Federated learning in food research

Federated learning in food research ☆ Zuzanna Fendor * , Bas H.M. van der Velden , Xinxin Wang, Andrea Jr. Carnoli , Osman Mutlu, Ali Hürriyeto˘ glu Wageningen Food Safety Research (WFSR), Part of Wageningen University & Research, Akkermaalsbos 2, 6708 WB, Wageningen, the Netherlands ARTICLE INFO Keywords: Federated learning Food Food safety Data privacy Machine learning Literature review ABSTRACT The use of machine learning in food research is sometimes limited due to data sharing obstacles such as data ownership and privacy requirements. Federated learning is a technique to potentially alleviate these obstacles because it allows to train machine learning models locally, keeping the data private and sharing only the learned parameters. In this review we investigate the use of federated learning in food research. First, we outline a framework that describes the variants of federated learning implementations. Then, we provide an overview of applications of federated learning in food research. Next, we discuss the performance of the models trained with federated learning, and reasons for the use of federated learning. Finally, we categorize the encountered federated learning applications within the federated learning framework. In the discussion, we highlight the knowledge gaps and discuss the potential novel applications. In this review we examined a total of 86 papers published between 2019 and 2024. The current applications encompass crop disease monitoring, yield prediction, quality assessment, and pesticide residue risk analysis. We observed the general trend of centralized horizontal federated learning, and identified the absence of vertical federated learning, federated transfer learning, and decentralized architectures as research gaps. 1. Introduction Lack of access to safe and nutritious food is detrimental to the health and wellbeing of people. The sustainable development goals aim to end hunger, achieve food security, improve nutrition, and promote sustainable agriculture [1]. One approach to contribute to this goal is to use machine learning models to gain insight in current issues and help to predict future scenarios [2,3]. Machine learning models are algorithms that learn patterns from data to perform a task, such as classifying the content of an image. For instance, a model can be trained to determine whether the fruit in an image is a pear or an apple. The model learns by making adjustments to its parameters based on the data to improve its performance. To ensure adequate and reliable performance, a lot of data is needed to reflect the variety of scenarios, or the model will be biased. When the data originates from just one source, a model might have a limited view of reality. For instance, an apple orchard might have access to many examples of apples, but just a few examples of pears, making the model better at recognizing apples rather than pears. Thus, data sharing between different data sources is beneficial to the size and variety of datasets, and to the model performance. Data sharing can be challenging when multiple data owners are involved. This is especially the case with private or sensitive data. Data sharing can be challenging due to legal restrictions and regulations concerning privacy, technical limitations, and data owners being competitors. In the food domain, data owners can be reluctant to share their data due to fear for liability, bad publicity, or loss of business advantage over competitors [4]. Using federated learning, a machine learning model can be trained without disclosing this private or sensitive data [5–7]. In other words, federated learning is a type of privacy-preserving machine learning. Within federated learning the data and the model training are local (i.e., the data does not leave the data’s owner storage), and only the model parameters are shared for model improvement [5]. ☆ Preprint citation: This manuscript was previously uploaded as a preprint, see Fendor et al. (2024). Fendor Z., Van der Velden B.H.M, Wang X., Carnoli A Jr., Mutlu O., Hürriyeto˘ glu A. (2024). Federated learning in food research. arXiv. https://doi.org/10.4855 0/arXiv.2406.06202. * Corresponding author. Akkermaalsbos 2, Gebouw 123, 6708 WB, Wageningen, the Netherlands. E-mail address: [email protected] (Z. Fendor). Contents lists available at ScienceDirect Journal of Agriculture and Food Research journal homepage: www.sciencedirect.com/journal/journal-of-agriculture-and-food-research https://doi.org/10.1016/j.jafr.2025.102238 Received 9 April 2025; Received in revised form 11 July 2025; Accepted 4 August 2025 Journal of Agriculture and Food Research 23 (2025) 102238 Available online 5 August 2025 2666-1543/© 2025 The Authors. Published by Elsevier B.V. This is an open access article under the CC BY license ( http://creativecommons.org/licenses/by/4.0/ ). Machine learning modeling in food research often involves considering data scattered across various research facilities or competing commercial producers reluctant to share their information. Therefore, privacy-preserving machine learning methods, like federated learning, can be very useful for this field. Surprisingly, despite the presence of extensive reviews of federated learning applications have been published in other fields (e.g. agriculture, smart cities, air pollution monitoring, and medicine) ([8–10]; D. C. [11–15]), we have not found a broader review of federated learning in food research. Our work fills this gap by summarizing current achievements, applications, and reasons for the use of federated learning in food research and highlighting encountered pitfalls. The main research questions we aim to answer in this review are. •“What food research problems are currently being solved by federated learning?” •“What are the gaps of federated learning use in food research?” We structured this paper as follows. In section 2, we define the federated learning framework. Section 3provides an overview of the current application space of federated learning in food and categorizes all included papers into this framework. In section 4, we give an outlook for future use of federated learning in the food domain. 2. Federated learning framework The core principle of federated learning is uniquely defined as training the model on locally stored data and only sharing the model parameters. Multiple variants of federated learning exist, and their implementation is detailed by the challenge at hand. In this section we will explain the essential variants of the federated learning framework that we will also use to categorize the papers examined. Federated learning was introduced as “a decentralized approach of leaving the data distributed on the mobile devices and learning a shared model by locally computed updates’’ [16]. Following this definition federated learning is applied when each computer (i.e., client node) possesses and does not share their own data. In the federated learning context, each client node trains a machine learning model on their local and private data. Then they share the model parameters, which are aggregated to yield a global model. Such an aggregation step can be performed using different strategies, the simplest of which is averaging [17]. Finally, each local model is updated with the parameters of the new global model. The process of sending the global model to the client nodes, training, and aggregating the results repeats until a maximum number of repetitions or a predefined performance minimum (e.g., model accuracy) is reached. Federated learning has been expanded to a broader range of variants, since its introduction by McMahan et al. 18–24]. We categorize federated learning approaches by type of server architecture, the clients, and data partitioning to form a framework. We divide server architecture into centralized and decentralized [22,24], type of client nodes into cross-device and cross-silo [18,19,24], and data partitioning amongst clients into horizontal federated learning, vertical federated learning, and federated transfer learning [24]. We chose to characterize federated learning using a framework that highlights its collaborative aspect. Our framework focused on the larger questions of who the participants are (type of client), what each participant contributes (data partitioning), and how they communicate (server architecture). Alternative frameworks for categorizing federated learning exist, for instance including privacy mechanism, applicable machine learning model, and methods for solving systems heterogeneity [23]. This type of categorization is focused on the technical details of the implementation. 2.1. Server architecture The distinction between centralized and decentralized federated learning (Fig. 1) is determined by the presence of a central server. In centralized federated learning, a central server is responsible for the aggregation of the results from the client nodes. Such a server is the communication and control center that facilitates the management and regulation of the federated learning process. Letting the server manage the training process presents several advantages. First, a central server releases the client nodes from the burden of both storage and aggregation of the models [22]. Second, the communication between the central server and the client nodes is generally better regulated and protected from eventual malicious attacks [22]. Finally, the presence of only one central server makes the federated learning system easier to operate and manage. Unfortunately, regulating the training process through a single central server has the disadvantage of introducing a single point of failure. In other words, if the central server fails the federated learning process stops, halting the training process because the local results cannot be aggregated. Decentralized federated learning includes many combinations of client nodes and related connections, and communication protocols such as gossip learning where the client nodes communicate one-on-one with randomly selected peers [25] In this category of federated learning, there is direct communication between the nodes. This means that the nodes share their learned model parameters among each other to continue learning on the received model or to aggregate the parameters from the received model with their own model. Decentralized federated learning makes the central server obsolete, removing a potential communication bottleneck, a single point of failure, and the need of assuming trustfulness of the central server [22]. In exchange, it increases the computational burden on the client nodes, and requires communication protocols between nodes to be implemented, which introduces new problems. The network topology (i.e., fully connected, partially connected, or node clustering) and the communication protocol (i.e., synchronous or asynchronous) used in a distributed federated learning system can influence the efficiency of the communication, but also robustness, flexibility, security, fault tolerance, and communication cost [26]. 2.2. Type of client nodes The type of client nodes, shown in Fig. 2, can be distinguished in cross-silo and cross-device [19,24]. This distinction has practical implications to the computational resources and the communication bottlenecks. Cross-silo federated learning usually refers to different organizations such as research institutions or companies that own a large, isolated data source, i.e., a data silo. The number of client nodes and therefore communications are limited, which leads to fewer communication bottlenecks. In cross-silo federated learning, there are usually few carefully selected participants, reducing the risk of a malicious client node trying to disrupt or take unfair advantage of the federated learning system. With this configuration there is also less risk of client nodes being unable to participate due to performance requirements because it is easier to control the quality of the hardware with just few participating client nodes. However, the presence of slow or unresponsive client nodes has a larger influence on model training because a participating node being dropped is more drastic in the cross-silo setting. Cross-device federated learning can involve millions of mobile or internet of things (IoT) devices. Consequently, there is less oversight on the participating devices, and, in many cases, there will be more variance in the hardware leading to reduced reliability. This type of federated learning faces larger hardware heterogeneity by design but often presents a large quantity and variety of data. Federated learning can also be applied to edge computing scenarios. Edge computing is the concept of transferring data from data sources, like sensors, IoT devices, or mobile devices, to a geographically proximate edge server (Xia et al., 2021). The edge server is an intermediate step between performing all training on the data-producing devices and Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 2 collecting the data from all the devices at one place. Federated learning in edge computing scenarios can be classified as both cross-device and cross-silo depending on the number of participating edge servers and the amount of data stored at each edge server (Xia et al., 2021). We classified all scenarios employing edge servers as cross-device, because it better reflects the aspect of online data collection from different devices with limited computing resources. 2.3. Data partitioning Federated learning can also be determined by the way data is partitioned amongst the nodes, as Fig. 3 shows. The distinction in data partitioning is important because it determines how the information is spread across the client nodes. Data partitioning can be distinguished in horizontal federated learning, vertical federated learning, and federated transfer learning [21,23,24]. In horizontal federated learning, each node has different samples with the same, or largely similar features. The aim of horizontal federated learning is to increase the number of training samples with the same set of features. For instance, multiple wheat farmers can collectively train a model to identify wheat diseases from images. The features of the data will not vary, because the wheat and the diseases are expected to look similarly across farms. But because the farmers photograph different wheat plants, the sample space becomes larger. This data partitioning can be applied to both cross-device and cross-silo federated learning [19]. In vertical federated learning there is a large overlap in samples, but the overlap in features is smaller than in horizontal federated learning. This means that the same individual sample is present across different client nodes with different features. Vertical federated learning is used to increase the feature space of the training data [27]. To illustrate, suppose a shipping company and an apple sauce factory possess information about the same batch of apples. The shipping company owns information about the conditions of the apples during the transport. The apple sauce factory collects information about the quality of apples, together with descriptive measures such as weight. The two companies can employ vertical federated learning to collaboratively train a model for the prediction of the quality of apples. Vertical federated learning typically assumes a different training protocol from horizontal federated learning: all parties need to align the samples in their data, a process called entity alignment [20]. Often the label, in this example the quality of apples, is held by one of the client nodes which acts as a central server. Each client node trains their local part of the model that learns to represent their features. With this federated learning configuration, the intermediate results from the models of each client are incorporated by the central server into a combined model. Then, the central server assesses the combined model by comparing how well the labels match the predictions. Finally, each client node receives client node-specific feedback from the central server (Y. [20]). The participation of all client nodes is necessary for making model predictions, because the Fig. 1. Server archtecture types in federated learning. Figure on the left is a schematic representation of a centralized federated learning architecture. The client nodes communicate their local model updates to the central server that aggregates the results into one model. The aggregated model is communicated back to the client nodes. Figure on the right is an example of a decentralized federated learning architecture, where the client nodes communicate the model updates to each other. The aggregation of the models is performed by the client nodes. Fig. 2. Types of client nodes in federated learning. Figure on the left displays cross-silo federated learning. In cross-silo federated learning a few organizations, instances, or companies that own large private datasets (data silos) collaborate to train a shared model. Figure on the right represents cross-device federated learning. Cross-device federated learning involves data collected and stored by many (thousands or millions) devices, possibly owned by the same organization. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 3 model needs input from the features spread across the clients. In federated transfer learning, neither the samples nor the features overlap, or they overlap very slightly [28]. Federated transfer learning aims to improve both the number of samples and the number of features by combining data from different, yet related domains, sources or tasks. Federated transfer learning trains the models in such a way that the mutual information between the tasks is leveraged to help both of them. For example, two companies that collect animal movement sensor data from different breeds of goat and different types of enclosure (free-range rocky terrain vs commercial farm) will own data of different individuals with different behavioral patterns. The free-range goats may display more jumping. The raw data from the commercial goats may be dampened due to the soft terrain. Federated transfer learning enriches the model with more knowledge of different features of goat behavior encountered in different situations. This approach leverages the knowledge acquired from one domain to the other [29]. After the completion of the federated learning training, the model can be fine-tuned to fit the needs of the client better, while still retaining knowledge gathered from the other related domain. Vertical federated learning and federated transfer learning typically are applied in a cross-silo setting. These partitioning types require additional communication to accommodate entity matching. Moreover, each client node needs to be identifiable, which is not necessarily the case in the cross-device setting [19]. 3. Federated learning in food This section summarizes our findings of federated learning in food. Section 3.1 describes the methods and the search strategy. Section 3.2 summarizes the found literature from different perspectives. Section 3.2.1 provides an overview of federated learning applications in food research. Section 3.2.2 touches upon federated learning model performance in found literature. Section 3.2.3 lists reasons for applying federated learning to food research. Finally, Section 3.3.4 uses the federated learning framework from Section 2 to summarize the use of federated learning in food research. 3.1. Methods We selected the literature for this review systematically. We considered a paper as eligible if it contained a use case of federated learning in the food domain. The papers were ideally an implementation of the use case; however, some proposed solutions were also included as long as they were concrete enough. We excluded papers that only contained a brief generic mention like “can be applied to agriculture”, and those that were not written in English. We obtained the papers through Google Scholar, Scopus, Web of Science, IEEE Xplore, and CAB abstracts on the August 15, 2024. We included the exact search queries and their descriptions in the appendix. The search query had two components. The first component included the terms: federated learning, federated approach. The second component included the terms related to food that can be roughly categorized into supply-chain related terms, adulteration related terms, various individual food items, microbial hazards, chemical hazards, and others. We kept the list of included terms as complete as possible. We limited the search to the title, the abstract, and the keywords. The review focused on “classic” peer-reviewed literature, however encountered “non-classic” literature (i.e., dissertations, conference papers or preprint articles) that met the criteria was also included. Due to the importance of conference proceedings in fast-moving fields such as computer science, we included conference papers on par with journal articles. References to dissertations were explicitly mentioned as such. Additionally, we searched for papers from 2016 onwards, because 2016 was the year of the first publication on federated learning. In total we collected 548 papers across two rounds. Two researchers checked the papers independently for relevance by reading the titles and abstracts. On top of this, we scanned the references in relevant articles for other relevant literature (snowball approach); however, this did not lead to the discovery of any earlier unseen publications. 3.2. Results We included 86 distinct papers. The publication dates span between 2019 and 2024. Out of those 86 papers, 9 (10 %) proposed but did not implement federated learning. Fig. 4 shows the reviewing process. 3.2.1. Application overview Fig. 5 provides a comprehensive overview of federated learning applications across various domains in food research, summarizing the found literature on the topic, highlighting six major categories: food safety hazards, food quality, food production process monitoring and optimization, food fraud, food supply chain, and agriculture-oriented applications. Table 1 shows the detailed list of publications for each category. 3.2.1.1. Food safety hazards 3.2.1.1.1. Chemical hazards. Federated learning applicability in monitoring chemical food safety hazard is critical, as monitoring contaminants such as pesticide residues and heavy metals—known to cause adverse health effects when exceeding safety limits [112]—helps ensure the safety of the food supply. In case of pesticide residues, the only publicly available data is often only the pass rate - proportion of tested food samples that comply with regulation of certain foods [30]. This restricted data availability makes it challenging to conduct research on bioaccumulation of pesticides because products are only described in a qualitative way: under or above the maximum allowed pesticide residue limit. The sum of smaller quantities that individually are compliant with the residue limit cannot be measured this way. The actual quantitative values related to the Fig. 3. Data partitioning in federated learning. Horizontal federated learning: each local dataset shares features (weight and color) but contains apples from different farms. Vertical federated learning: The model combines data of the same apple entity that is matched across datasets. Each dataset contains different features. In this example the features are collected at different sites: a farm and a factory. Federated transfer learning: similar to vertical federated learning, however there is less overlap between the samples. Instead of entity matching all the samples used for training, the process often involves learning a shared feature space. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 4 amount of pesticide residues in food are often scattered across different departments and are not sharable due to data ownership. The only study that makes use of federated learning in this field is that of J. Yu et al. [30] in which the aim is to assess the pesticide residue risk in fruits and vegetables in China. To enable the usage of those data silos, the authors create an entropy risk model that can be computed in parts at each local dataset and combined at the central server. The model gives a more nuanced view on the pesticide residue risks of foods compared to the pass rate based on the qualitative values. This federated learning approach enables more informative predictions of pesticide risk, while keeping the data private. Heavy metals are another chemical hazard for human health [113]. One federated learning application for detecting heavy metals in soil uses near-infrared spectroscopic data from IoT remote sensing [32]. Horizontal federated learning is applied to an IoT decentralized sensing system to deal with the large scale of sensing by removing the need to collect data from all areas. However, the details of the implementation were not reported in this paper. A perspective paper [31] proposes detecting and predicting heavy metal contamination of soil and water using federated learning using an edge cloud server. Federated learning is presented as a potential way of facilitating the use of various data types, satellite imagery in combination of in-situ sampling data, gathered at the edges of the edge cloud server while keeping the sources private. 3.2.1.1.2. Biological hazards. Qian et al. [4] reviewed potential federated learning applications in biological food safety domains, particularly those involving pathogen contamination. Their review paper proposes a wide range of potential food safety application domains for federated learning and other private data sharing methods. The authors identify key food safety data types with varying privacy concerns, and propose, amongst others, federated learning as a solution to enhance models by privately sharing data across organizations. For example, graphic information system models that combine microbial data and environmental data, which are costly to gather, can benefit from federated learning to improve spatial and temporal data distribution without exposing sensitive information [4]. Quantitative microbial risk assessment models for, among others, the prediction of foodborne illness or recall likelihoods can reduce uncertainty and improve validation by securely integrating data from multiple sources [4]. Agent-based models for pathogen prediction and public health models for foodborne disease outbreaks can benefit from federated learning by data sharing to improve the parameter estimates and to improve data varieties [4]. 3.2.1.2. Food quality. Federated learning has been applied in water quality assessment and monitoring, described in the papers by Park et al. [33], Vellingiri et al. [34], and Zhu [35]. Clean water is not only essential for consumption, but it is also needed for agriculture. If contaminated, irrigation water can lead to food hazards in crops [114]. That is why essential water sources, such as natural water bodies or water from treatment companies, ought to be frequently monitored. The traditional type of water quality monitoring requires more hands-on personnel [115]. Furthermore, some water bodies are difficult or dangerous for personnel to reach due to factors such as remote locations. [35]. These factors make using federated learning to automate the process beneficial. Park et al. [33] tackle the problem of monitoring and prediction of green tide; an overgrowth of algae in water. The data was gathered on the nation-wide scale in South Korea using smart IoT sensors, edge servers, and a central cloud. The IoT sensors enable monitoring of various green tide indicators such as temperature, pH, and dissolved oxygen. The data cannot feasibly be sent to a central server in real-time, due to the large geographic spread of the sensor locations and the limited range of the sensor signals. Furthermore, it is not efficient to have a single access point for data collection. This is why multiple local Fig. 4. Flowchart of the paper selection and reviewing process. The reviewers looked at the title and the abstract first to assess the relevance. If the relevance could not be determined on the title and abstract alone, the text body was scanned for more information. Fig. 5. Bar chart showing the number of papers per federated learning category. The colors represent a grouping of categories (from top to bottom in chart): dark green: Food safety hazards, light green: Food quality, light blue: Food production process monitoring and optimization, light pink: Food fraud light brown: Food supply chain, dark brown: Agriculture oriented applications. (For interpretation of the references to color in this figure legend, the reader is referred to the Web version of this article.) Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 5 Table 1 Federated learning applications in food research. “*” indicates articles that proposed the use of federated learning for a certain goal without conducting experiments. Application Subapplication Authors (year) Modality Data type Data Partition Food safety hazards Chemical [30] Tabular/time series Pesticide detection Horizontal   [31]* Imaging, Tabular Several types of (unspecified) data    [32] Imaging NIR spectra of soil samples Horizontal Biological [4]* Various Various  Food quality Food quality [33] Tabular/time series Sensor data Horizontal   [34] Tabular Sensor data Horizontal   [35] Imaging Camera monitoring Horizontal   [36] Tabular Sensor data Horizontal   [37] Tabular Sensor data, spectral data Horizontal   [38] Imaging Ultraviolet visible and near infrared spectrograms Horizontal   [39]* Imaging Spectral images  Food production process monitoring and optimization Animal monitoring [40] Imaging Chicken images Horizontal   [41] Tabular Sensor data Horizontal   [42]* Tabular Milk content, behavior, genome data    [43] Tabular Sensor data Horizontal   [44] Imaging Spectrogram (visual representation of sound recordings) Horizontal Process monitoring and optimization [45] Time series Ultrasonic sensor Transfer learning   [46] Tabular Irrigation demand, evapotranspiration, infiltration from cropland Horizontal   [47] Time series Water consumption Unspecified   [48] Imaging Water foam images Horizontal   [49] Tabular World states and actions Horizontal   [50]* Tabular Sensor data    [51] Tabular Sensor data Horizontal   [52]* Tabular Sensor data    [14] Tabular New York Water consumption from Kaggle Horizontal   [53] Imaging Images of weeds Horizontal   [54] Tabular NIR imaging Horizontal   [55] Tabular Sensors Horizontal   [56] Tabular Sensors Horizontal   [57] Time series Sensors Horizontal   [58] Time series Weather, water level Horizontal Food fraud Food fraud [59] Tabular Fraud type, product category, year, origin country, control country Horizontal Food supply chain Food supply chain [60] Tabular Sampling data of daily food supervision and management from the National Market Supervision Administration Horizontal/ Vertical Security of food systems [61] Time series Network traffic, sensor, actuator Horizontal   [62] Time series Sensor data Horizontal   [63] Time series Traffic data Horizontal   [64] Tabular Approved number of blocks, online time, net stake investment Horizontal   [65] Imaging Food images Horizontal   [66] Tabular Network traffic data, sensor data Horizontal   [67] Sensor Temperature, ripeness Horizontal   [68] Time series Traffic data Horizontal   [69] Sensor Soil, temperature, nutrients Transfer learning Consumer side [70] Tabular Beer preferences Horizontal   [71] Tabular Survey, genome Horizontal Agriculture oriented applications Digitalization [72]* Sensor data    [73]* Imaging, Sensor Fruit images and E-Nose data  Crop disease [74]* Imaging, Sensor Rice images    [75] Imaging Strawberry images Horizontal   [76] Imaging Strawberry images Horizontal   [77] Imaging Strawberry images Horizontal   [78] Imaging Banana leaf images Horizontal   [79] Imaging Banana leaf images Horizontal   [80] Imaging Banana leaf images Horizontal   [81] Imaging Banana leaf images Horizontal   [82] Imaging Guava images Horizontal   [83] Imaging Jute images Horizontal   [84] Imaging Sunflower images Horizontal   [85] Imaging Parsley images Horizontal   [86] Imaging Coffee images Horizontal   [87] Imaging Wheat images Horizontal   [88] Imaging Maize images Horizontal   [89] Imaging UAV-gathered images Horizontal/ Vertical (continued on next page) Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 6 edge servers are used to collect the sensor data. Horizontal federated learning was applied to use the sensor data on the edge servers to train models locally and aggregate the parameters into a global model. Another water quality application focuses on predicting pollution of the Cauvery River in India [34]. The choice for applying federated learning is motivated from the data privacy perspective and focuses more on enabling collaboration between different parties, in this case water monitoring sites. This data is similar to the data from Park et al. [33], with the difference of the data being collected at monitoring site-level, making it a cross-silo problem. Kathen et al. [36] use a fleet of autonomous surface vehicles for water quality monitoring by scouting water bodies and looking for pollution peaks. They use centralized federated learning with edge server structure in the second stage of their research, where they gather more detailed data and train specialized models for each zone of the water body. After this, the models from different zones are used to update the centrally learned initial model. The final water quality example by Zhu [35], employs sensors, in this case cameras, edge servers, and centralized horizontal federated learning setting. The author uses a convolutional neural network, which is a computer vision model, to assess the quality of the water based on chromatic information in image data. Federated learning is also applied to milk quality assessment in liquid [37] and powder [38] form. The milk quality is dependent on its contents and can be analyzed using light spectra from various ranges of light. Vimalajeewa et al. [37] use Mid-Infrared Spectroscopy samples for liquid milk quality assessment. This work compares the state-of-the-art techniques for analyzing Mid-Infrared Spectroscopy data to their own composite model in both federated and non-federated settings. Gulati et al. [38] use ultraviolet illumination visible spectral images collected for both pure and ad-hoc adulterated skimmed milk powder. The authors use this case study as a proof-of-concept for combining blockchain technology and federated learning in supply chains to achieve a federated learning workflow that ensures protection against malicious attacks on performance from both the server and the clients. Such a federated learning architecture transfers some of the functionality of a centralized server to a distributed ledger that can choose which local models to aggregate and when to aggregate them based on predefined criteria. Finally, the perspective review by Müller-Maatsch et al. [39] suggests the usage of federated learning for improving the usability of portable miniature spectral devices for measuring safety, authenticity, and quality of various food commodities. The challenge they address is the large bottleneck caused by the need to calibrate and re-calibrate such instruments; a task for which the necessary databases are often device specific, food specific, or contamination specific. Due to such heterogeneity across these datasets, eventual similarities between databases are not being leveraged to their full potential. In consequence, every change is met with a lot of repetitive work. A federated learning network is proposed to link a data sample to different data sources such as reference methods database and spectral imaging database. 3.2.1.3. Food production process monitoring and optimization. Federated learning supports diverse applications in food production process monitoring and optimization, including animal monitoring (e.g., dayage classification, health, and welfare), assessing dairy cattle genetic potential, and monitoring beer fermentation among other related tasks. Machine learning models have been widely used in this domain. However, developing accurate machine learning models requires large and diverse datasets that a single company may not possess. Collaboration can contribute to better models, but sharing data with competitors raises privacy and competitive concerns. Federated learning addresses this issue by combining data sources securely, enhancing product quality, efficiency, and innovation while maintaining competitive boundaries and data privacy. 3.2.1.3.1. Animal monitoring. The first application we encountered was chicken day-age classification by Y. Huang et al. [40]. Knowing the exact age of a chicken helps to assess the health of the animal, to choose the optimal breeding cycle, feeding cycle, and slaughter time of respectively chickens to reduce costs and environmental pollution. The model takes chicken images and trains them in a federated manner on horizontally partitioned local data. Each local model is individualized through filter pruning, and federated learning is applied as a form of distributed learning. Another aspect to monitor is the animal health and welfare that can be assessed through observing the animal’s behavior. Animal activity recognition traditionally relies on time-consuming and labor-intensive visual and behavioral observation [116]. Nowadays, it can also be automated using wearable motion data sensors on the animals, as shown Table 1 (continued) Application Subapplication Authors (year) Modality Data type Data Partition   [90] Imaging Orchard and apple images Horizontal   [91] Imaging Fruit images Horizontal   [92] Imaging Hyperspectral tomato images Horizontal   [93] Imaging UAV-gathered images Horizontal   [94] Imaging Rice leaf images Horizontal   [95] Imaging Orange images Horizontal   [96] Imaging Plant images Horizontal   [97] Tabular, Imaging, Sensor Air and soil humidity and temperature data, soil NPK and EC, and camera HSVs Horizontal   [98] Imaging PlantVillage dataset (images from different crops) Horizontal   [99] Imaging Wheat images Horizontal   [58] Imaging PlantVillage dataset (images from different crops) Horizontal   [100] Imaging Rice leaf images Horizontal   [101] Imaging Wheat images Horizontal Yield prediction [102] Tabular Weather and soil sensor data Horizontal    Imaging Satellite images Horizontal   [103] Tabular Sensor data Horizontal   [104] Tabular Breeding test phenotypes, environmental meteorology Vertical   [105] Tabular Weather, soil, environment sensors Horizontal   [106] Tabular Climatic data Horizontal   [107] Tabular Weather and soil sensor data Horizontal   [108] Imaging UAV-gathered images Horizontal Food image classification [109] Imaging Food images Transfer learning   [110] Imaging Fruit images Horizontal   [111] Imaging Near-infrared hyperspectral pasture images Horizontal Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 7 from Mao et al. [41] and Arshad et al. [43]. The raw data can be interpreted with models trained to classify the behaviors into categories like eating, sleeping, walking etcetera. Mao et al. [41] highlight that, although federated learning has the potential of enlarging the size of available data, animal behavior patterns can vary within individuals and between farms leading to client drift, where each client starts to deviate from one another during local training. Furthermore, gradient conflicts might occur during aggregation. Consequently, this work proposes solutions to both problems to enhance convergence towards a global model applicable to all individual animals. We can find another application of federated learning relatable to animal behavior in the work of Borgianni et al. [44], where the authors built a federated learning model to predict the presence of honeybees in hives based on the sound that can be recorded on-site. Gengler [42] proposes utilizing sensor data, usually meant for management applications, to assess and improve the genetic potential of dairy cattle. One of the challenges listed was the governance and ownership of the sensor data. The author hypothesizes that federated learning will create opportunities for new animal evaluation methods, for instance through training on continuously updated data. The training based on the raw sensor data can result in phenotypic sensor algorithms or even federated genomic prediction models that could enhance breeding research. 3.2.1.3.2. Process monitoring and optimization. Next application addresses the challenges of the monitoring of fermentation processes of beer using ultrasonic sensors and temperature. Due to the length of the fermentation process, it is difficult to collect enough data to train models for each vessel separately, especially for small and medium enterprises. The ultrasonic waves can vary each time a sensor is attached, even if the same procedure is followed and the sensor type and vessel materials stay the same [117]. Each fermentation vessel produces different patterns that are not directly transferable. Bowler et al. [45] test domain adaptation strategies to combine data obtained from a laboratory (source domain) and industrial data (target domain). Federated transfer learning is used as a privacy-friendly version of domain adaptation. Federated transfer learning is compared against combining the data and fine-tuning the models on the target domain. Akbari et al. [49] use synchronous centralized federated learning for optimizing the paths of UAV based edge devices in a way that it minimizes the age of information and the network energy usage of real time monitoring for smart farming. Another example of federated learning in smart farming can be found in Siniosoglou et al. [57], where they used a centralized, horizontal strategy to build a neural network model to predict future field conditions from historical data. Federated learning for process monitoring is also present in waterrelated research. Cui et al. [46] suggest federated learning as a future extension of their groundwater level modelling research. Federated learning is proposed for a groundwater level IoT system composed of sensors, edge servers and a central cloud server to obtain up-to-date data. Elhachmi and Kobbane [47] use a similar approach to build a federated learning linear regression model on simulated water data to predict water consumption. The research conducted by Singh et al [55] and Singh et al [56] focuses on the implementation of edge computing for the development of federated learning system for water management and irrigation, respectively. Another example of federated learning based on edge computing can be found in Thonglek and Maipradit [58]. In this study, the authors build a federated learning model to predict the water level using historical data on water and weather. Bera et al. [51] implement edge computing to reduce energy requirements and connectivity challenges related to a federated learning system for irrigation monitoring. Mato et al. [48] use models for texture segmentation to measure the amount of foam in wastewater to optimize the best moment for its removal. Horizontal, centralized cross-silo federated learning is applied on two tanks within one wastewater treatment plant to test the possibilities for extending the model to multiple wastewater treatment plants. Finally, Pei et al. [50] suggest combining federated learning with “Internet of Underwater Things” for applications such as water quality monitoring, underwater life detection, flood forecasting, underwater detection, and marine energy transfer. 3.2.1.4. Food fraud. Federated learning is applied to the food safety research on predicting food fraud based on historical data. In their paper, Gavai et al. [59] simulate a scenario where known food fraud cases for different geographic locations are confidential and not shared beyond their borders. Such experimental setting allowed to combine the food fraud data from the European Union Rapid Alert for Food and Feed 1 database and the United States Economically Motivated Adulteration 2 database. The data was divided among three clients. One client got the data from the United States database only, and the other two got the data from the European Union split by “before and after 2014”, leading to a realistic data split scenario where data is distributed in a heterogenous way. The authors train Bayesian Network models on local data, combined data, and the local data in federated learning setting to compare the federated and non-federated scenarios. 3.2.1.5. Food supply chain. This section describes the uses of federated learning across food supply chains. Food supply chains often include multiple stages from production to consumption. Data sharing across the different stages is not always possible or optimal. Data sharing may create delays [67], or privacy or security vulnerabilities. Moreover, trust is an essential aspect of supply chains [118,60]. This area lends itself to federated learning as recognized by T.H. Liu et al. [119]. Within food research, X. Zhang et al [60] look specifically into collaboration across food supply chains through the use of vertical and horizontal federated learning, paired with blockchain technology. They test their architecture in an artificial experimental setup using a cowpea safety dataset. 3.2.1.5.1. Security of food systems. Federated learning can be applied to ensure the security of the food systems. Food systems increasingly rely on IoT technology, exposing them to cybersecurity threats. For instance, an attack targeting a water pump caused a failure of a water treatment plant in Illinois in 2011 [120]. Federated learning fits well with the use of machine learning to enhance security against cyber-attacks in industrial IoT settings [61] by allowing data to remain localized. This privacy-preserving training approach addresses the stakeholder’s reluctance to share sensitive and private data. Using a real-world Secure Water Treatment dataset, Jahromi et al. [61] train a model to effectively detect previously unseen attacks in a horizontal federated scenario, combining data from different sources. A similar federated learning application is introduced by Bahadoripour et al. [66] where centralized horizontal federated learning is used for training a cyber-attack detection model for industrial control systems. The authors apply domain adaptation as a preprocessing step to map the data from each client into a shared feature space before applying federated learning, transforming it from a federated transfer learning problem to a horizontal federated learning problem. Finally, Zeng et al. [68] has introduce an adversarial sample generation method and a training method to enhance adversarial robustness of federated learning systems in the secure water treatment domain. In the face of IoT sensors being more frequently employed on farms, it is important to take the safety of food systems into consideration to prevent vulnerabilities to hacker attacks [121]. The work of P. Kumar et al. [62], Friha et al. [63], and Praharaj et al. [69] focus on this use case of the protection of agricultural IoT using federated learning intrusion models. Alrashdi et al. [65] integrate contrastive learning to enhance the model’s ability to distinguish between clean and adversarial examples of food images, thereby improving robustness of federated learning in 1 Rapid Alert for Food and Feed Window: https://webgate.ec.europa. eu/rasff-window/screen/search. 2 Economically Motivated Adulteration: https://www.fda.gov/food/compli ance-enforcement-food/economically-motivated-adulteration-food-fraud. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 8 internet of visual things systems. As federated learning may not ensure full security, additional improvements have been proposed. Bala and Kaur [64] highlight the use of federated learning and blockchain in the light of crop insurance. Crop insurance requires companies to gather personal data from the farmers. Blockchain as a storage solution is stated to offer transparency and real-time traceability, but at the same time requires a solid consensus strategy to be secure and prevent malicious actors from infiltrating the system. The authors suggest a solution using federated learning to train a model for calculating trust scores that can be used for a reliable Proof-Of-Stake mechanism for blockchain. 3.2.1.5.2. Consumer side of the food supply chain. From the consumer perspective of the food supply chain, recent research has also explored how federated learning can support modelling individual preferences, which can streamline the customization of product offerings. Z. Liu et al. [70] apply federated learning to consumer preference of beer brands. In this scenario, each simulated consumer finds different aspects of beer important. Federated learning is implemented to keep the consumers’ preferences private by only sharing their reward based on the choices made during training, without disclosing which aspects they find important. When the consumers’ preferences are modelled using the collective regret obtained through federated learning, the simulated optimal beer choice is reached faster and with less collective accumulated regret. Next, in a more recent work, Sudharson et al. [71] propose a system that utilizes user preferences collected via a survey, health goals, microbiome data, and wearable device readings to recommend informed dietary choices. 3.2.1.6. Agriculture oriented applications. Agriculture-oriented applications of federated learning indirectly influence multiple aspects of the food system such as food safety, nutrient availability, and product quality. Better yield prediction and tailored crop selection reduce pesticide reliance, preserve food safety, and promote a stable, healthful food supply. 3.2.1.6.1. Federated learning in a broader scope of agriculture digitalization. Federated learning can be viewed from the broader perspective of the digitalization of the agriculture. A literature review by Abbasi et al. [122] captures and analyses various developments in smart farming. The digital technologies that were included in the description were the IoT technology, wireless sensor networks, cloud and edge computing, autonomous robotics systems, big data and analytics, machine learning, deep learning, decision support systems, cyber-physical systems, and digital twins. Federated learning was named as a solution to “digital transformation’s cyber-security and data privacy challenges”. Salem et al. [72] introduced a future vision for smart farming in United Arab Emirates that incorporated federated learning with edge server technology to timely detect adverse conditions and automatically react to the signals to optimize farming through digitalization. 3.2.1.6.2. Crop disease. Crop disease identification and severity estimation is a prevalent use case in the food research. Here, federated learning can be useful to tackle challenges related to data sharing such as privacy concerns and communication cost. Many papers focus on image data for various fruits and vegetables, such as rice ([74,94]; Tripahty et al., 2024), strawberries([75]; N. [76]; S. [77]), bananas (A. [78]; S. [79]; S. [80]; V. [81]), guava (D. [82]), jute (K. [83]), sunflower (J. [84]), parsley [85], coffee (V. [86]), wheat [87,99], maize [88], and soybean [89]. Two papers are about disease identification in orchards [90,91]. Mamba Kabala et al. [98] use the PlantVillage dataset for an extensive exploration of federated learning for disease identification on four types of crops. They investigate the effect of using different values for number of clients, communication rounds, and iterations, as well as different neural network architectures for their federated learning setup. Finally, one paper implements a system for multiple agricultural pest identification independent of the plant species cultivated [93]. All of the papers focus on imaging data of healthy and diseased plants. Khan et al. [93], Konyar and Gahrooei [92] and C. Yu et al. [89] focus on images obtained through unmanned aerial vehicle (UAV) devices. C. Yu et al. [89] and Konyar and Gahrooei [92] consider data obtained by UAV devices and collected at the edge servers. Khan et al. [93] mentioned UAV technology as the goal, however the experiments are performed on a public dataset without simulating the UAV scenario. Mehta, Kukreja, and Gupta [88] mention the problem of specificity and the capabilities of identifying diseases on the individual plant-level being low when remote sensing like drones or satellite images are being utilized. Their solution focuses on the individual-level images of diseased plants. The remaining papers on this topic use openly accessible datasets [88] enriched with images from growers and experts [87]. 3.2.1.6.3. Yield prediction. Federated learning enables secure, collaborative yield prediction model training across diverse agricultural conditions to guide preventive measures, optimize crop variety selection, and facilitate private data utilization in both cross-silo and crossdevice settings. Three publications include yield prediction of soybean crops [102, 103] and maize crop varieties [104]. Durrant et al. [102] use the soybean prediction problem as a use case to demonstrate horizontal federated learning and its potential for data sharing in agriculture. The task is to predict the average observed yield per unit area within American counties that pose as data silos. They use sequences of remote satellite images taken before harvest, and tabular data of the weather and soil as two different data-driven scenarios. Manoj et al. [103] focus on the same task of federated soybean yield prediction and also applied centralized, horizontal federated learning in a cross-silo setting. Their motivation was to contribute to the agricultural production risk management demands by overcoming data sharing obstacles. A. Li et al. [107] use the same task to test a model pruning algorithm that respects different local distributions and increases communication efficiency with the same federated learning settings. Q. Zhang et al. [104] use vertical, cross-silo federated learning for the selection of the maize variety based on the optimal yield prediction for collaborative plant breeding. The authors use data collected from 248 trial sites that posed as client nodes. They consider varieties of maize to be the overlapping samples and phenotypic traits and environmental conditions to be features that could differ across client nodes. The authors apply a federated random forest model that requires some central knowledge of the features and samples. The central server receives encoded and encrypted aligned feature codes to perform decision-tree-type training. We identified two publications on the selection of the correct crop from a wider range of plant species based on the weather and soil conditions [105,106]. Bera et al. [105] considers ten plant species to create a crop prediction model that can be used in an app for crop recommendation called CropReco. Bera et al. [105] use federated learning to enable model training on the edge servers that gathered data directly from the sensors, which we consider to be a cross-device setting. Idoje et al. [106] investigate federated learning Gaussian Naïve Bayes models that can predict the optimal choice between rice, maize or chickpea crops based on climatic features in a horizontally partitioned dataset. 3.2.1.6.4. Food image classification. Federated learning approaches have also been explored for more specialized tasks in image classification. Ooi et al [54] show examples of federated learning aiming to improve the identification of weed plants to optimize pesticide spraying. Herbicide used on crops can be harmful to both people working in the fields and ingesting the products. V. Huang et al. [111] perform experiments to test client drop out resilience of federated learning models in the context of weed detection. The other two papers in this category [109,110] are likewise not written from the perspective of solving agricultural problems. Instead, they focus on advancing the field of federated learning with food image classification as an example use case. 3.2.2. Federated learning performance in the food research Some included papers only propose future directions of federated learning; therefore, they do not include performance results [4,31,39, Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 9 tuna OR tilapia OR bacteria OR "Bacillus cereus" OR "Campylobacter jejuni" OR "Clostridium botulinum" OR "Clostridium perfringens" OR "e. coli" OR "Escherichia coli" OR "Listeria monocytogenes" OR Shigella OR "Staphylococcus aureus" OR salmonella OR "Vibrio cholerae" OR "Vibrio parahaemolyticus" OR "Vibrio vulnificus" OR "Yersinia enterocolitica" OR "acromobacter sakazakii" OR "Bacteriophage" OR "Enteric Virus" OR "Hepatitis A virus" OR Norovirus OR "Norwalk virus" OR "Rota virus" OR prion OR "Mad Cow Disease" OR "food poisoning" OR fungus OR mycotoxin* OR aflatoxin OR Deoxynivalenol OR "Ochratoxin A" OR Fumonisin OR Patulin OR arsenic OR cadmium OR Mercury OR "heavy metals" OR pesticide* OR pest* OR instecticide* OR fungicide* OR herbicide* OR Azoxystrobin OR Captan OR Clethodim OR Thiocarbamate OR "allergenic" OR "Anaphylactic shock" OR "allergen*") OR TS=(crop* OR agriculture)) NOT TS = (“fine-grained”OR cancer) Google scholar ("federated learning" OR "federated approach") AND (food AND (safety OR monitoring OR supply OR chain OR processing OR market* OR distribution OR storage OR handling OR farm* OR fraud OR hygiene OR surveillance OR agribusiness) OR (food AND (adulteration OR contaminants OR tempering OR fraud) OR bacteria OR “Bacillus cereus” OR “Campylobacter jejuni” OR “Clostridium botulinum” OR “Clostridium perfringens” OR "e. coli" OR "Escherichia coli" OR ”Listeria monocytogenes” OR Shigella OR “Staphylococcus aureus” OR salmonella OR “Vibrio cholerae” OR “Vibrio parahaemolyticus” OR “Vibrio vulnificus” OR “Yersinia enterocolitica” OR “Cronobacter sakazakii” OR “Bacteriophage” OR “Enteric Virus” OR “Hepatitis A” virus OR Norovirus OR “Norwalk virus” OR “Rota virus” OR fungus OR mycotoxin* OR pesticide* OR pest* OR vegetable* OR fruit* OR meat OR fish OR grain OR cereal OR legume* OR milk OR dairy OR egg* OR crop* OR agriculture) CAB abstracts CAB abstracts is a database already limited to articles on agriculture and life sciences. Since using just federated learning and federated approach as search terms resulted in only 15 hits, we decided not to narrow it down any more than that. ("federated learning" OR "federated approach") Scopus (126 hits total): TITLE-ABS-KEY ("federated learning" OR "federated approach") AND (TITLE-ABS-KEY (food* AND (safety OR monitoring OR supply OR chain OR processing OR market* OR distribution OR storage OR handling OR farm* OR fraud OR hygiene OR surveillance OR agribusiness)) OR TITLE-ABS-KEY (food AND (adulteration OR contaminants OR tempering OR fraud)) OR TITLE-ABS-KEY (water OR food* OR vegetable* OR fruit* OR meat OR fish OR grain OR cereal OR legume* OR milk OR dairy OR egg* OR maize OR rice OR wheat OR flour OR potato* OR onion* OR tomato* OR lettuce OR carrot* OR pepper* OR cucumber* OR celery OR broccoli OR mushroom OR spinach OR cabbage* OR bean* OR cauliflower OR garlic OR asparagus OR banana* OR apple* OR grape* OR strawberries OR melon* OR avocado* OR blueberries OR mandarins OR oranges OR peach* OR pineapple* OR lemon* OR poultry OR chicken* OR pig* OR duck* OR bovine OR beef OR broiler OR salmon OR tuna OR tilapia OR bacteria OR "Bacillus cereus" OR "Campylobacter jejuni" OR "Clostridium botulinum" OR "Clostridium perfringens" OR "e. coli" OR "Escherichia coli" OR "Listeria monocytogenes" OR shigella OR "Staphylococcus aureus" OR salmonella OR "Vibrio cholerae" OR "Vibrio parahaemolyticus" OR "Vibrio vulnificus" OR "Yersinia enterocolitica" OR "acromobacter sakazakii" OR "Bacteriophage" OR "Enteric Virus" OR "Hepatitis A virus" OR norovirus OR "Norwalk virus" OR "Rota virus" OR prion OR "Mad Cow Disease" OR "food poisoning" OR fungus OR mycotoxin* OR aflatoxin OR deoxynivalenol OR "Ochratoxin A" OR fumonisin OR patulin OR arsenic OR cadmium OR mercury OR "heavy metals" OR pesticide* OR pest* OR instecticide* OR fungicide* OR herbicide* OR azoxystrobin OR captan OR clethodim OR thiocarbamate OR "allergenic" OR "Anaphylactic shock" OR "allergen*") OR TITLE-ABS-KEY (crop* OR agriculture)) AND NOT TITLE-ABS-KEY ("fine-grained" OR cancer) IEEE divided into: (83 hits total) 1. ("federated learning" OR "federated approach") AND (food* AND (safety OR monitoring OR supply OR chain OR processing OR market* OR distribution OR storage OR handling OR farm* OR fraud OR hygiene OR surveillance OR agribusiness)) 15 hits 2. ("federated learning" OR "federated approach") AND (food AND (adulteration OR contaminants OR tempering OR fraud)) 3. ("federated learning" OR "federated approach") AND (water OR food* OR vegetable* OR fruit* OR meat OR fish OR grain OR cereal OR legume* OR milk OR dairy OR egg* OR maize OR rice) NOT (FINE-grained OR cancer) 63 hits 4. ("federated learning" OR "federated approach") AND (OR wheat OR flour OR potato* OR onion* OR tomato* OR lettuce OR carrot* OR pepper* OR cucumber* OR celery OR broccoli OR mushroom OR spinach OR cabbage* OR bean* OR cauliflower OR garlic OR asparagus 5. ("federated learning" OR "federated approach") AND (banana* OR apple* OR grape* OR strawberries OR melon* OR avocado* OR blueberries OR mandarins OR oranges OR peach* OR pineapple* OR lemon*) 6. ("federated learning" OR "federated approach") AND (poultry OR chicken* OR pig* OR duck* OR bovine OR beef OR broiler OR salmon OR tuna OR tilapia) 7. ("federated learning" OR "federated approach") AND (bacteria OR "Bacillus cereus" OR "Campylobacter jejuni" OR "Clostridium botulinum" OR "Clostridium perfringens" OR "e. coli" OR "Escherichia coli" OR "Listeria monocytogenes" OR Shigella OR "Staphylococcus aureus" OR salmonella OR "Vibrio cholerae" OR "Vibrio parahaemolyticus" OR "Vibrio vulnificus") 8. ("federated learning" OR "federated approach") AND (OR "Yersinia enterocolitica" OR "acromobacter sakazakii" OR "Bacteriophage" OR "Enteric Virus" OR "Hepatitis A virus" OR Norovirus OR "Norwalk virus" OR "Rota virus" OR prion OR "Mad Cow Disease" OR "food poisoning") 9. ("federated learning" OR "federated approach") AND (fungus OR mycotoxin* OR aflatoxin OR Deoxynivalenol OR "Ochratoxin A" OR Fumonisin OR Patulin OR arsenic OR cadmium OR Mercury OR "heavy metals" OR pesticide* OR pest*) 5 hits 10. ("federated learning" OR "federated approach") AND (instecticide* OR fungicide* OR herbicide* OR Azoxystrobin OR Captan OR Clethodim OR Thiocarbamate OR "allergenic" OR "Anaphylactic shock" OR "allergen*") Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 16 11.2 Nomenclature Abbreviation    CAB Commonwealth Agricultural Bureaux EFRA Extreme Food Risk Analysis ENISA European Union Agency for Cybersecurity ICT Information and Communications Technology IEEE Institute of Electrical and Electronics Engineers IID Independent and Identically Distributed IoT Internet of Things LVVN Netherlands Ministry of Agriculture, Fisheries, Food Security and Nature UAV Unmanned Aerial Vehicle   11.3 Detailed categorization of literature within the federated learning framework This table contains information used for creating Fig. 6. It categorizes each study according to the three distinctions in the federated learning framework in Section 2. Reference Architecture Data Partition Cross-silo/Cross-device [75] Centralized Horizontal Cross-silo [78] Centralized Horizontal Cross-silo [49] Centralized Horizontal Cross-device [94] Centralized Horizontal Cross-device [65] Centralized Horizontal Cross-device [43] Centralized Horizontal Cross-silo [66] Centralized Horizontal Cross-silo [64] Centralized Horizontal Cross-silo [105] Centralized Horizontal Cross-silo [51] Centralized Horizontal Cross-device [44] Centralized Horizontal Cross-silo [45] Centralized Transfer learning Cross-silo [89] Centralized Horizontal /Vertical Cross-device [95] Decentralized Horizontal Cross-device [32] Unspecified Horizontal Cross-device [96] Centralized Horizontal Cross-silo [46] Centralized Horizontal Cross-device [82] Centralized Horizontal Cross-silo [90] Centralized Horizontal Cross-silo [97] Centralized Horizontal Cross-device [102] Centralized Horizontal Cross-silo [14] Centralized Horizontal Unspecified [47] Centralized Unspecified Cross-device [101] Centralized Horizontal Cross-device [63] Centralized Horizontal Cross-device [59] Centralized Horizontal Cross-silo [38] Centralized/Decentralized Horizontal Cross-silo [109] Centralized Transfer learning Cross-device [111] Centralized Horizontal Cross-device [106] Centralized Horizontal Cross-silo [84] Centralized Horizontal Cross-silo [139] Centralized Transfer learning Cross-silo [30] Centralized Horizontal Cross-silo [61] Centralized Horizontal Cross-silo [83] Centralized Horizontal Cross-silo [36] Centralized Horizontal Cross-device [53] Centralized Horizontal Cross-device [93] Centralized Horizontal Unspecified [92] Centralized Horizontal Cross-device [67] Decentralized Horizontal Cross-silo [107] Centralized Horizontal Cross-silo [98] Centralized Horizontal Cross-silo [103] Centralized Horizontal Cross-silo [41] Centralized Horizontal Cross-silo [48] centralized Horizontal Cross-silo [88] Centralized Horizontal Cross-silo [87] Centralized Horizontal Cross-silo [99] Centralized Horizontal Unspecified [76] Centralized Horizontal Cross-silo [118] Centralized Horizontal Cross-device (continued on next page) Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 17 (continued) Reference Architecture Data Partition Cross-silo/Cross-device [108] Decentralized Horizontal Cross-device [54] Centralized Horizontal Cross-device [62] Centralized Horizontal Cross-device [33] Centralized Horizontal Cross-device [50] Centralized Unspecified Cross-device [110] Centralized Horizontal Cross-silo [69] Centralized Transfer Learning Cross-device [104] Centralized Vertical Cross-silo [79] Centralized Horizontal Cross-silo [80] Centralized Horizontal Cross-silo [77] Centralized Horizontal Cross-silo [56] Centralized Horizontal Cross-device [55] Centralized Horizontal Unspecified [57] Centralized Horizontal Cross-silo [71] Centralized Horizontal Cross-device [126] Decentralized Horizontal Cross-device [58] Centralized Horizontal Cross-device [91] Centralized Horizontal Cross-silo [100] Centralized Horizontal Cross-silo [86] Centralized Horizontal Cross-silo [81] Centralized Horizontal Cross-silo [34] Centralized Horizontal Cross-silo [37] Centralized Horizontal Cross-silo [85] Centralized Horizontal Cross-silo [60] Centralized/Decentralized Horizontal/Vertical Cross-device /Cross-silo [40] Centralized Horizontal Cross-device [68] Centralized Horizontal Cross-silo [35] Centralized Horizontal Cross-device Data availability No data was used for the research described in the article. References [1] Goal 2 | department of economic and social affairs. (n.d.). Retrieved January 12, 2024, from https://sdgs.un.org/goals/goal2. [2] X. Wang, Y. Bouzembrak, A.O. Lansink, H.J. van der Fels-Klerx, Application of machine learning to the monitoring and prediction of food safety: a review, Compr. Rev. Food Sci. Food Saf. 21 (1) (2022) 416–434, https://doi.org/ 10.1111/1541-4337.12868. [3] W. Mu, G.A. Kleter, Y. Bouzembrak, E. Dupouy, L.J. Frewer, F.N. Radwan Al Natour, H.J.P. Marvin, Making food systems more resilient to food safety risks by including artificial intelligence, big data, and internet of things into food safety early warning and emerging risk identification tools, Compr. Rev. Food Sci. Food Saf. 23 (1) (2024) e13296, https://doi.org/10.1111/1541-4337.13296. [4] C. Qian, Y. Liu, C. Barnett-Neefs, S. Salgia, O. Serbetci, A. Adalja, J. Acharya, Q. Zhao, R. Ivanek, M. Wiedmann, A perspective on data sharing in digital food safety systems, in: Critical Reviews in Food Science and Nutrition, 2022, https:// doi.org/10.1080/10408398.2022.2103086. [5] H.B. McMahan, E. Moore, D. Ramage, S. Hampson, B.A.y. Arcas, CommunicationEfficient Learning of Deep Networks from Decentralized Data (arXiv: 1602.05629). arXiv, 2023, https://doi.org/10.48550/arXiv.1602.05629. [6] N. Rieke, J. Hancox, W. Li, F. Milletarì, H.R. Roth, S. Albarqouni, S. Bakas, M. N. Galtier, B.A. Landman, K. Maier-Hein, S. Ourselin, M. Sheller, R.M. Summers, A. Trask, D. Xu, M. Baust, M.J. Cardoso, The future of digital health with federated learning, npj Digit. Med. 3 (1) (2020), https://doi.org/10.1038/ s41746-020-00323-1. Article 1. [7] G.A. Kaissis, M.R. Makowski, D. Rückert, R.F. Braren, Secure, privacy-preserving and federated machine learning in medical imaging, Nat. Mach. Intell. 2 (6) (2020), https://doi.org/10.1038/s42256-020-0186-1. Article 6. [8] M. Ili´ c, M. Ivanovi´ c, Federated learning—opportunities and application challenges, in: N.T. Nguyen, J. Botzheim, L. Guly´ as, M. Nunez, J. Treur, G. Vossen, A. Kozierkiewicz (Eds.), Advances in Computational Collective Intelligence, Springer, Nature Switzerland, 2023, pp. 481–492, https://doi.org/ 10.1007/978-3-031-41774-0_38. [9] J.C. Jiang, B. Kantarci, S. Oktug, T. Soyata, Federated learning in smart city sensing: challenges and opportunities, Sensors 20 (21) (2020), https://doi.org/ 10.3390/s20216230. Article 21. [10] L. Li, Y. Fan, M. Tse, K.-Y. Lin, A review of applications in federated learning, Comput. Ind. Eng. 149 (2020) 106854, https://doi.org/10.1016/j. cie.2020.106854. [11] D.C. Nguyen, Q.-V. Pham, P.N. Pathirana, M. Ding, A. Seneviratne, Z. Lin, O. Dobre, W.-J. Hwang, Federated learning for smart healthcare: a survey, ACM Comput. Surv. 55 (3) (2022), https://doi.org/10.1145/3501296, 60:1-60:37. [12] J. Xu, B.S. Glicksberg, C. Su, P. Walker, J. Bian, F. Wang, Federated learning for healthcare informatics, Journal of Healthcare Informatics Research 5 (1) (2021) 1–19, https://doi.org/10.1007/s41666-020-00082-4. [13] K.R. ˇ Zalik, M. ˇ Zalik, A review of federated learning in agriculture, Sensors 23 (23) (2023), https://doi.org/10.3390/s23239566. Article 23. [14] M. El Hanjri, H. Kabbaj, A. Kobbane, A. Abouaomar, Federated learning for water consumption forecasting in smart cities, in: ICC 2023-IEEE International Conference on Communications, IEEE, 2023, May, pp. 1798–1803. [15] R.S. Antunes, C. Andr´ e da Costa, A. Küderle, I.A. Yari, B. Eskofier, Federated learning for healthcare: systematic review and architecture proposal, ACM Transactions on Intelligent Systems and Technology 13 (4) (2022), https://doi. org/10.1145/3501813, 54:1-54:23. [16] M. Abadi, A. Chu, I. Goodfellow, H.B. McMahan, I. Mironov, K. Talwar, L. Zhang, Deep learning with differential privacy, in: Proceedings of the 2016 ACM SIGSAC Conference on Computer and Communications Security, 2016, October, pp. 308–318. [17] P. Qi, D. Chiaro, A. Guzzo, M. Ianni, G. Fortino, F. Piccialli, Model aggregation techniques in federated learning: a comprehensive survey, Future Gener. Comput. Syst. 150 (2024) 272–293, https://doi.org/10.1016/j.future.2023.09.008. [18] C. Huang, J. Huang, X. Liu, Cross-Silo Federated Learning: Challenges and Opportunities (arXiv:2206.12949), 2022 arXiv, http://arxiv.org/abs/2206.12 949. [19] P. Kairouz, H.B. McMahan, B. Avent, A. Bellet, M. Bennis, A.N. Bhagoji, K. Bonawitz, Z. Charles, G. Cormode, R. Cummings, R.G.L. D’Oliveira, H. Eichner, S.E. Rouayheb, D. Evans, J. Gardner, Z. Garrett, A. Gasc´ on, B. Ghazi, P. B. Gibbons, S. Zhao, Advances and open problems in federated learning, Foundations and Trends® in Machine Learning 14 (1–2) (2021) 1–210, https:// doi.org/10.1561/2200000083. [20] Y. Liu, Y. Kang, T. Zou, Y. Pu, Y. He, X. Ye, Y. Ouyang, Y. Zhang, Q. Yang, Vertical federated learning: concepts, advances, and challenges, IEEE Trans. Knowl. Data Eng. 36 (7) (2024) 3615–3634, https://doi.org/10.1109/tkde.2024.3352628. [21] Q. Yang, Y. Liu, T. Chen, Y. Tong, Federated Machine Learning: Concept and Applications (arXiv:1902.04885). arXiv, 2019. http://arxiv.org/abs/1902.04885. [22] L. Yuan, L. Sun, P.S. Yu, Z. Wang, Decentralized Federated Learning: a Survey and Perspective (arXiv:2306.01603). arXiv, 2023, https://doi.org/10.48550/ arXiv.2306.01603. [23] C. Zhang, Y. Xie, H. Bai, B. Yu, W. Li, Y. Gao, A survey on federated learning, Knowl. Base Syst. 216 (2021) 106775, https://doi.org/10.1016/j. knosys.2021.106775. [24] B.S. Guendouzi, S. Ouchani, H. El Assaad, M. El Zaher, A systematic review of federated learning: challenges, aggregation methods, and development tools, J. Netw. Comput. Appl. 220 (2023) 103714, https://doi.org/10.1016/j. jnca.2023.103714. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 18 [25] C. Hu, J. Jiang, Z. Wang, ). Decentralized Federated Learning: a Segmented Gossip Approach (arXiv:1908.07782, 2019, https://doi.org/10.48550/ arXiv.1908.07782 arXiv. [26] E.T.M. Beltr´ an, M.Q. P´ erez, P.M.S. S´ anchez, S.L. Bernal, G. Bovet, M.G. P´ erez, G. M. P´ erez, A.H. Celdr´ an, Decentralized federated learning: fundamentals, state of the art, frameworks, trends, and challenges, IEEE Communications Surveys & Tutorials 25 (4) (2023) 2983–3013, https://doi.org/10.1109/ comst.2023.3315746. [27] A. Khan, M.T. Thij, A. Wilbik. Vertical Federated Learning: a Structured Literature Review. Knowledge and Information Systems, 2025, https://doi.org/10.1007/ s10115-025-02356-y. [28] Y. Liu, Y. Kang, C. Xing, T. Chen, Q. Yang, A secure federated transfer learning framework, IEEE Intell. Syst. 35 (4) (2020) 70–82, https://doi.org/10.1109/ MIS.2020.2988525. [29] S. Saha, T. Ahmad, Federated transfer learning: concept and applications, Intell. Artif. 15 (1) (2021) 35–44, https://doi.org/10.3233/IA-200075. [30] J. Yu, Y. Chen, Z. Wang, J. Liu, B. Huang, Food risk entropy model based on federated learning, APPLIED SCIENCES-BASEL 12 (10) (2022), https://doi.org/ 10.3390/app12105174. [31] Z.M. Yaseen, The next generation of soil and water bodies heavy metals prediction and detection: new expert system based edge cloud server and federated learning technology, Environ. Pollut. 313 (2022), https://doi.org/ 10.1016/j.envpol.2022.120081. [32] H. Chen, J. Xie, L. Xu, Q. Feng, Q. Lin, K. Cai, Feature selection for portable spectral sensing data of soil using broad learning network in fusion With fuzzy technique, IEEE Sens. J. 24 (5) (2024) 5644–5653, https://doi.org/10.1109/ JSEN.2023.3331026. [33] S. Park, S. Jung, H. Lee, J. Kim, J.-H. Kim, Large-scale water quality prediction using federated sensing and learning: a case study with real-world sensing bigdata, SENSORS 21 (4) (2021), https://doi.org/10.3390/s21041462. [34] J. Vellingiri, K. Kalaivanan, P. Gopinath M, C. Gobinath, S. Prabhakar Rontala, R. Sarathkumar, Strategies for classifying water quality in the cauvery river using a federated learning technique, International Journal of Cognitive Computing in Engineering 4 (2023) 187–193, https://doi.org/10.1016/j.ijcce.2023.04.004. [35] J. Zhu, Research on the water quality intelligent monitoring system of river gushing based on federal learning, J. Phys. Conf. 2010 (1) (2021), https://doi. org/10.1088/1742-6596/2010/1/012150. [36] M.J.T. Kathen, F. Peralta, P. Johnson, I.J. Flores, D.G. Reina, AquaFeL-PSO: an informative path planning for water resources monitoring using autonomous surface vehicles based on multi-modal PSO and federated learning, Ocean. Eng. 311 (2024), https://doi.org/10.1016/j.oceaneng.2024.118787. [37] D. Vimalajeewa, C. Kulatunga, D.P. Berry, S. Balasubramaniam, A service-based joint model used for distributed learning: application for smart agriculture, IEEE Transactions on Emerging Topics in Computing 10 (2) (2022) 838–854, https:// doi.org/10.1109/TETC.2020.3048671. [38] M. Gulati, N. Dadkhah, B. Groß, G. Wunder, J. Glavonjic, A. Pavlovic, A. Tomcic, BETA-FL: blockchain-event triggered asynchronous federated learning in supply chains, EasyChair (2023). [39] J. Müller-Maatsch, F.R. Bertani, A. Mencattini, A. Gerardino, E. Martinelli, Y. Weesepoel, S. van Ruth, The spectral treasure house of miniaturized instruments for food safety, quality and authenticity applications: a perspective, Trends Food Sci. Technol. 110 (2021) 841–848. [40] Y. Huang, X. Yang, J. Guo, J. Cheng, H. Qu, J. Ma, L. Li, A high-precision method for 100-Day-Old classification of chickens in edge computing scenarios based on federated computing, Animals 12 (24) (2022), https://doi.org/10.3390/ ani12243450. [41] A. Mao, E. Huang, H. Gan, K. Liu, FedAAR: a novel federated learning framework for animal activity recognition with wearable sensors, Animals 12 (16) (2022). https://search.ebscohost.com/login.aspx?direct=true&db=lah&AN=202204138 92&site=ehost-live. [42] N. Gengler, Challenges and opportunities for evaluating and using the genetic potential of dairy cattle in the new era of sensor data from automation, J. Dairy Sci. 102 (6) (2019) 5756–5763. https://search.ebscohost.com/login.aspx?direct =true&db=lah&AN=20193330242&site=ehost-live. [43] J. Arshad, A. Irtisam, T. Arif, M.S. Rasheed, S.T. Chauhdary, M.K.I. Rahmani, R. Almajalid, A federated learning model for intelligent cattle health monitoring system using body area sensors and IoT, Egyptian Informatics Journal 27 (2024), https://doi.org/10.1016/j.eij.2024.100488. [44] L. Borgianni, M.S. Ahmed, D. Adami, S. Giordano, Spectrogram based bee sound analysis with DNNs: a step toward federated learning approach, in: 2023 4th International Symposium on the Internet of Sounds, IEEE, 2023, October, pp. 1–8, https://doi.org/10.1109/IEEECONF59510.2023.10335200. [45] A.L. Bowler, M.P. Pound, N.J. Watson, Domain adaptation and federated learning for ultrasonic monitoring of beer fermentation, Fermentation 7 (4) (2021). https ://search.ebscohost.com/login.aspx?direct=true&db=lah&AN=20220197921&s ite=ehost-live. [46] F. Cui, Z.A. Al-Sudani, G.S. Hassan, H.A. Afan, S.J. Ahammed, Z. Yaseen, Boosted artificial intelligence model using improved alpha-guided grey wolf optimizer for groundwater level prediction: comparative study and insight for federated learning technology, J. Hydrol. 606 (2022), https://doi.org/10.1016/j. jhydrol.2021.127384. [47] J. Elhachmi, A. Kobbane, A federated learning approach for water distribution networks monitoring, in: Proceedings—2022 9th International Conference on Wireless Networks and Mobile Communications, WINCOM 2022, 2022, https:// doi.org/10.1109/WINCOM55661.2022.9966455. [48] J.C. Mato, S.G. V´ azquez, J.F. Aguila, A.D. Rodríguez, X. Lin, L.G. G´ andara, J. S. Seoane, J.S. Castro, Foam segmentation in wastewater treatment plants, WATER 16 (3) (2024), https://doi.org/10.3390/w16030390. [49] M. Akbari, A. Syed, W.S. Kennedy, M. Erol-Kantarci, AoI-Aware energy-efficient SFC in UAV-aided smart agriculture using asynchronous federated learning, IEEE Open Journal of the Communications Society 5 (2024) 1222–1242, https://doi. org/10.1109/OJCOMS.2024.3363132. [50] J. Pei, W. Liu, L. Wang, C. Liu, A.K. Bashir, Y. Wang, Fed-IoUT: opportunities and challenges of federated learning in the internet of underwater things, IEEE Internet of Things Magazine 6 (1) (2023) 108–112, https://doi.org/10.1109/ IOTM.001.2200127. [51] S. Bera, T. Dey, A. Mukherjee, D.B.S. De, FLAG: federated learning for sustainable irrigation in agriculture 5.0, IEEE Trans. Consum. Electron. 70 (1) (2024) 2303–2310, https://doi.org/10.1109/TCE.2024.3370373. [52] W.K. Cheng, B. Yaik Ooi, T.B. Tan, Y. -L. Chen. Edge-Cloud Architecture for Precision Aquaculture, 2023 International Conference on Consumer Electronics - Taiwan (ICCE-Taiwan), PingTung, Taiwan, 2023, pp. 355–356, https://doi.org/ 10.1109/ICCE-Taiwan58799.2023.10226981. [53] A. Kaushal, O. Almurshed, O. Almoghamis, A. Alabbas, N. Auluck, B. Veeravalli, O. Rana, SHIELD: a secure heuristic integrated environment for load distribution in rural-AI, Future Gener. Comput. Syst. 161 (2024) 286–301, https://doi.org/ 10.1016/j.future.2024.07.026Khan. [54] M.P.L. Ooi, S. Sohail, V.G. Huang, N. Hudson, M. Baughman, O. Rana, H. Nagra, Measurement and applications: exploring the challenges and opportunities of hierarchical federated learning in sensor applications, IEEE Instrum. Meas. Mag. 26 (9) (2023) 21–31. [55] N. Singh, M. Adhikari, Edge-centric collaborative federated learning for irrigation management of paddy fields using agriculture sensor data processing, IEEE Sensors Letters 8 (8) (2024), https://doi.org/10.1109/LSENS.2024.3422417. [56] N. Singh, T. Tripathi, M. Adhikari, Communication-efficient federated learning for Real-time applications in edge networks, in: 2023 IEEE International Conference on Advanced Networks and Telecommunications Systems (ANTS), IEEE, 2023, pp. 527–532. [57] I. Siniosoglou, K. Xouveroudis, V. Argyriou, T. Lagkas, D. Margounakis, A.A. A. Boulogeorgos, P. Sarigiannidis, Applying federated learning on decentralized smart farming: a case study, in: 2023 IEEE International Conference on Communications Workshops (ICC Workshops), IEEE, 2023, pp. 1295–1300, https://doi.org/10.1109/ICCWorkshops57953.2023.10283681. [58] K. Thonglek, A. Maipradit, Hierarchical federated learning for predicting water levels: a case study in Thailand, in: 2024 16th International Conference on Computer and Automation Engineering (ICCAE), IEEE, 2024, pp. 218–222, https://doi.org/10.1109/ICCAE59995.2024.10569296. [59] A. Gavai, Y. Bouzembrak, W. Mu, F. Martin, R. Kaliyaperumal, J. van Soest, A. Choudhury, J. Heringa, A. Dekker, H. Marvin, A Federated Learning Approach to Data Sharing in a Food Supply Chain, 2022. [60] X. Zhang, X. Tan, J. Xu, S. Luo, Z. Qi, A trusted sharing model for risk information of food full-process and all-information based on blockchain and federated learning, in: 2023 IEEE International Conference on Blockchain (Blockchain), IEEE, 2023, pp. 186–191, https://doi.org/10.1109/ Blockchain60715.2023.00039, 2023. [61] A.N. Jahromi, H.K. Schulich, A. Dehghantanha, Deep federated learning-based cyber-attack detection in industrial control systems, in: 2021 18th International Conference on Privacy, Security and Trust, PST 2021, 2021, https://doi.org/ 10.1109/PST52912.2021.9647838. [62] P. Kumar, G.P. Gupta, R. Tripathi, PEFL: deep privacy-encoding-based federated learning framework for smart agriculture, IEEE MICRO 42 (1) (2022) 33–40, https://doi.org/10.1109/MM.2021.3112476. [63] O. Friha, M.A. Ferrag, L. Shu, L. Maglaras, K.-K.R. Choo, M. Nafaa, FELIDS: federated learning-based intrusion detection system for agricultural internet of things, J. Parallel Distr. Comput. 165 (2022) 17–31, https://doi.org/10.1016/j. jpdc.2022.03.003. [64] K. Bala, P.D. Kaur, A novel game theory based reliable proof-of-stake consensus mechanism for blockchain, Transactions on Emerging Telecommunications Technologies 33 (9) (2022), https://doi.org/10.1002/ett.4525. [65] I. Alrashdi, K.M. Sallam, A. Alqazzaz, B. Arain, I.A. Hameed, A contrastive learning approach for enhanced robustness for strengthening federated intelligence in internet of visual things, Internet of Things (Netherlands) 26 (2024), https://doi.org/10.1016/j.iot.2024.101206. [66] S. Bahadoripour, H. Karimipour, A.N. Jahromi, A. Islam, An explainable multimodal model for advanced cyber-attack detection in industrial control systems, Internet of Things (Netherlands) 25 (2024), https://doi.org/10.1016/j. iot.2024.101092. [67] C.A. Lee, K.M. Chow, H.A. Chan, D.P.K. Lun, Decentralized governance and artificial intelligence policy with blockchain-based voting in federated learning, Frontiers in Research Metrics and Analytics 8 (2023) 1035123. [68] G.Q. Zeng, J.M. Shao, K.D. Lu, G.G. Geng, J. Weng, Automated federated learning-based adversarial attack and defence in industrial control systems, IET Cyber-Systems and Robotics 6 (2) (2024) e12117, https://doi.org/10.1049/ csy2.12117. [69] L. Praharaj, M. Gupta, D. Gupta, Hierarchical federated transfer learning and digital twin enhanced secure cooperative smart farming, in: 2023 IEEE International Conference on Big Data (Bigdata), IEEE, 2023, pp. 3304–3313. [70] Z. Liu, L. Song, C. Fragouli, Federated multi-armed bandits with vector rewards for aspect-based recommendations, in: 2022 IEEE Global Communications Conference, GLOBECOM 2022—Proceedings, 2022, pp. 1079–1084, https://doi. org/10.1109/GLOBECOM48099.2022.10001533. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 19 [71] K. Sudharson, S. Lakshiya, M. Lokesh, Fettle: unveiling a personalized nutritional odyssey through cutting-edge AI and data integration, in: 2024 5th International Conference for Emerging Technology (INCET), IEEE, 2024, pp. 1–6, https://doi. org/10.1109/INCET61516.2024.10592948. [72] R. Salem, B. Aidaros, F. Belqasmi, Sustainable Industrial Agriculture in the UAE: Leveraging Iot and AI Technologies, 2024, pp. 1–6, https://doi.org/10.1109/ URC62276.2024.10604647. [73] S. Seilov, B. Pandey, A. Nurzhaubayev, D. Abildinov, A. Konyrkhanova, B. Zhursinbek, Federated Learning Based Smart Horticulture and Smart Storage of Fruits Using E-Nose, and Blockchain: A Proposed Model, in: 2024 IEEE 3rd International Conference onAI in Cybersecurity (ICAIC), Houston, TX, USA, 2024, pp. 1–5, https://doi.org/10.1109/ICAIC60265.2024.10433834. [74] M. Aggarwal, V. Khullar, N. Goyal, Contemporary and futuristic intelligent technologies for rice leaf disease detection, in: 2022 10th International Conference on Reliability, Infocom Technologies and Optimization (Trends and Future Directions), ICRITO 2022, 2022, https://doi.org/10.1109/ ICRITO56286.2022.9965113. [75] A. Bansal, S. Vats, S. Mittal, V. Kukreja, S. Mehta, Strawberries at the nexus of tech: Federated learning and CNN for disease severity classification. https://doi. org/10.1109/SMART59791.2023.10428253, 2023. [76] N. Thapliyal, V. Kukreja, N. Garg, S. Mehta, J. Anand, Empowering Farmers: Federated Learning CNN for Accurate Strawberry Leaf Disease, 2024, pp. 1–6, https://doi.org/10.1109/ICRITO61523.2024.10522397. [77] S. Vats, M.S. Khanna, V. Kukreja, S. Mehta, Strawberry Leaf Disease Severity Decoded for Agriculture: A Federated Learning CNN Approach 2 (2024) 1–6, https://doi.org/10.1109/IATMSI60426.2024.10503249. [78] A. Suryavanshi, V. Kukreja, S. Mehta, S. Malhotra, M. Manwal, Agriculture Advances: Federated Learning Cnn’s for Combatting Banana Leaf Diseases, 2024, pp. 1–6, https://doi.org/10.1109/INCET61516.2024.10592894. [79] S. Chattopadhyay, M. Manwal, V. Kukreja, S. Mehta, Transforming AgroDiagnostics: Banana Leaf Diseases Through Federated Learning Cnns, 2024, https://doi.org/10.1109/INOCON60754.2024.10512313, 1–6. [80] S. Malhotra, M. Manwal, V. Kukreja, S. Mehta, Technological Synergy in Agriculture: a Federated Learning Cnns Against Banana Leaf Diseases, 2024, pp. 1–6, https://doi.org/10.1109/ICRITO61523.2024.10522379. [81] V. Sharma, V. Kukreja, S. Mehta, M. Aeri, Bananaleafnet: Federated Learning Cnns for Multiclass Disease Identification in Banana Crops, 2023, pp. 1–6, https://doi.org/10.1109/ICSTCEE60504.2023.10584854. [82] D. Malik, V. Kukreja, S. Mehta, A. Gupta, V. Singh, Mitigating the Impact of Guava Leaf Diseases Using Cnns and Federated Learning, 2023, pp. 1–6, https:// doi.org/10.1109/ASIANCON58793.2023.10270236. [83] K. Rajput, A. Garg, V. Kukreja, S. Mehta, Agricultural Intelligence: Federated Learning Cnn’s Models for Jute Leaf Disease Analysis, 1–6, 2024, https://doi.org/ 10.1109/ICRITO61523.2024.10522218. [84] J. Joshi, M. Aeri, V. Kukreja, S. Mehta, Revolutionizing in Agriculture: Federated CNN Models for Sunflower Leaf Diseases, 2024, pp. 1–6, https://doi.org/ 10.1109/ICRITO61523.2024.10522441. [85] V. Kukreja Vishalli, S. Mehta, P. Das, V. Singh, Dissecting Parsley Leaf Disease Severity: a Federated Learning and CNN Perspective, 2023, pp. 1–6, https://doi. org/10.1109/ASIANCON58793.2023.10270195. [86] V. Jindal, V. Kukreja, A. Bhattacherjee, U. Garg, S. Mehta, Sustainable Coffee Production: a Federated Learning Framework with CNN for Disease Detection and Classification, 2023, https://doi.org/10.1109/ICCCIS60361.2023.10425772, 776–781. [87] S. Mehta, V. Kukreja, A. Gupta, Revolutionizing maize disease management with federated learning CNNs: a decentralized and privacy-sensitive approach, in: 2023 4th International Conference for Emerging Technology (INCET), 2023, pp. 1–6, https://doi.org/10.1109/INCET57972.2023.10170499. [88] S. Mehta, V. Kukreja, S. Vats, Empowering farmers with AI: federated learning of CNNs for wheat diseases multi-classification, in: 2023 4th International Conference for Emerging Technology (INCET), 2023, pp. 1–6. [89] C. Yu, S. Shen, K. Zhang, H. Zhao, Y. Shi, Energy-aware device scheduling for joint federated learning in edge-assisted internet of agriculture things, in: IEEE Wireless Communications and Networking Conference, WCNC (Vols. 2022-April, 2022, pp. 1140–1145, https://doi.org/10.1109/WCNC51071.2022.9771547. [90] F. Deng, W. Mao, Z. Zeng, H. Zeng, B. Wei, Multiple diseases and pests detection based on federated learning and improved faster R-CNN, IEEE Trans. Instrum. Meas. 71 (2022), https://doi.org/10.1109/TIM.2022.3201937. [91] J. Tong, Y. Zou, Z. Zeng, F. Deng, Detection technology of several pests and diseases in orchard based on federated learning and improved faster R-CNN. CAIBDA 2022; 2nd International Conference on Artificial Intelligence, Big Data and Algorithms, 2022, pp. 1–4. [92] E. Konyar, M.R. Gahrooei, Federated generalized scalar-on-tensor regression, J. Qual. Technol. 56 (1) (2024) 20–37, https://doi.org/10.1080/ 00224065.2023.2246600. [93] F.S. Khan, S. Khan, Mohd N.H. Mohd, A. Waseem, M.N.A. Khan, S. Ali, R. Ahmed, Federated learning-based UAVs for the diagnosis of plant diseases, in: 8th International Conference on Engineering and Emerging Technologies, ICEET 2022, 2022, https://doi.org/10.1109/ICEET56468.2022.10007133. [94] M.A. Al-Betar, A.K. Abasi, Z.A.A. Alyasseri, S. Fraihat, R.F. Mohammed, A communication-efficient federated learning framework for sustainable development using lemurs optimizer, Algorithms 17 (4) (2024), https://doi.org/ 10.3390/a17040160. [95] C. Carrascosa, F. Enguix, M. Rebollo, J.A. Rincon, Consensus-based learning for MAS: definition, implementation and integration in IVEs, International Journal of Interactive Multimedia and Artificial Intelligence 8 (3) (2023) 21–32, https://doi. org/10.9781/ijimai.2023.08.004. [96] S. Choubey, D. Divya, Lightweight Federated Transfer Learning for Plant Leaf Disease Detection and Classification Across Multiclient cross-silo Datasets, 2024, p. 82, https://doi.org/10.1051/bioconf/20248205018. [97] H. Devaraj, S. Sohail, M. Ooi, B.Y. Li, N. Hudson, M. Baughman, K. Chard, R. Chard, E. Casella, I. Foster, O. Rana, RuralAI in tomato farming: integrated sensor system, distributed computing, and hierarchical federated learning for crop health monitoring, IEEE Sensors Letters 8 (5) (2024), https://doi.org/ 10.1109/LSENS.2024.3384935. [98] D. Mamba Kabala, A. Hafiane, L. Bobelin, R. Canals, Image-based crop disease detection with federated learning, Sci. Rep. 13 (1) (2023) 19220. [99] B.L. Muhammad, M.B. Kusharki, Federated Learning for Collaborative Crop Disease Monitoring in Wheat Production 1 (2023) 1–5, https://doi.org/10.1109/ ICMEAS58693.2023.10379272. [100] R. Tripathy, J. Mandala, S.R. Pappu, G.K.D. Gopisetty, Optimization based rice leaf disease classification in federated learning, Multimed. Tool. Appl. (2024), https://doi.org/10.1007/s11042-023-18085-9. [101] M. Fahim-Ul-Islam, A. Chakrabarty, S.T. Ahmed, R. Rahman, H.H. Kwon, M. J. Piran, A comprehensive approach towards wheat leaf disease identification leveraging transformer models and federated learning, IEEE Access (2024), https://doi.org/10.1109/ACCESS.2024.3438544, 1–1. [102] A. Durrant, M. Markovic, D. Matthews, D. May, J. Enright, G. Leontidis, The role of cross-silo federated learning in facilitating data sharing in the agri-food sector, Comput. Electron. Agric. 193 (2022), https://doi.org/10.1016/j. compag.2021.106648. [103] T. Manoj, K. Makkithaya, V.G. Narendra, A federated learning-based crop yield prediction for agricultural production risk management, in: 2022 IEEE Delhi Section Conference, DELCON 2022, 2022, https://doi.org/10.1109/ DELCON54057.2022.9752836. [104] Q. Zhang, X. Zhao, Y. Han, F. Yang, S. Pan, Z. Liu, K. Wang, C. Zhao, Maize yield prediction using federated random forest, in: Computers and Electronics in Agriculture, 210, 2023, https://doi.org/10.1016/j.compag.2023.107930. [105] S. Bera, T. Dey, A. Mukherjee, R. Buyya, E-CropReco: a dew-edge-based multiparametric crop recommendation framework for internet of agricultural things, J. Supercomput. 79 (11) (2023) 11965–11999, https://doi.org/10.1007/s11227023-05131-4. [106] G. Idoje, T. Dagiuklas, M. Iqbal, Federated learning: crop classification in a smart farm decentralised network, in: Smart Agricultural Technology, 5, 2023, https:// doi.org/10.1016/j.atech.2023.100277. [107] A. Li, M. Markovic, P. Edwards, G. Leontidis, Model pruning enables localized and efficient federated learning for yield forecasting and data sharing, Expert Syst. Appl. 242 (2024) 122847. [108] J. Nie, J. Jiang, Y. Li, J. Li, Y. Qiao, S. Ercisli, UAVEC-FLchain: distributed multiregional jujube orchard joint yield estimation for secure agricultural-IoT applications, Internet of Things 25 (2024) 101143. [109] Y.-T. Hsieh, C.-Y. Lee, C.-C. Lin, P. Liu, J.-J. Wu, A bicameralism voting framework for combining knowledge from clients into better prediction, in: Proceedings—2019 IEEE International Conference on Big Data, Big Data 2019, 2019, pp. 298–306, https://doi.org/10.1109/BigData47090.2019.9005528. [110] D. Polap, M. Wozniak, A hybridization of distributed policy and heuristic augmentation for improving federated learning approach, Neural Netw. 146 (2022) 130–140, https://doi.org/10.1016/j.neunet.2021.11.018. [111] V. Huang, S. Sohail, M. Mayo, T.L. Botran, M. Rodrigues, C. Anderson, M. Ooi, Keep it Simple: Fault Tolerance Evaluation of Federated Learning with Unreliable Clients, 2023, pp. 1–3, https://doi.org/10.1109/CLOUD60044.2023.00024. [112] S. Ali, M.I. Ullah, A. Sajjad, Q. Shakeel, A. Hussain, Environmental and health effects of pesticide residues, in: Inamuddin, M.I. Ahamed, E. Lichtfouse (Eds.), Sustainable Agriculture Reviews 48: Pesticide Occurrence, Analysis and Remediation Vol. 2 Analysis, Springer International Publishing, 2021, pp. 311–336, https://doi.org/10.1007/978-3-030-54719-6_8. [113] K. Rehman, F. Fatima, I. Waheed, M.S.H. Akash, Prevalence of exposure of heavy metals and their impact on health consequences, J. Cell. Biochem. 119 (1) (2017) 157–184, https://doi.org/10.1002/jcb.26234. [114] M. Steele, J. Odumeru, Irrigation water as source of foodborne pathogens on fruit and vegetables, J. Food Protect. 67 (12) (2004) 2839–2849, https://doi.org/ 10.4315/0362-028X-67.12.2839. [115] J.O. Ighalo, A.G. Adeniyi, G. Marques, Internet of things for water quality monitoring and assessment: a comprehensive review, in: A.E. Hassanien, R. Bhatnagar, A. Darwish (Eds.), Artificial Intelligence for Sustainable Development: Theory, Practice and Future Applications, Springer International Publishing, 2021, pp. 245–259, https://doi.org/10.1007/978-3-030-51920-9_13. [116] R. Müller, L. Schrader, A new method to measure behavioural activity levels in dairy cows, Appl. Anim. Behav. Sci. 83 (4) (2003) 247–258, https://doi.org/ 10.1016/S0168-1591(03)00141-2. [117] A.L. Bowler, N.J. Watson, Transfer learning for process monitoring using reflection-mode ultrasonic sensing, Ultrasonics 115 (2021) 106468, https://doi. org/10.1016/j.ultras.2021.106468. [118] M. Navidimehr, A.M. Khalilabad, H. Amintoosi, M. Allahbakhsh, An LPWANbased framework for food supply chain management using blockchain and federated multi-task learning, in: 2024 8th International Conference on Smart Cities, Internet of Things and Applications (Sciot), IEEE, 2024, pp. 207–212. [119] T.H. Liu, B. Ye, Z. Qin, J. Zhang, The high-performance solution with federated learning in supply chain system, ACM International Conference Proceeding Series (2022) 241–245, https://doi.org/10.1145/3532213.3532249. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 20 [120] E. Nakashima, Foreign hackers targeted U.S. water plant in apparent malicious cyber attack, expert says. https://www.washingtonpost.com/blogs/checkpo int-washington/post/foreign-hackers-broke-into-illinois-water-plant-controlsystem-industry-expert-says/2011/11/18/gIQAgmTZYN_blog.html, 2023. [121] Y. Lu, L.D. Xu, Internet of things (IoT) cybersecurity research: a review of current research topics, IEEE Internet Things J. 6 (2) (2019) 2103–2115, https://doi.org/ 10.1109/JIOT.2018.2869847. [122] R. Abbasi, P. Martinez, R. Ahmad, The digitization of agricultural Industry–A systematic literature review on agriculture 4.0, Smart Agricultural Technology 2 (2022) 100042. [123] V. Hassija, V. Chamola, V. Gupta, S. Jain, N. Guizani, A survey on supply chain security: application areas, security threats, and solution architectures, IEEE Internet Things J. 8 (8) (2020) 6222–6246. [124] T. Li, A.K. Sahu, A. Talwalkar, V. Smith, Federated learning: challenges, methods, and future directions, IEEE Signal Process. Mag. 37 (3) (2020) 50–60, https://doi. org/10.1109/MSP.2020.2975749. [125] Y. Zhao, M. Li, L. Lai, N. Suda, D. Civin, V. Chandra, Federated learning with NonIID data. https://doi.org/10.48550/arXiv.1806.00582, 2018. [126] K. Thonglek, P. Rattanatamrong, Decentralized federated learning for agricultural plant diseases identification on edge devices, in: 2023 18th International Joint Symposium on Artificial Intelligence and Natural Language Processing (iSAINLP), IEEE, 2023, pp. 1–6, https://doi.org/10.1109/iSAINLP60301.2023.10354694. [127] J. Qian, B. Dai, B. Wang, Y. Zha, Q. Song, Traceability in food processing: problems, methods, and performance Evaluations—A review, Crit. Rev. Food Sci. Nutr. 62 (3) (2020) 679–692, https://doi.org/10.1080/10408398.2020.1825925. [128] A. Khan, M. ten Thij, A. Wilbik, Vertical federated learning: a structured literature review, Knowl Inf Syst 67 (2025) 3205–3243. https://doi-org.ezproxy.library. wur.nl/10.1007/s10115-025-02356-y. [129] A. Hosna, E. Merry, J. Gyalmo, Z. Alom, Z. Aung, M.A. Azim, Transfer learning: a friendly introduction, J. Big Data 9 (1) (2022), https://doi.org/10.1186/s40537022-00652-w. [130] E.T. Martínez Beltr´ an, M.Q. P´ erez, P.M.S. S´ anchez, S.L. Bernal, G. Bovet, M. G. P´ erez, G.M. P´ erez, A.H. Celdr´ an, Decentralized federated learning: fundamentals, state of the art, frameworks, trends, and challenges, IEEE Communications Surveys & Tutorials 25 (4) (2023) 2983–3013, https://doi.org/ 10.1109/COMST.2023.3315746. [131] Y. Bouzembrak, M. Klüche, A. Gavai, H.J.P. Marvin, Internet of things in food safety: literature review and a bibliometric analysis, Trends Food Sci. Technol. 94 (2019) 54–64, https://doi.org/10.1016/j.tifs.2019.11.002. [132] C. Little, M. Elliot, R. Allmendinger, Federated learning for generating synthetic data: a scoping review, International Journal of Population Data Science 8 (1) (2023) 2158. [133] M. Rigaki, S. Garcia, A survey of privacy attacks in machine learning, ACM Comput. Surv. 56 (4) (2023) 1–34. [134] L. Lyu, H. Yu, X. Ma, C. Chen, L. Sun, J. Zhao, Q. Yang, P.S. Yu, Privacy and robustness in federated learning: attacks and defenses, IEEE Transact. Neural Networks Learn. Syst. (2022) 1–21, https://doi.org/10.1109/ TNNLS.2022.3216981. [135] C. Gentry, S. Halevi, Implementing gentry’s fully-homomorphic encryption scheme, in: Annual International Conference on the Theory and Applications of Cryptographic Techniques, 2011, May, pp. 129–148. [136] A. Malatras, I. Agrafiotis, M. Adamczyk, Securing Machine Learning Algorithms [Report/Study]. ENISA, 2021. https://www.enisa.europa.eu/publications/ securing-machine-learning-algorithms. [137] N. Truong, K. Sun, S. Wang, F. Guitton, Y. Guo, Privacy preservation in federated learning: an insightful survey from the GDPR perspective, Comput. Secur. 110 (2021) 102402, https://doi.org/10.1016/j.cose.2021.102402. [138] Directive - 2016/943 - En - Eur-Lex. (n.d.). https://eur-lex.europa.eu/eli/dir/ 2016/943/oj/eng. [139] J. Liu, Z. Chang, G. Min, Z. Han, Incentive mechanism design for federated learning in multi-access edge computing, in: 2022 IEEE Global Communications Conference, GLOBECOM 2022—Proceedings, 2022, pp. 3454–3459, https://doi. org/10.1109/GLOBECOM48099.2022.10000933. Z. Fendor et al. Journal of Agriculture and Food Research 23 (2025) 102238 21