Data handling of homomorphic encrypted data used for machine learning model training in o-ran

Abstract

Embodiments of the disclosure describe a method for data handling of homomorphic encrypted data used for machine learning model training in an Open Radio Access Network (O-RAN). The method includes generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling. The method further includes transmitting, by the data source entity, the generated data to a model training entity. The generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model. The trained ML model is transferred from the model training entity to a model inference entity. The method further includes transmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

Description

DATA HANDLING OF HOMOMORPHIC ENCRYPTED DATA USED FOR MACHINE LEARNING MODEL TRAINING IN O-RANCROSS-REFERENCE TO RELATED APPLICATIONS

[0001] This application claims priority to Indian Provisional Patent Application No.202511007468, filed on January 29, 2025, and to Indian Non-Provisional Patent Application No.202511007468, filed on August 26, 2025 the entire contents of which are incorporated herein by reference.FIELD

[0002] The present disclosure relates to data handling of homomorphic encrypted data used for machine learning model training in an Open Radio Access Network (O-RAN).BACKGROUND

[0003] The information disclosed in this background section is only for enhancement of understanding of the general background of the disclosure and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art already known to a person skilled in the art.

[0004] Homomorphic Encryption (HE) is a cryptographic method that allows calculations to be performed on encrypted data without the need for decryption. The HE is especially useful for Machine Learning (ML) tasks, which is important to keep training data confidential and private while still enabling effective model training. The HE supports secure data sharing and processingin fields like telecommunication, healthcare, and finance. For instance, organizations can train ML models using encrypted datasets, ensuring that the original data remains protected from any unauthorized access. This improves data privacy and meets strict data protection regulations.SUMMARY

[0005] This summary is provided to introduce a selection of concepts, in a simplified format, that are further described in the detailed description of the disclosure. This summary is neither intended to identify key or essential inventive concepts of the disclosure nor is it intended for determining the scope of the disclosure.

[0006] According to one embodiment of the present disclosure, a method is disclosed. The method includes generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling. The method further includes transmitting, by the data source entity, the generated data to a model training entity. The generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model. The trained ML model is transferred from the model training entity to a model inference entity. The method further includes transmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, and to obtain a plaintext output at the model inference entity.

[0007] According to one embodiment of the present disclosure, an apparatus is disclosed. The apparatus is configured to generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling. The apparatus is configured to transmit the generated data to a model training entity. The generated data with the HE and plain text labeling is used at the model trainingentity to train a Machine Learning (ML) model. The trained ML model is transferred from the model training entity to a model inference entity. The apparatus is configured to transmit the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, and to obtain a plaintext output at the model inference entity.

[0008] According to one embodiment of the present disclosure, a non-transitory computer-readable medium stores instructions. The one or more instructions are executed by an apparatus that comprises one or more processors. The one or more processors are configured to generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling. The one or more processors are configured to transmit the generated data to a model training entity. The generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model. The trained ML model is transferred from the model training entity to a model inference entity. The one or more processors are configured to transmit the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, and to obtain a plaintext output at the model inference entity.

[0009] To further clarify the advantages and features of the present disclosure, a more particular description of the disclosure will be rendered by reference to specific embodiments thereof, which are illustrated in the appended drawings. It is appreciated that these drawings depict only typical embodiments of the disclosure and are therefore not to be considered limiting of its scope. The disclosure will be described and explained with additional specificity and detail in the accompanying drawings.BRIEF DESCRIPTION OF THE DRAWINGS

[0010] Features, aspects, and advantages of embodiments of the disclosure will be described below with reference to the accompanying drawings, in which like reference numerals denote like elements, and wherein:FIG. 1 is a diagram of an example of an implementation environment in which systems and / or methods, described herein, may be implemented, according to an embodiment as disclosed herein;FIG. 2 is a sequence flow diagram illustrating various message exchanges between a data source and other network entities for managing homomorphic encrypted data, according to an embodiment as disclosed herein;FIG. 3 is an example scenario of various message exchanges between an Open Radio Access Network (0-RAN) Distributed Unit (DU) and a Near Real-Time RAN Intelligent Controller (RT-RIC) for managing the homomorphic encrypted data, according to an embodiment as disclosed hereinFIG. 4 is a flow diagram illustrating a method for managing the homomorphic encrypted data utilized in training machine learning models within an 0-RAN, according to an embodiment as disclosed herein;FIG. 5 is a flow diagram illustrating a method for managing the homomorphic encrypted data utilized in training machine learning models within an O-RAN, according to another embodiment as disclosed herein; andFIG. 6 illustrates a diagram of example components of an apparatus, according to an embodiment as disclosed herein.DETAILED DESCRIPTION

[0011] The following detailed description of example embodiments refers to the accompanying drawings. The present disclosure provides illustrations and descriptions, but is not intended to be exhaustive or to limit the implementations to the precise form disclosed. Modifications and variations are possible in light of the present disclosure or may be acquired from practice of the implementations. Further, one or more features or components of one embodiment may be incorporated into or combined with another embodiment (or one or more features of another embodiment). Additionally, the flowchart and description of operations provided below relate to at least one of the embodiments in the present disclosure. It should be noted that it is possible to make other embodiments that do not exactly match the flowchart and its description. It is understood that in other embodiments, one or more operations may be omitted, one or more operations may be added, or one or more operations may be performed simultaneously (at least in part).

[0012] It will be apparent that systems and / or methods, described herein, may be implemented in different forms of hardware, software, or a combination of hardware and software. The actual specialized control hardware or software code used to implement these systems and / or methods should not limit their implementations. Thus, the operation and behavior of the systems and / or methods are described herein without reference to specific software code. It is understood that software and hardware may be designed to implement the systems and / or methods based on the description herein.

[0013] Even though particular combinations of features are recited in the claims and / or disclosed in the specification, the particular combinations are not intended to limit the disclosure of implementations. In fact, many of these features may be combined in ways not specifically recited in the claims and / or disclosed in the specification. Even if a dependent claim directly depends on only one claim, the present disclosure may indicate that the dependent claim is dependent on other claims in the claim set.

[0014] No element, act, or instruction used herein should be construed as critical or essential unless explicitly described as such. Also, as used herein, the articles “a” and “an” (in other words, nouns not mentioned in the plural) are intended to include one or more items, and may be used interchangeably with “one or more.” Also, as used herein, the terms “has,” “have,” “having,” “include,” “including,” or the like are intended to be open-ended terms. Further, the phrase “based on” is intended to mean “based, at least in part, on” unless explicitly stated otherwise. Furthermore, expressions such as “at least one of [A] and [B],” “[A] and / or [B],” or “at least one of [A] or [B]” are to be understood as including only A, only B, or both A and B.

[0015] The foregoing disclosure provides illustration and description, but is not intended to be exhaustive or to limit the implementations to the precise form disclosed. Modifications and variations are possible in light of the above disclosure or may be acquired from practice of the implementations.

[0016] In the present disclosure, specific tasks may be performed using Artificial Intelligence / Machine Learning (AI / ML) models. An AI / ML model is a model generated using one or more Al technologies, one or more ML algorithms, or both, and generates output data based on input data. This output data is used to perform tasks. Tasks performed using AI / ML models includethose generally referred to as intellectual tasks, such as classification, prediction, natural language processing, etc.

[0017] Although Al and ML are explained separately, ML is a technology included in Al. In ML, instead of being explicitly programmed for a specific task, systems can improve their performance over time by identifying patterns and making inferences from training data. Typically, the generation of ML models includes data collection, model training, and model inference. Data collection involves gathering and preprocessing data to be used for training and inference. Model training involves developing and validating models using the collected data. Model inference involves applying the trained models to new data to generate new output data and perform tasks.

[0018] Machine learning includes various types of learning methods such as supervised learning, unsupervised learning, reinforcement learning, semi-supervised learning, self-supervised learning, transudative learning, transfer learning, meta learning, and the like. These types of learning methods can be appropriately selected according to the embodiments. Unless otherwise specified, the application of types not mentioned in this description is not precluded. Additionally, the structure of ML models may vary depending on the embodiments and learning methods, and is not limited to the methods disclosed. Furthermore, ML includes deep learning, which uses models that include neural networks. Deep learning models may include, for example, Deep Neural Networks (DNNs), Convolutional Neural Networks (CNNs), etc.

[0019] It should be noted that the AI / ML models presented hereinafter are examples and are not limited to the illustrated AI / ML models. They can be modified or altered by using different Al or ML algorithms. The configuration of the neural network is not limited to the configuration disclosed in the present disclosure and can be modified.

[0020] Throughout this disclosure, the abbreviations for the terminology used are listed in the table below.Table-1

[0021] Homomorphic Encryption (HE) is a cryptographic method that enables computations on encrypted data without the need for decryption. This cryptographic method offers several benefits, including, for example:a. Data privacy: Sensitive data can be processed and analyzed without exposing data content, even to third-party services or cloud providers.b. Collaboration: Organizations can share and collaborate on data without revealing sensitive information.c. Delegated computation: Organizations can outsource computations to untrusted servers while maintaining data confidentiality.

[0022] Despite the abovementioned advantages, several challenges / problems / limitations are encountered in the existing HE methods, which are mentioned below.

[0023] One problem identified herein is that when a Machine Learning (ML) model is trained using data encrypted with the HE, the output of the ML model remains encrypted. To access this inference output, a private key used for the HE at a data source must be available to a recipient. This poses a practical problem, as sharing the private key would undermine the security provided by encryption (e.g., HE).

[0024] For instance, in the telecommunications industry, protecting user data is crucial due to privacy regulations and the sensitive nature of communication information. Consider a scenario where a telecommunications company wants to collaborate with a network analytics firm to enhance service offerings by analyzing encrypted call data. The telecommunications company employs the HE to encrypt all user call data, ensuring that even if this user call data is shared with the analytics firm, this user call data remains confidential and unreadable to unauthorized parties. The analytics firm develops the ML model aimed at predicting network congestion based on call patterns. Using the encrypted call data provided by the telecommunications company, the analytics firm can train the ML model without ever decrypting the sensitive information, HE’s ability to perform computations on encrypted data. Once the ML model is trained, the ML model generates encrypted predictions regarding potential congestion in the network. However, to access thesepredictions, the analytics firm requires the private key used for the HE. Sharing the private key poses a significant challenge, as sharing the private key would compromise the encryption and the confidentiality of user data.

[0025] Additionally, existing Open Radio Access Network (O-RAN) alliance specifications outline the architecture for ML training and inference. However, they do not address how to handle data when the training or inference input is encrypted with the HE.

[0026] Therefore, there is a need to tackle the above-mentioned challenges / problems / limitations and explore alternative approaches for managing HE data in the context of machine learning within an O-RAN.

[0027] To address the above-mentioned challenges / problems / limitations, a disclosed method provides a unique strategy for data handling of homomorphic encrypted data used for machine learning model training in the O-RAN. The disclosed method may perform a plaintext labelling of an input data and train the ML model using encrypted inputs and plaintext labels. The labels in plaintext can be applied to regression and classification types of models. This enables the user of the ML model inference output to directly utilize without needing to perform decryption of the output, as described in conjunction with FIG. 1 to FIG. 6.

[0028] In some example embodiments, the disclosed method has several advantages over the existing methodologies, for example, which are mentioned herein. The private key need not be transferred to the entity performing model inference to decrypt the model inference output. Reduced computational resources are needed at the time of model inference. If model inference outputs homomorphic encrypted data, then the decryption requires a lot of resources at the entity performing model inference. And these resources might not always be available at that entity toperform the decryption. Reduced latency (temporal resources) to use the model inference output. If the private key is not provided to the entity performing model inference, then the output must be transmitted back to the entity (data source) holding the private key to perform output decryption, and the decrypted data may be transmitted back to the model inference entity.

[0029] In some example embodiments, the data that is used to train a model for regression or classification (supervised machine learning) at a data source performs the HE on the data to protect the privacy. The labels or the output corresponding to one or more training data elements are not encrypted and are stored in plaintext form along with the associated / corresponding encrypted data element, as described in conjunction with FIG. 1 to FIG. 6.

[0030] Referring now to the drawings, and more particularly to FIGS. 1 to 6, where similar reference characters denote corresponding features consistently throughout the figures, their preferred embodiments are shown.

[0031] FIG. 1 is a diagram of an example of an implementation environment 100 in which systems and / or methods, described herein, may be implemented, according to an embodiment as disclosed herein. The implementation environment 100 includes a User Equipment (UE) 110, a service environment 120, and a network 130. The service environment 120 includes one or more subenvironments 121-1 to 121-N (collectively and / or interchangeably referred hereinafter as 121). To illustrate this, FIG. 1 shows, for convenience, examples of a 1stsub-environment 121- 1, a 2ndsubenvironment 121-2, and an Nthsub-environment 121-N (where N is any natural number).

[0032] The UE 110 is connected to the network 130, and the network 130 is connected to the service environment 120. The connections may be wired, wireless, or a combination of both wired and wireless. The UE 110 and the service environment 120 are connected via the network 130.

[0033] The UE 110 is a device that communicates with the service environment 120. The UE 110 receives information from the service environment 120 and / or sends information to the service environment 120. Also, the UE 110 may generate and / or store information to be transmitted, as necessary. Also, the UE 110 may store and / or process information that is received, as necessary.

[0034] The example FIG. 1 refers to the “UE”. However, it should be understood by those skilled in the art that general terms such as “user device”, “terminal”, “terminal device”, “communication device”, “unmanned aerial vehicle (drone)”, “satellite device” and “communication terminal” can be used interchangeably with the term “UE”.

[0035] For example, the UE 110 may include a computing device (e.g., a desktop computer, a laptop computer, a tablet computer, a handheld computer, a smart speaker, a server, etc.), a mobile phone (e.g., a smartphone, a radiotelephone, etc.), a wearable device (e.g., a pair of smart glasses or a smart watch), or a similar device.

[0036] The service environment 120 is an environment that communicates with the UE 110 to provide one or more services. The service environment 120 receives information from the UE 110 and / or sends information to the UE 110. Also, the service environment 120 may generate and / or store information to be transmitted, as necessary. Also, the service environment 120 may store and / or process information that is received, as necessary. For example, the service environment 120 may provide computing resources as one of the services. It should be noted that the service is not limited to being provided to the UE 110; it may also be provided to devices other than the UE 110. For example, based on communication from the UE 110, the service may perform processes such as anomaly detection or traffic analysis and notify the results to a predetermined destination.

[0037] The example FIG. 1 refers to the “service environment”. The term “service environment” is used to refer to the broader context within which services operate. For example, cloud environments, platforms, computing systems, network systems, and cloud systems generally represent the environments in which services are conducted, and these are included within the “service environment”. However, the “service environment” is not limited to these examples. Additionally, the specific types of environments within the “service environment” are not restricted. For instance, cloud environments and cloud systems can be categorized as private cloud, public cloud, hybrid cloud, or multi-cloud, all of which are included within the “service environment”.

[0038] The one or more services provided by the service environment 120 is not specifically limited and can be adjusted according to the embodiments. For example, the one or more services may include a service that provides information to the HE 110, a service that stores information from the HE 110, or a service that performs processing based on information from the HE 110 and returns the results of the processing.

[0039] In an embodiment, the service environment 120 may also provide computing resources, such as the service. The computing resources can be hardware resources and / or software resources. For example, applications, processors, memory, and storage can be included in the provided computing resources. Each computing resource can communicate with other computing resources via wired connections, wireless connections, or a combination of wired and wireless connections.

[0040] The provided computing resources can be actual resources (also referred to as physical resources) and / or virtual resources. Furthermore, means of virtualization for virtual resources can be selected as appropriate. That is, in this disclosure, the use of adjectives such as “virtual” or“virtualized” to describe names does not imply that they are virtualized by a specific means of virtualization. For example, “virtual machine” refers to software that operates like an actual computer, realized through means of virtualization, and it is not intended to exclude those realized by specific means of virtualization, such as hypervisors or containers. Conversely, when means of virtualization, such as hypervisors or containers, are mentioned in this disclosure, it is merely cited as a general method of implementation. It should also be interpreted that embodiments implemented with other virtualization means are also disclosed. Also, the services may also be provided using resources virtualized by different means.

[0041] The service environment 120 includes one or more devices, such as servers and network devices, which provide services or perform processes. The placement of these devices within the service environment 120 can be determined as appropriate. Additionally, if the service environment 120 includes one or more sub-environments 121, the placement of devices can be determined based on predetermined policies for each sub-environment 121. For example, devices related to the first service may be placed in the 1st sub-environment 121-1, and devices related to the second service may be placed in the 2nd sub-environment 121-2. In another example, devices expected to have a higher load than a predetermined threshold may be placed in the 1st subenvironment 121-1, while devices expected to have a lower load than the predetermined threshold may be placed in the 2nd sub-environment 121-2. In this way, specific devices can be placed in specific sub -environments 121. Conversely, each sub-environment 121 can be specialized for a particular purpose.

[0042] In some example embodiments, the 1stsub-environment 121-1 may relate to a data source, the 2ndsub-environment 121-2 may relate to a model training entity, and a 3rdsub-environment 121-3 (not shown in FIG. 1) may relate to a model interference entity.

[0043] In an embodiment, all processes executed in a single service may run within a single service environment, or in multiple service environments. Multiple processes executed in a single service could be provided by different service environments.

[0044] The network 130 is a network that exchanges information between the UE 110 and the service environment 120. The network 130 includes one or more wired and / or wireless networks.

[0045] For example, the network 130 may include a cellular network (e.g., a Fifth Generation (5G) network, a Long-Term Evolution (LTE) network, a Third Generation (3G) network, a Code Division Multiple Access (CDMA) network, etc.), a Public Land Mobile Network (PLMN), a Local Area Network (LAN), a Wide Area Network (WAN), a Metropolitan Area Network (MAN), a telephone network (e.g., the Public Switched Telephone Network (PSTN)), a private network, an ad hoc network, an intranet, the Internet, a fiber optic-based network, or the like, a NonTerrestrial Network (NTN), and / or a combination of these or other types of networks.

[0046] The network 130 can be a part of a network. For example, in a 5G network that includes a RAN, a transport network, and a core network, the network 130 can be at least one of the RAN, the transport network, or the core network. For example, the service environment 120 could be in the core network, in which case the network 130 could correspond to a network that is a combination of a RAN and a transport network and is part of the 5G network.

[0047] The number and arrangement of devices and networks shown in FIG. 1 are provided as an example. It should be understood that any changes that may be implemented by those skilled inthe art, such as the addition or rearrangement of well-known devices or networks at the time of implementation, are included in this disclosure.

[0048] FIG. 2 is a sequence flow diagram illustrating various message exchanges between a data source and other network entities for managing homomorphic encrypted data, according to an embodiment as disclosed herein. The sequence flow diagram represents a method 200 that may execute multiple operations for data handling of homomorphic encrypted data used for machine learning model training in the O-RAN, which are given below. In one or more embodiments, the method 200 may utilize one or more network entities for the data handling. An example of the one or more network entities may include a data source entity 200a, a model training entity 200b, and a model inference entity 200c.

[0049] In some example embodiments, consider a first scenario where each of the data source entity 200a, the model training entity 200b, and the model inference entity 200c operates as an individual entity within the O-RAN. For the first scenario, the method 100 may execute multiple operations for the data handling (e.g., 201, 202, 203a, 204a, 205, 206, and 207), which are given below.

[0050] At operation 201, the method 200 includes generating, at the data source entity 200a, data with the HE and plain text labeling (unencrypted). Example of the HE may include, but is not limited to, Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), and a Fully Homomorphic Encryption (FHE). In some example embodiments, the data source entity 200a is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN IntelligentController (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), and other O-RAN entities that require data privacy protection.

[0051] At operation 202, the method 200 includes transmitting, by the data source entity 200a, the generated data (training data) to the model training entity 200b. In some example embodiments, the model training entity 200b is associated with one of the O-CU, the O-DU, the O-RU, the Near-RT RIC, the Non-RT RIC, the SMO, an xApp, an rApp, and one or more external entities with computational resources (e.g., GPU, accelerator, etc.).

[0052] At operation 203a, the method 200 includes performing a Machine Learning (ML) model training at the model training entity 200b using the generated data with the HE and plain text labeling. In some example embodiments, the trained ML model is stored at the model training entity 200b.

[0053] At operation 204a, the method 200 includes transferring the trained ML model from the model training entity 200b to the model inference entity 200c. At operation 205, the method 200 includes transmitting, by the data source entity 200a, the HE data to the model inference entity 200c. At operation 206, the method 200 includes performing, by the model inference entity 200c, one or more interference operations using the HE data and the transferred trained ML model. Finally, as a result, at the operation 207, the method 200 includes obtaining a plaintext output at the model inference entity 200c. In other words, the model inference output is utilized directly by a user without the need to decrypt the output. In some example embodiments, the model inference entity 200c is associated with one of the O-CU, the O-DU, the O-RU, the Near-RT RIC, the Non- RT RIC, the SMO, the xApp, and the rApp.

[0054] In some example embodiments, consider a second scenario where the data source entity 200a and the model training entity 200b operate as a unified entity within the O-RAN. Further, each of the data source entity 200a and the model inference entity 200c operates as an individual entity within the O-RAN. For the second scenario, the method 100 may execute multiple operations for the data handling (e.g., 201, 203b, 204b, 205, 206, and 207), which are given below.

[0055] At operation 201, the method 200 includes generating, at the data source entity 200a, data with the HE and plain text labeling (unencrypted). At operation 203b, the method 200 includes performing, at the data source entity 200a, one or more operations to train the ML model based on the HE and plain text labeling. In the second scenario, the trained ML model is stored at the data source entity 200a. At operation 204b, the method 200 includes transmitting, by the data source entity 200a, the trained ML model from the model training entity to the model inference entity 200c. At operation 205, the method 200 includes transmitting, by the data source entity 200a, the HE data to the model inference entity 200c. At operation 206, the method 200 includes performing, by the model inference entity 200c, one or more interference operations using the HE data and the transferred trained ML model. Finally, as a result, at the operation 207, the method 200 includes obtaining the plaintext output at the model inference entity 200c.

[0056] FIG. 3 is an example scenario 300 of various message exchanges between the O-DU 300a and the Near RT-RIC 300b for managing homomorphic encrypted data, according to an embodiment as disclosed herein.

[0057] In this example scenario 300, the O-DU 300a operates as the data source for Signal-to-Interference-plus-Noise Ratio (SINR) data of different UEs connected to the O-RUs. The O-DU 300a may be configured to share one or more O-RU locations, neighbour cell information, and theSINR information of the UEs with the Near RT-RIC 300b, to optimize handover decisions based on the SINR values. In this example scenario 300, the location of the O-RU is not important, but the relation between different O-RU locations (neighbour cell or not) plays a key role in handover decisions. Hence, the historical data of O-RU location information, Cell IDs, Temporary UE identifiers (Cell Radio Network Temporary Identifier (C-RNTI)), neighbour cell information, and SINR may be encrypted using the HE at the O-DU 300a.

[0058] In this example scenario 300, or a third scenario, where the model training entity and the model inference entity operate as a unified entity (e.g., Near RT-RIC 300b) within the O-RAN. Further, the data source entity (e.g., O-DU 300a) and the model training entity / the model inference entity (e.g., Near RT-RIC 300b) operate as an individual entity within the O-RAN.

[0059] To manage the homomorphic encrypted data, various operations are performed by at least one of the O-DU 300a and the Near RT-RIC 300b, which are given below.

[0060] At the initial operations (301, 302, and 303), the O-DU 300a (data source) may perform the HE on the dataset and associate plaintext labels (Hand Over (HO) decision / historical) with the encrypted data, once an E2 interface is established between the O-DU 300a and the Near RT-RIC 300b

[0061] At operation 304, the O-DU 300a may transmit the HE data along with corresponding plaintext handover decisions (historical) to the Near RT-RIC 300b. At operation 305, the Near RT-RIC 300b may employ a machine learning algorithm to train the ML model for predicting handover decisions. At operation 306, the O-DU 300a may transmit the HE data to the Near RT-RIC 300b (model inference entity). At operation 307, upon receiving the HE data, the Near RT- RIC 300b may perform one or more interference operations using the HE data and the transferredtrained ML model to predict the HO decision. Finally, as a result, at the operation 308, the Near RT-RIC 300b may obtain the plaintext output associated with the HO decision.

[0062] In some example embodiments, if the ML model is trained and used (for inference) at the Near RT-RIC 300b, then the model training and inference entities are the same (Near RT-RIC 300b)

[0063] In some example embodiments, the HO decision may be a set of two values, as outlined below.a. A Boolean value of performing an HO for a given UE.b. Cell ID corresponding to the O-RU that the UE may be handed over to, as part of the HO procedure.

[0064] In some example embodiments, there are several variations within the disclosed method (e.g., refer to FIG. 2) where the different operations can be performed at different O-RAN and external entities as described in the previous FIGs. The model training entity 200b and the model inference entity 200c can be the same, and in such a case, the HE data for inference is received at the entity that is acting as both the training and inference entity (e.g., 200b and 200c). Also, there might not be an offline training phase for training the model, and the inference output (e.g., HO decision as described in the embodiment) could be used to perform online training of the ML model.

[0065] In some example embodiments, the disclosed method (e.g., refer to FIG. 2) may be used in an RIC platform, the xApps, the rApps, the O-CU, the O-DU, and the SMO.

[0066] FIG. 4 is a flow diagram illustrating a method 400 for managing HE data utilized in training ML models within the O-RAN, according to an embodiment as disclosed herein. The method 400 may execute multiple operations on the HE data, which are given below.

[0067] At operation 401, the method 400 includes generating data with the HE and the plain text labelling. At operation 402, the method 400 includes transmitting the generated data to the model training entity 200b to train the ML model. At operation 403, the method 400 includes transferring the trained ML model from the model training entity 200b to the model inference entity 200c. At operation 404, the method 400 includes transmitting the HE data to the model inference entity 200c. At operation 405, the method 400 includes performing one or more interference operations using the HE data and the transferred trained ML model . At operation 406, the method 400 includes obtaining the plaintext output at the model inference entity 200c. Further, a detailed description related to the various operations of FIG. 4 is covered in the description related to FIG. 2 and FIG.3 and is omitted herein for the sake of brevity.

[0068] FIG. 5 is a flow diagram illustrating a method 500 for managing the homomorphic encrypted data utilized in training machine learning models within the O-RAN, according to another embodiment as disclosed herein. The method 500 may execute multiple operations on the HE data, which are given below.

[0069] At operation 501, the method 500 includes generating data with the HE and the plain text labelling. At operation 502, the method 500 includes performing the one or more operations to train the ML model based on the HE and plain text labelling. In the disclosed method 500, the data source entity 200a and the model training entity 200b operate as the unified entity within the O- RAN. At operation 503, the method 500 includes transmitting the trained ML model from themodel training entity 200b to the model inference entity 200c. In the disclosed method 500, each of the data source entity 200a and the model inference entity 200c operates as the individual entity within the O-RAN. At operation 504, the method 500 includes transmitting the HE data to the model inference entity 200c. At operation 505, the method 500 includes performing the one or more interference operations using the HE data and the transferred trained ML model. At operation 506, the method 500 includes obtaining the plaintext output at the model inference entity 200c.Further, a detailed description related to the various operations of FIG. 5 is covered in the description related to FIG. 2 and FIG. 3 and is omitted herein for the sake of brevity.

[0070] The various actions, acts, blocks, steps, or the like in the flow diagrams may be performed in the order presented, in a different order, or simultaneously. Further, in some embodiments, some of the actions, acts, blocks, steps, or the like may be omitted, added, modified, skipped, or the like without departing from the scope of the disclosure.

[0071] FIG. 6 illustrates a diagram of example components of an apparatus 600, according to an embodiment as disclosed herein. As shown in FIG. 6, the apparatus 600 comprises a processor 610, a memory 620, a storage component 630, an input component 640, an output component 650, a communication interface 660, and a bus 670. In one embodiment, the apparatus 600 may relate to at least one of the data source entity 200a, the model training entity 200b, and the model inference entity 200c, or any other network device.

[0072] The processor 610, as used herein, means any type of computational circuit that may comprise hardware elements and software elements. The processor 610 may be embodied as a multi-core processor, a single core processor, or a combination of one or more multi-core processors and / or one or more single core processors, a distributed processing system, or the like.The processor 610 may be a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), an Accelerated Processing Unit (APU), an Application-Specific Integrated Circuit (ASIC), or another type of processing component.

[0073] The memory 620 includes a non-transitory computer readable medium. Memory 620 includes a Random-Access Memory (RAM), a Read Only Memory (ROM), and / or another type of dynamic or static storage device (e.g., a flash memory, a magnetic memory, and / or an optical memory) that stores information and / or instructions for use by processor 610. The memory 620 comprises machine-readable instructions which are executable by the processor 610. These machine-readable instructions, when executed by the processor 610 cause the processor 610 to perform one or more method steps of an embodiment described above.

[0074] The storage component 630 stores information and / or software related to the operation and use of the apparatus 600. For example, the storage component 630 may include a hard disk (e.g., a magnetic disk, an optical disk, a magneto-optic disk, and / or a solid-state disk), a Compact Disc (CD), a Digital Versatile Disc (DVD), a floppy disk, a cartridge, a magnetic tape, and / or another type of non-transitory computer-readable medium, along with a corresponding drive.

[0075] The input component 640 is configured to receive information, such as user input. For example, the input component 640 may include, but not be limited to, a touchscreen display, a keyboard, a keypad, a mouse, a button, a switch, and / or a microphone. Additionally, or alternatively, the input component 640 may include a sensor for sensing information (e.g., a Global Positioning System (GPS), an accelerometer, a gyroscope, and / or an actuator).

[0076] The output component 650 is configured to provide output information from the apparatus 600. For example, the output component 650 may be, but is not limited to, a display, a speaker, instructions to an external device, and / or one or more Light-Emitting Diodes (LEDs).

[0077] The communication interface 660 is an interface that provides a communication connection to other devices, such as external devices and internal devices. The connection by the communication interface 660 can be a wired connection, a wireless connection, or a combination of wired and wireless connections, and can be a direct connection or an indirect connection via a communication network that exists between the apparatus 600 and other devices. In other words, the standard of the communication interface 660 is not limited.

[0078] The bus 670 acts as an interconnect between the processor 610, the memory 620, the storage component 630, the input component 640, the output component 650, and the communication interface 660 of the apparatus 600. The bus 670 may include a wired interconnection or a wireless interconnection.

[0079] The number and arrangement of components shown in FIG.6 are provided as an example. In practice, the apparatus 600 may include additional components, fewer components, different components, or differently arranged components than those shown in FIG. 6. Additionally, or alternatively, a set of components (e.g., one or more components) of the apparatus 600 may perform one or more functions described as being performed by another set of components of the apparatus 600. Further, one or more method steps described in any of the embodiments may be performed utilizing the apparatus 600 in communication with one another.

[0080] Examples of the techniques and apparatus described herein include, but are not limited to, the following enumerated embodiments:[1] A method comprising:generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmitting, by the data source entity, the generated data to a model training entity, wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.[2] The method as described in [1], comprising:in response to generating, at the data source entity, the data with the HE and the plain text labeling;performing, at the data source entity, one or more operations to train the ML model based on the HE and plain text labeling,wherein the data source entity and the model training entity operate as a unified entity within an Open Radio Access Network (O-RAN); transmitting, by the data source entity, the trained ML model from the model training entity to the model inference entity,wherein each of the data source entity and the model inference entity operates as an individual entity within the O-RAN; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain the plaintext output at the model inference entity.[3] The method as described in any of [1] to [2], wherein each of the data source entity, the model training entity, and the model inference entity operates as an individual entity within an Open Radio Access Network (O-RAN).[4] The method as described in any of [1] to [3], wherein the model inference entity comprises the ML model trained using the data generated with the HE and plain text labeling.[5] The method as described in any of [1] to [4], wherein the HE comprises at least one of a Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), and a Fully Homomorphic Encryption (FITE).[6] The method as described in any of [1] to [5], wherein the data source entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), and other O-RAN entities.[7] The method as described in any of [1] to [6], wherein the model training entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), an xApp, an rApp, and one or more external entities with computational resources.[8] The method as described in any of [1] to [7], wherein the model inference entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), an xApp, and an rApp.[9] The method as described in any of [1] to [8], wherein the trained MU model is stored at the model training entity.

[0010] The method as described in any of [1] to [9], wherein the trained ML model is stored at the data source entity.

[0011] An apparatus configured to:generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmit, by the data source entity, the generated data to a model training entity, wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmit, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

[0012] The apparatus as described in

[0011] , the apparatus is configured to:in response to generate, at the data source entity, the data with the HE and the plain text labeling;perform, at the data source entity, one or more operations to train the ML model based on the HE and plain text labeling,wherein the data source entity and the model training entity operate as a unified entity within an Open Radio Access Network (O-RAN);transmit, by the data source entity, the trained ML model from the model training entity to the model inference entity,wherein each of the data source entity and the model inference entity operates as an individual entity within the O-RAN; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain the plaintext output at the model inference entity,wherein the trained ML model is stored at the data source entity.

[0013] The apparatus as described in any of

[0011] to

[0012] , wherein each of the data source entity, the model training entity, and the model inference entity operates as an individual entity within an Open Radio Access Network (O-RAN).

[0014] The apparatus as described in any of

[0011] to

[0013] , wherein the model inference entity comprises the ML model trained using the data generated with the HE and plain text labeling.

[0015] The apparatus as described in any of

[0011] to

[0014] , wherein the HE comprises at least one of a Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), and a Fully Homomorphic Encryption (FHE).

[0016] The apparatus as described in any of

[0011] to

[0015] , wherein the data source entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), and other O-RAN entities.

[0017] The apparatus as described in any of

[0011] to

[0016] , wherein the model training entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), aNon-RT RIC, a Service Management and Orchestration (SMO), an xApp, an rApp, and one or more external entities with computational resources.

[0018] The apparatus as described in any of

[0011] to

[0017] , wherein the inference entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), aNon-RT RIC, a Service Management and Orchestration (SMO), an xApp, and an rApp.

[0019] The apparatus as described in any of

[0011] to

[0018] , wherein the trained ML model is stored at the model training entity.

[0020] A non-transitory computer-readable medium storing instructions, the instructions comprising:one or more instructions that, when executed by an apparatus, the apparatus comprising one or more processors, cause the one or more processors to:generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmit, by the data source entity, the generated data to a model training entity,wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmit, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

[0021] A method comprising:generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmitting, by the data source entity, the generated data to a model training entity, wherein the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is utilized at the model training entity and a model inference entity; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the trained ML model, to obtain a plaintext output at the model inference entity.

[0022] The method as described in

[0021] ,wherein each of the data source entity and the model training entity operates as an individual entity within an Open Radio Access Network (O-RAN); andwherein each of the model training entity and the model inference entity operates as a unified entity within the O-RAN.

[0023] The method as described in any of

[0021] to

[0022] ,wherein the HE comprises at least one of a Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), and a Fully Homomorphic Encryption (FHE);wherein the data source entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Realtime RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), and other O-RAN entities;wherein the model training entity is associated with one of the O-CU, the O-DU, the O-RU, the Near-RT RIC, the Non-RT RIC, the SMO, an xApp, an rApp, and one or more external entities with computational resources; andwherein the inference entity is associated with one of the O-CU, the O-DU, the O- RU, the Near-RT RIC, the Non-RT RIC, the SMO, the xApp, and the rApp.

[0081] The embodiments disclosed herein can be implemented through at least one software program running on at least one hardware device and performing network management functions to control the elements. The elements can be at least one of a hardware device or a combination of hardware devices and software modules.

[0082] While specific language has been used to describe the disclosure, any limitations arising on account of the same are not intended. As would be apparent to a person in the art, various working modifications may be made to the method in order to implement the inventive concept as taught herein.

[0083] The drawings and the forgoing description give examples of embodiments. Those skilled in the art will appreciate that one or more of the described elements may well be combined into a single functional element. Alternatively, certain elements may be split into multiple functional elements. Elements from one embodiment may be added to another embodiment. For example, orders of processes described herein may be changed and are not limited to the manner described herein.

[0084] Moreover, the actions of any flow diagram need not be implemented in the order shown; nor do all of the acts necessarily need to be performed. Also, those acts that are not dependent on other acts may be performed in parallel with the other acts. The scope of embodiments is by no means limited by these specific examples. Numerous variations, whether explicitly given in the specification or not, such as differences in structure, dimension, and use ofmaterial, are possible. The scope of embodiments is at least as broad as given by the following claims.

[0085] Benefits, other advantages, and solutions to problems have been described above with regard to specific embodiments. However, the benefits, advantages, solutions to problems, and any component(s) that may cause any benefit, advantage, or solution to occur or become more pronounced are not to be construed as a critical, required, or essential feature or component of any or all the claims.

[0086] The foregoing description of the specific embodiments will so fully reveal the general nature of the embodiments herein that others can, by applying current knowledge, readily modify and / or adapt for various applications such specific embodiments without departing from the generic concept, and, therefore, such adaptations and modifications should and are intended to be comprehended within the meaning and range of equivalents of the disclosed embodiments. It is to be understood that the phraseology or terminology employed herein is for the purpose of description and not of limitation. Therefore, while the embodiments herein have been described in terms of at least one embodiment, those skilled in the art will recognize that the embodiments herein can be practiced with modification within the spirit and scope of the embodiments as described herein.

Claims

We claim:

1. A method comprising:generating, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmitting, by the data source entity, the generated data to a model training entity, wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, and wherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

2. The method as claimed in claim 1, comprising:in response to generating, at the data source entity, the data with the HE and the plain text labeling,performing, at the data source entity, one or more operations to train the ML model based on the HE and plain text labeling,wherein the data source entity and the model training entity operate as a unified entity within an Open Radio Access Network (O-RAN); transmitting, by the data source entity, the trained ML model from the model training entity to the model inference entity,wherein each of the data source entity and the model inference entity operates as an individual entity within the O-RAN; andtransmitting, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain the plaintext output at the model inference entity.

3. The method as claimed in claim 1, wherein each of the data source entity, the model training entity, and the model inference entity operates as an individual entity within an Open Radio Access Network (O-RAN).

4. The method as claimed in claim 1, wherein the model inference entity comprises the ML model trained using the data generated with the HE and plain text labeling.

5. The method as claimed in claim 1, wherein the HE comprises at least one of a Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), or a Fully Homomorphic Encryption (FHE).

6. The method as claimed in claim 1, wherein the data source entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non- RT RIC, a Service Management and Orchestration (SMO), or other O-RAN entities.

7. The method as claimed in claim 1, wherein the model training entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT R1C), a Non- RT RIC, a Service Management and Orchestration (SMO), an xApp, an rApp, or one or more external entities with computational resources.

8. The method as claimed in claim 1, wherein the model inference entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O- RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non-RT RIC, a Service Management and Orchestration (SMO), an xApp, or an rApp.

9. The method as claimed in claim 1, wherein the trained ML model is stored at the model training entity.

10. The method as claimed in claim 2, wherein the trained ML model is stored at the data source entity.

11. An apparatus configured to:generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmit, by the data source entity, the generated data to a model training entity,wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmit, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

12. The apparatus as claimed in claim 11, the apparatus is configured to:in response to generate, at the data source entity, the data with the HE and the plain text labeling,perform, at the data source entity, one or more operations to train the ML model based on the HE and plain text labeling,wherein the data source entity and the model training entity operate as a unified entity within an Open Radio Access Network (O-RAN);transmit, by the data source entity, the trained ML model from the model training entity to the model inference entity,wherein each of the data source entity and the model inference entity operates as an individual entity within the O-RAN; andtransmit, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain the plaintext output at the model inference entity,wherein the trained ML model is stored at the data source entity.

13. The apparatus as claimed in claim 11, wherein each of the data source entity, the model training entity, and the model inference entity operates as an individual entity within an Open Radio Access Network (O-RAN).

14. The apparatus as claimed in claim 11, wherein the model inference entity comprises the ML model trained using the data generated with the HE and plain text labeling.

15. The apparatus as claimed in claim 11 , wherein the HE comprises at least one of a Somewhat Homomorphic Encryption (SHE), a Partially Homomorphic Encryption (PHE), or a Fully Homomorphic Encryption (FHE).

16. The apparatus as claimed in claim 11, wherein the data source entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non- RT RIC, a Service Management and Orchestration (SMO), or other O-RAN entities.

17. The apparatus as claimed in claim 11, wherein the model training entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O- RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), aNon-RT RIC, a Service Management and Orchestration (SMO), an xApp, an rApp, or one or more external entities with computational resources.

18. The apparatus as claimed in claim 11, wherein the inference entity is associated with one of an O-RAN Centralized Unit (O-CU), an O-RAN Distributed Unit (O-DU), an O-RAN Radio Unit (O-RU), a Near-Real-time RAN Intelligent Controller (Near-RT RIC), a Non- RT RIC, a Service Management and Orchestration (SMO), an xApp, or an rApp.

19. The apparatus as claimed in claim 11, wherein the trained ML model is stored at the model training entity.

20. A non-transitory computer-readable medium storing instructions, the instructions comprising:one or more instructions that, when executed by an apparatus, the apparatus comprising one or more processors, cause the one or more processors to:generate, at a data source entity, data with a Homomorphic Encryption (HE) and plain text labeling;transmit, by the data source entity, the generated data to a model training entity,wherein the generated data with the HE and plain text labeling is used at the model training entity to train a Machine Learning (ML) model, andwherein the trained ML model is transferred from the model training entity to a model inference entity; andtransmit, by the data source entity, the HE data to the model inference entity, to perform one or more interference operations using the HE data and the transferred trained ML model, to obtain a plaintext output at the model inference entity.

Images