CNN), adapted for image classification tasks. The local training process generates model updates (e.g., gradients or model weights) that encapsulate the knowledge learned from the farm's data.</p>
<h4>Secure Aggregation of Model Updates</h4>
<p>Instead of sharing raw data, only the computed model updates are transmitted from individual farms to a central server or an aggregation mechanism. This aggregation process is designed to be secure and privacy-preserving. Techniques such as secure multi-party computation (SMC) or differential privacy can be integrated at this stage to further protect the individual contributions from being inferred from the aggregated updates [17]. Blockchain technology can also be employed to enhance the security, transparency, and immutability of the aggregation process [2, 9, 15, 21, 22].</p>
<h4>Global Model Deployment and Inference</h4>
<p>The aggregated model updates are used to refine a global pest recognition model. This global model represents the collective knowledge learned across all participating farms. Once trained, the global model can be deployed back to individual farms for local inference, enabling real-time pest identification and management without requiring data to leave the farm.</p>
<h3>Federated Learning Algorithms and Privacy-Enhancing Techniques</h3>
<p>Several FL algorithms can be adapted for this framework. The most common is Federated Averaging (FedAvg), which iteratively aggregates model weights from local clients [7]. Other variations, like Federated Proximal (FedProx), can help address issues of non-IID data distributions across farms [11].</p>
<p>To bolster privacy guarantees, the following techniques are considered:</p>
<ul>
<li><strong>Differential Privacy (DP):</strong> Adding carefully calibrated noise to model updates before aggregation can provide formal privacy guarantees against reconstruction attacks [4, 7].</li>
<li><strong>Secure Multi-Party Computation (SMC):</strong> Enables multiple parties to jointly compute a function (e.g., model aggregation) over their inputs while keeping those inputs private [13].</li>
<li><strong>Homomorphic Encryption (HE):</strong> Allows computations to be performed on encrypted data, meaning model updates can be aggregated while remaining encrypted [11].</li>
<li><strong>Blockchain Integration:</strong> As mentioned, blockchain can provide a decentralized, transparent, and auditable ledger for managing model updates and ensuring trust among participants [2, 9, 15, 21, 22].</li>
</ul>
<h3>Evaluation Metrics</h3>
<p>The performance of the proposed FL framework will be evaluated using standard metrics for image classification and pest recognition, alongside metrics that assess the effectiveness of the privacy-preserving mechanisms:</p>
<p><strong>Model Performance:</strong></p>
<table>
<thead>
<tr>
<th>Metric</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>Accuracy</td>
<td>Overall proportion of correctly identified pest instances.</td>
</tr>
<tr>
<td>Precision</td>
<td>Proportion of correctly identified positive identifications (e.g., true pests) out of all instances predicted as positive.</td>
</tr>
<tr>
<td>Recall (Sensitivity)</td>
<td>Proportion of true positive instances that were correctly identified.</td>
</tr>
<tr>
<td>F1-Score</td>
<td>The harmonic mean of precision and recall, providing a single measure of accuracy.</td>
</tr>
</tbody>
</table>
<p><strong>Privacy Guarantees:</strong></p>
<table>
<thead>
<tr>
<th>Metric</th>
<th>Description</th>
</tr>
</thead>
<tbody>
<tr>
<td>Epsilon (ε) and Delta (δ) (for DP)</td>
<td>Quantify the privacy loss. Lower values indicate stronger privacy.</td>
</tr>
<tr>
<td>Membership Inference Attack Success Rate</td>
<td>Measures the ability of an attacker to determine if a specific data point was used in training.</td>
</tr>
<tr>
<td>Model Inversion Attack Success Rate</td>
<td>Measures the ability of an attacker to reconstruct private data samples from model updates.</td>
</tr>
</tbody>
</table>
<p>The framework's robustness will also be assessed against challenges such as data imbalance, communication overhead, and variations in data quality across different farms [10, 24].</p>
<figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/federated-learning-frameworks-for-privacy-preserving-pest-recognition-in-multi-farm-data-ecosystems-j77rw/figure-2-1779808605039.octet-stream" alt="Federated Learning Data Flow Diagram" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 2. Federated Learning Data Flow Diagram</figcaption></figure>
<p>The integration of these components and techniques aims to create a robust and trustworthy FL system for pest recognition, fostering collaboration while upholding strict data privacy standards essential for smart farming ecosystems [1, 5, 6, 12, 14, 18, 19, 20].</p>
<h2>Results</h2>
[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object],[object Object]
<h2>Discussion</h2>
<p>The findings of this study underscore the significant potential of federated learning (FL) frameworks to enable privacy-preserving pest recognition within multi-farm data ecosystems. By decentralizing model training, FL addresses the critical need to protect sensitive farm-specific data, a major hurdle for traditional centralized machine learning approaches. The ability to train models locally and aggregate only model updates offers a robust solution for data confidentiality and ownership concerns, which are paramount in smart farming [24].</p><p>However, the practical implementation of FL in real-world agricultural settings presents several challenges. Data heterogeneity, a pervasive issue in agriculture, can significantly impact model performance and convergence. Farms often have unique environmental conditions, crop types, pest variations, and data collection methodologies, leading to non-IID (non-independent and identically distributed) data distributions across participants [25]. This heterogeneity can exacerbate communication bottlenecks, as the aggregation of diverse model updates may require more complex algorithms or lead to slower convergence. Furthermore, the scalability of FL solutions is a concern when dealing with a large number of farms, each potentially contributing varying amounts of data and computational resources.</p><p>To address these challenges, several strategies can be explored. Techniques for handling data imbalance and non-IID data, such as federated averaging with adaptive optimization or personalized FL approaches, are crucial. Enhancing communication efficiency through model compression, gradient quantization, or asynchronous updates can mitigate communication overheads [13, 20]. The integration of blockchain technology, as explored in some frameworks [2, 9, 15, 21, 22], can further bolster security, transparency, and trust by providing an immutable ledger for tracking model updates and participant contributions, thereby addressing some of the security implications in smart farming [24].</p><p>The role of edge computing is also pivotal in enhancing FL for agricultural applications. Deploying FL models on edge devices or farm-level servers can reduce the reliance on constant network connectivity, minimize latency, and enable real-time pest detection and decision-making. This aligns with the trend towards distributed AI and TinyML [27], where processing occurs closer to the data source. Remote sensing technologies [30], coupled with FL, can provide broader data coverage and enable large-scale pest monitoring, with FL ensuring that the raw high-resolution imagery remains on the farm.</p><p>Future research should focus on developing more robust and adaptive FL algorithms specifically tailored for agricultural data. Investigating novel privacy-enhancing techniques beyond differential privacy and secure multi-party computation, such as homomorphic encryption, will be essential to strengthen privacy guarantees [11, 17]. The development of standardized FL frameworks for agriculture, along with comprehensive benchmarking of different approaches under realistic farm conditions, is also necessary. Furthermore, exploring the integration of FL with other emerging technologies like digital twins [29] could lead to more sophisticated and predictive pest management systems.</p><p>The following table summarizes key FL frameworks and their potential applications in agriculture:</p>
<figure class="article-figure"><img src="https://smnxsewcdnayrztrrghn.supabase.co/storage/v1/object/public/journal-assets/scholarly/federated-learning-frameworks-for-privacy-preserving-pest-recognition-in-multi-farm-data-ecosystems-j77rw/figure-3-1779808612001.octet-stream" alt="Federated Learning Frameworks for Privacy-Preserving Pest Recognition" loading="lazy" style="max-width:100%;height:auto;" /><figcaption>Figure 3. Federated Learning Frameworks for Privacy-Preserving Pest Recognition</figcaption></figure>
<p>The table below outlines the challenges and potential solutions for implementing FL in multi-farm data ecosystems:</p>
<table>
<thead>
<tr>
<th>Challenge</th>
<th>Description</th>
<th>Potential Solutions</th>
</tr>
</thead>
<tbody>
<tr>
<td>Data Heterogeneity</td>
<td>Variations in data distribution, quality, and characteristics across farms (non-IID data) [25].</td>
<td>Personalized FL, robust aggregation algorithms, data augmentation on local farms.</td>
</tr>
<tr>
<td>Communication Overhead</td>
<td>Large model sizes and frequent updates can strain network bandwidth, especially in rural areas.</td>
<td>Model compression, gradient quantization, asynchronous updates, efficient communication protocols [13, 20].</td>
</tr>
<tr>
<td>Scalability</td>
<td>Managing a large number of participating farms and their computational resources.</td>
<td>Hierarchical FL, optimized resource allocation, efficient participant selection strategies.</td>
</tr>
<tr>
<td>Privacy and Security</td>
<td>Ensuring robust protection of raw farm data and model integrity against potential attacks [24].</td>
<td>Differential privacy, secure multi-party computation, homomorphic encryption, blockchain integration [2, 9, 15, 21, 22].</td>
</tr>
<tr>
<td>Model Robustness</td>
<td>Ensuring model performance across diverse farm environments and pest types.</td>
<td>Transfer learning, domain adaptation techniques, continuous model retraining with new data.</td>
</tr>
</tbody>
</table>
<p>The following table provides a comparative overview of different FL architectures relevant to pest recognition:</p>
<table>
<thead>
<tr>
<th>FL Architecture</th>
<th>Key Features</th>
<th>Suitability for Pest Recognition</th>
</tr>
</thead>
<tbody>
<tr>
<td>Federated Averaging (FedAvg)</td>
<td>Simple aggregation of model weights.</td>
<td>Baseline; effective for homogeneous data but may struggle with significant heterogeneity.</td>
</tr>
<tr>
<td>Federated Stochastic Gradient Descent (FedSGD)</td>
<td>Aggregates gradients rather than weights.</td>
<td>Can be more communication-efficient but requires more frequent synchronization.</td>
</tr>
<tr>
<td>Personalized Federated Learning (PFL)</td>
<td>Trains a global model while allowing local models to adapt to specific farm data.</td>
<td>Highly suitable for addressing data heterogeneity across farms [3].</td>
</tr>
<tr>
<td>Federated Transfer Learning (FTL)</td>
<td>Leverages pre-trained models and fine-tunes them on local farm data.</td>
<td>Beneficial when local datasets are small or for adapting models to new pest types.</td>
</tr>
<tr>
<td>Blockchain-enhanced FL</td>
<td>Integrates blockchain for enhanced security, transparency, and auditability.</td>
<td>Addresses trust and security concerns, crucial for collaborative data ecosystems [2, 9, 21, 22].</td>
</tr>
</tbody>
</table>
<p>In conclusion, FL offers a promising pathway towards privacy-preserving pest recognition in smart farming. While challenges related to data heterogeneity, communication, and scalability persist, ongoing research and technological advancements, including edge computing and blockchain integration, are paving the way for more secure, efficient, and robust solutions. Further exploration into hybrid approaches and tailored algorithms will be critical for realizing the full potential of FL in transforming agricultural data ecosystems.</p>
<h2>Conclusion</h2>
<p>This research has demonstrated the significant potential of federated learning (FL) frameworks in addressing the critical challenge of privacy-preserving pest recognition within multi-farm data ecosystems. By enabling collaborative model training without direct data sharing, FL empowers farms to collectively enhance pest identification capabilities while strictly safeguarding their sensitive, proprietary data. This decentralized approach aligns with the growing need for data security and ownership in smart farming, fostering trust and facilitating broader participation in data-driven agricultural advancements (Tsion, 2023; Śmietanka et al., 2020; Bonawitz et al., 2021). The integration of FL, potentially augmented by blockchain technologies for enhanced security and transparency (Mahmood & Jusas, 2022; MengJuan & Jiang, 2022; Jiang et al., 2023), offers a robust solution for building intelligent pest recognition systems that respect farm-level privacy. While challenges such as data heterogeneity and communication overhead persist, the benefits of shared intelligence and enhanced privacy preservation are substantial. Ultimately, the adoption of FL frameworks promises to accelerate the development and deployment of sophisticated, privacy-aware pest management strategies, thereby contributing to more resilient, sustainable, and data-secure agricultural practices across the sector (Gupta et al., 2020; Araújo et al., 2021).</p>
<h2>References</h2>
<ol class="references">
<li>Tsion, A. (2023). Federated Deep Learning for Privacy-Preserving Analytics in Distributed Data Ecosystems. <em>American International Journal of Computer Science and Technology</em>, <em>5</em>. https://doi.org/10.63282/3117-5481/aijcst-v5i6p101</li>
<li>Mahmood, Z., Jusas, V. (2022). Blockchain-Enabled: Multi-Layered Security Federated Learning Platform for Preserving Data Privacy. <em>Electronics</em>, <em>11</em>(10), 1624. https://doi.org/10.3390/electronics11101624</li>
<li>Kundu, S. (2023). Multi-Brain Federated Learning for Decentralized AI: Collaborative, Privacy-Preserving Models Across Domains. <em>Journal of Artificial Intelligence, Machine Learning and Data Science</em>, <em>1</em>(2), 2559-2562. https://doi.org/10.51219/jaimld/subhasis-kundu/547</li>
<li>Śmietanka, M., Pithadia, H., Treleaven, P. (2020). Federated Learning for Privacy-Preserving Data Access. <em>SSRN Electronic Journal</em>. https://doi.org/10.2139/ssrn.3696609</li>
<li>Woubie, A., Backstrom, T. (2021). Federated Learning for Privacy-Preserving Speaker Recognition. <em>IEEE Access</em>, <em>9</em>, 149477-149485. https://doi.org/10.1109/access.2021.3124029</li>
<li>Śmietanka, M., Pithadia, H., Treleaven, P. (2021). Federated learning for privacy-preserving data access. <em>International Journal of Data Science and Big Data Analytics</em>, <em>1</em>(2), 1. https://doi.org/10.51483/ijdsbda.1.2.2021.1-13</li>
<li>Bonawitz, K., Kairouz, P., McMahan, B., Ramage, D. (2021). Federated Learning and Privacy. <em>Queue</em>, <em>19</em>(5), 87-114. https://doi.org/10.1145/3494834.3500240</li>
<li>Mandala, V. (2017). Federated Mesh Architectures for Privacy-Preserving Data Engineering in Multi-Cloud Environments. <em>Global Research and Development Journals</em>. https://doi.org/10.70179/2dvv3v83</li>
<li>MengJuan, C., Jiang, W. (2022). Federated Learning with Blockchain for Privacy-Preserving Data Sharing in Internet of Vehicles. <em>SSRN Electronic Journal</em>. https://doi.org/10.2139/ssrn.4166488</li>
<li>Treleaven, P., Smietanka, M., Pithadia, H. (2022). Federated Learning: The Pioneering Distributed Machine Learning and Privacy-Preserving Data Technology. <em>Computer</em>, <em>55</em>(4), 20-29. https://doi.org/10.1109/mc.2021.3052390</li>
<li>Elahi, M., Cui, H., Kaosar, M. (2023). Survey: An Overview on Privacy Preserving Federated Learning in Health Data. <em>Computer Networks and Communications</em>. https://doi.org/10.37256/cnc.1120231992</li>
<li>Jitendra Singh Chouhan, Amit Kumar Bhatt, Nitin Anand (2023). Federated Learning; Privacy Preserving Machine Learning for Decentralized Data. <em>Tuijin Jishu/Journal of Propulsion Technology</em>, <em>44</em>(1), 167-169. https://doi.org/10.52783/tjjpt.v44.i1.2234</li>
<li>Shakeer, S. M., Babu, M. R. (2024). A Study of Federated Learning with Internet of Things for Data Privacy
and Security using Privacy Preserving Techniques. <em>Recent Patents on Engineering</em>, <em>18</em>(1). https://doi.org/10.2174/1872212117666230112110257</li>
<li>Thota, S. (2023). Federated Learning Approaches for Privacy-Preserving Artificial Intelligence in Distributed Cloud Environments. <em>International Journal of Artificial Intelligence, Data Science, and Machine Learning</em>, <em>4</em>(3), 118-127. https://doi.org/10.63282/3050-9262.ijaidsml-v4i3p114</li>
<li>Polu, O. R. (2023). QUANTUM-RESILIENT AND BLOCKCHAIN-ENHANCED FEDERATED LEARNING IN CLOUD ECOSYSTEMS FOR ADVANCED PRIVACY-PRESERVING AI. <em>INTERNATIONAL JOURNAL OF INFORMATION TECHNOLOGY AND MANAGEMENT INFORMATION SYSTEMS</em>, <em>14</em>(2), 58-67. https://doi.org/10.34218/ijitmis_14_02_008</li>
<li>Han, X., Wang, L., Wu, J., Fang, X. (2024). Data Valuation for Vertical Federated Learning: A Model-free and Privacy-preserving Method. <em>SSRN Electronic Journal</em>. https://doi.org/10.2139/ssrn.4682439</li>
<li>Novikova, E., Fomichov, D., Kholod, I., Filippov, E. (2022). Analysis of Privacy-Enhancing Technologies in Open-Source Federated Learning Frameworks for Driver Activity Recognition. <em>Sensors</em>, <em>22</em>(8), 2983. https://doi.org/10.3390/s22082983</li>
<li>Verma, A., Gonzalez, M. (2023). Privacy-Preserving Federated Learning for Healthcare Data Sharing. <em>International Journal of Recent Advances in Engineering and Technology</em>, <em>12</em>(2), 14-20. https://doi.org/10.65521/intjournalrecadvengtech.v12i2.122</li>
<li>Sondinti, L. R. K., Yasmeen, Z. (2022). Analyzing Behavioral Trends in Credit Card Fraud Patterns: Leveraging Federated Learning and Privacy-Preserving Artificial Intelligence Frameworks. <em>Universal Journal of Business and Management</em>, <em>2</em>(1), 38-49. https://doi.org/10.31586/ujbm.2022.1224</li>
<li>PHONG, L. T., PHUONG, T. T., WANG, L., OZAWA, S. (2024). Frameworks for Privacy-Preserving Federated Learning. <em>IEICE Transactions on Information and Systems</em>, <em>E107.D</em>(1), 2-12. https://doi.org/10.1587/transinf.2023mui0001</li>
<li>Jiang, W., Chen, M., Tao, J. (2023). Federated learning with blockchain for privacy-preserving data sharing in Internet of vehicles. <em>China Communications</em>, <em>20</em>(3), 69-85. https://doi.org/10.23919/jcc.2023.03.006</li>
<li>Salah, K., Rehman, M. H. u., Nizamuddin, N., Al‐Fuqaha, A. (2019). Blockchain for AI: Review and Open Research Challenges. <em>IEEE Access</em>, <em>7</em>, 10127-10149. https://doi.org/10.1109/access.2018.2890507</li>
<li>Porambage, P., Okwuibe, J., Liyanage, M., Ylianttila, M., Taleb, T. (2018). Survey on Multi-Access Edge Computing for Internet of Things Realization. <em>IEEE Communications Surveys & Tutorials</em>, <em>20</em>(4), 2961-2991. https://doi.org/10.1109/comst.2018.2849509</li>
<li>Gupta, M., Abdelsalam, M., Khorsandroo, S., Mittal, S. (2020). Security and Privacy in Smart Farming: Challenges and Opportunities. <em>IEEE Access</em>, <em>8</em>, 34564-34584. https://doi.org/10.1109/access.2020.2975142</li>
<li>Araújo, S. O., Peres, R. S., Barata, J., Lidon, F. C., Ramalho, J. C. (2021). Characterising the Agriculture 4.0 Landscape—Emerging Trends, Challenges and Opportunities. <em>Agronomy</em>, <em>11</em>(4), 667-667. https://doi.org/10.3390/agronomy11040667</li>
<li>Mowla, M. N., Mowla, N., Shah, A. F. M. S., Rabie, K. M., Shongwe, T. (2023). Internet of Things and Wireless Sensor Networks for Smart Agriculture Applications: A Survey. <em>IEEE Access</em>, <em>11</em>, 145813-145852. https://doi.org/10.1109/access.2023.3346299</li>
<li>Abadade, Y., Temouden, A., Bamoumen, H., Benamar, N., Chtouki, Y., Hafid, A. (2023). A Comprehensive Survey on TinyML. <em>IEEE Access</em>, <em>11</em>, 96892-96922. https://doi.org/10.1109/access.2023.3294111</li>
<li>Gill, S. S., Wu, H., Patros, P., Ottaviani, C., Arora, P., Pujol, V. C. (2024). Modern computing: Vision and challenges. <em>Telematics and Informatics Reports</em>, <em>13</em>, 100116-100116. https://doi.org/10.1016/j.teler.2024.100116</li>
<li>Boas, J. L. d. V., Rodrigues, J. J. P. C., Alberti, A. M. (2022). Convergence of Distributed Ledger Technologies with Digital Twins, IoT, and AI for fresh food logistics: Challenges and opportunities. <em>Journal of Industrial Information Integration</em>, <em>31</em>, 100393-100393. https://doi.org/10.1016/j.jii.2022.100393</li>
<li>Victor, N., Maddikunta, P. K. R., Mary, D. R. K., Murugan, R., Chengoden, R., Gadekallu, T. R. (2024). Remote Sensing for Agriculture in the Era of Industry 5.0—A Survey. <em>IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing</em>, <em>17</em>, 5920-5945. https://doi.org/10.1109/jstars.2024.3370508</li>
</ol>
</article>