System and method to secure multi-party machine learning with federated encryption
Patent Information
- Application Number
- IN202541009707
- Authority / Receiving Office
- IN · IN
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2025-02-06
- Publication Date
- 2026-10-14
- Estimated Expiration
- 2045-02-06
Abstract
Description
Description:TECHNICAL FIELD
[001] The present invention relates generally to of machine learning and cryptographic protocols and, more particularly, to system and method for secure multi-party machine learning utilizing federated encryption protocols to ensure data confidentiality, integrity, and privacy during collaborative training and inference processes.BACKGROUND OF THE INVENTION
[002] In the rapid adoption of machine learning (ML) across industries has led to significant advancements in automation, predictive analytics, and decision-making processes. However, in many cases, ML models require access to vast amounts of sensitive and distributed data owned by multiple parties, such as healthcare providers, financial institutions, or telecommunications companies. The sharing of such data is often restricted due to concerns about privacy, regulatory compliance, and security risks.
[003] Federated learning (FL) has emerged as a promising approach to enable collaborative training of the ML models without directly sharing raw data. In FL, individual parties compute local updates to a global model using their private datasets. The local updates are aggregated centrally to refine the global model. However, the FL still suffers from vulnerabilities, including data leakage through gradient inversion attacks, malicious participants injecting poisoned updates, and unauthorized access to model parameters.
[004] Existing encryption-based techniques, such as homomorphic encryption, secure multiparty computation (SMPC), and differential privacy, offer partial solutions to these vulnerabilities. However, the existing encryption-based techniques either introduce significant computational overhead, fail to provide end-to-end security, or lack scalability.
[005] Thus, there remains a need for a system and method that integrates federated learning with advanced encryption protocols to enable secure multi-party machine learning.SUMMARY OF THE INVENTION
[006] This summary is provided to introduce a selection of concepts, in a simple manner, which is further described in the detailed description of the disclosure. This summary is neither intended to identify key or essential inventive concepts of the subject matter nor to determine the scope of the disclosure.
[007] In accordance with an embodiment of the present disclosure, a system for secure multi-party machine learning is disclosed. The system includes each of a plurality of participating nodes configured to perform local training on one or more private datasets, wherein each of the plurality of participating nodes configured to generate one or more model updates. Further, the system includes a central aggregator configured to receive the one or more model updates from the plurality of participating nodes and perform encrypted aggregation. Furthermore, the system includes a model update encryption module configured to aggregate one or more encrypted updates. In addition, the system includes a differential privacy module configured to inject noise into the one or more model updates to prevent one or more data reconstruction attacks. Further, the system includes an integrity verification module configured to detect malicious activities using one or more cryptographic hash functions and digital signatures.
[008] In accordance with an embodiment of the present disclosure, a method for secure multi-party machine learning is disclosed. The method includes establishing secure communication channels between each of a plurality of participating nodes and a central aggregator. Further, the method includes performing local training at the each of the plurality of participating nodes. Furthermore, the method includes encrypting one or more model updates before transmission over a communication network from the each of a plurality of participating nodes and the central aggregator. In addition, the method includes aggregating the one or more encrypted updates using a model update encryption module. Further, the method includes verifying an integrity of an aggregated model upon aggregating the one or more encrypted updates.
[009] One or more advantages of the prior art are overcome, and additional advantages are provided through the invention. Additional features are realized through the technique of the invention. Other embodiments and aspects of the disclosure are described in detail herein and are considered a part of the invention.BRIEF DESCRIPTION OF THE FIGURES
[010] The accompanying figures where like reference numerals refer to identical or functionally similar elements throughout the separate views and which together with the detailed description below are incorporated in and form part of the specification, serve to further illustrate various embodiments and to explain various principles and advantages all in accordance with the invention.FIG. 1 is a block diagram illustrating an exemplary environment for secure multi-party machine learning with federated encryption protocols, in accordance with an embodiment of the present disclosure;FIG. 2 is a block diagram illustrating an implementation of the system for secure multi-party machine learning with federated encryption protocols, in accordance with an embodiment of the present disclosure; andFIG. 3 is a process flow diagram illustrating an exemplary method for secure multi-party machine learning, in accordance with an embodiment of the present disclosure.
[011] Skilled artisans will appreciate the elements in the figures are illustrated for simplicity and clarity and have not necessarily been drawn to scale. For example, the dimensions of some of the elements in the figures may be exaggerated relative to other elements to help to improve understanding of embodiments of the present invention.DETAILED DESCRIPTION OF THE INVENTION
[012] While various embodiments of the invention have been shown and described herein, it will be obvious to those skilled in the art that such embodiments are provided by way of example only. Numerous variations, changes, and substitutions may occur to those skilled in the art without departing from the invention. It should be understood that various alternatives to the embodiments of the invention described herein may be employed. It shall be understood that different aspects of the invention can be appreciated individually, collectively, or in combination with each other.
[013] An environment and various implementations for ensures road safety by preventing drunk driving and enabling rapid emergency response. The environment and processes may be described with reference to FIG. 1 showing an architectural level schematic of a system in accordance with an implementation. Because FIG. 1 is an architectural diagram, certain details are intentionally omitted to improve the clarity of the description. The discussion of FIG. 1 will be organized as follows. First, the elements of the figure will be described, followed by their interconnections. Then, the use of the elements in the environment will be described in greater detail. The environment provides power of deep learning neural networks for secure multi-party machine learning.
[014] Referring now to the drawings, and more particularly to FIG. 1 through FIG. 3, where similar reference characters denote corresponding features consistently throughout the figures, there are shown preferred embodiments and these embodiments are described in the context of the following exemplary system and / or method.
[015] FIG. 1 is a block diagram 100 illustrating an exemplary environment for secure multi-party machine learning with federated encryption protocols, in accordance with an embodiment of the present disclosure. The federated encryption protocols may refer to encryption techniques or Machine Learning (ML) models where the plurality of participating nodes 104a, 104b, 104c…104n collaborate to train the system 102 without sharing private data. Federated encryption ensures that the data and model updates shared between the plurality of participating nodes 104a, 104b, 104c…104n and the aggregator 108 remain secure, private, and tamper-proof.
[016] According to FIG. 1, the exemplary environment 100 includes a system 102, a plurality of participating nodes 104a, 104b, 104c…104n, a network 106, and a central aggregator 108. The computing device 104 and the cloud server 108 may be connected through a network 106. In an embodiment, the network 106 may be an internet. The plurality of participating nodes 104a, 104b, 104c…104n may include laptops, desktops, servers, edge devices, or Internet Of Things (IoT) devices that are capable of performing local training on private datasets. The plurality of participating nodes 104a, 104b, 104c…104n may be equipped with sufficient Central Processing Unit (CPU) or Graphic Processing Unit (GPU) resources for training machine learning models.
[017] The plurality of participating nodes 104a, 104b, 104c…104n may include local data storage, such as hard drives or SSDs, for securely storing private datasets, model updates, and encryption keys. The plurality of participating nodes 104a, 104b, 104c…104n may include network cards or modules to establish secure communication with the central aggregator 108 over the network 106. Further, each of the plurality of participating nodes 104a, 104b, 104c…104n may include cryptographic modules or specialized encryption accelerators for efficient execution of encryption, decryption, and hashing operations. Further, each of the plurality of participating nodes 104a, 104b, 104c…104n may be configured to deploy an aggregated model for inference tasks with end-to-end encryption.
[018] In an embodiment of the present disclosure, the system 102 may be located within each of the plurality of participating nodes 104a, 104b, 104c…104n. In another embodiment, the system 102 may be connected to the each of the plurality of participating nodes 104a, 104b, 104c…104n.
[019] The central aggregator 108 may include high-performance servers or data centers with substantial computational power (e.g., multi-core CPUs, GPUs, or TPUs) to aggregate model updates and perform encrypted computations. Further, the central aggregator 108 may include centralized or distributed storage systems to manage encrypted model updates received from the each of the plurality of participating nodes 104a, 104b, 104c…104n. The central aggregator 108 may include high-speed network interfaces with support for secure communication protocols (e.g., TLS, IPsec). Further, the central aggregator 108 may include encryption processors configured to support homomorphic encryption, secure multiparty computation (SMPC), and hybrid encryption protocols.
[020] In an example, hospitals in different regions collaboratively train a diagnostic ML model on patient data while preserving the confidentiality of sensitive medical records using the system 102. Further, multiple banks collaborate to build a fraud detection system i.e., the system 102 using transaction data. The federated encryption protocols ensure compliance with financial data regulations. Furthermore, municipalities deploy the system 102 to train an ML model on traffic data collected from the IoT devices, ensuring data privacy for citizens.
[021] FIG. 2 is a block diagram 100 illustrating an implementation of the system for secure multi-party machine learning with federated encryption protocols, in accordance with an embodiment of the present disclosure. According to FIG. 2, the system 102 may include one or more hardware processors 202, a memory 204 and a storage unit 206. The one or more hardware processors 202, the memory 204 and the storage unit 206 may be communicatively coupled through a system bus 208 or any similar mechanism. The memory 204 may include the plurality of modules 210 in the form of programmable instructions executable by the one or more hardware processors 202. Further, the plurality of modules 210 may include a model update encryption module 212, a differential privacy module 214, an integrity verification module 216, and a secure communication establishing module 218.
[022] The model update encryption module 212 may be configured to encrypt the model updates generated by the plurality of participating nodes 104a, 104b, 104c…104n before transmission to the central aggregator 108. The model update encryption module 212 may ensure that the model updates remain confidential and protected from unauthorized access during communication. The model update encryption module 212 may be configured to employ hybrid encryption techniques, combining symmetric encryption for data confidentiality and asymmetric encryption for secure key exchange. The model update encryption module 212 may include hybrid encryption, secure multiparty computation, and homomorphic encryption modules to aggregate one or more encrypted updates.
[023] The differential privacy module 214 may be responsible for injecting calibrated noise into the model updates before encryption. The noise obfuscates sensitive information in the updates, ensuring that individual data points in the private datasets maynot be reconstructed, even if the model updates are intercepted or analyzed.
[024] The integrity verification module 216 may ensure the authenticity and integrity of model updates and the aggregated model. The integrity verification module 216 may be configured to employ cryptographic hash functions to detect any tampering or malicious activities. Additionally, digital signatures may be used to verify the identity of the sender and confirm that updates originate from authorized participants. The integrity verification module may be configured to use one or more hash trees to validate the one or more model updates. The hash tree may include a hierarchical data structure used for efficient and secure verification of data integrity. The hash tree may be widely used in cryptographic applications, including blockchain, secure communications, and distributed computing.
[025] The secure communication establishing module 218 may be configured to establish encrypted communication channels between the each of the plurality of participating nodes 104a, 104b, 104c…104n and the central aggregator 108. The secure communication establishing module 218 may be configured to utilize protocols such as Transport Layer Security (TLS) and public-key infrastructure (PKI) to authenticate endpoints and securely exchange encryption keys, ensuring data integrity and confidentiality during transmission.
[026] A module may typically refer to a discrete unit or component of a larger system or software application that performs a specific function or set of related functions. Modules may be often designed to be modular, meaning they may be easily integrated into or removed from the overall system 102 without affecting other components.
[027] The one or more hardware processors 202, as used herein, means any type of computational circuit, such as, but not limited to, a microprocessor unit, microcontroller, complex instruction set computing microprocessor unit, reduced instruction set computing microprocessor unit, very long instruction word microprocessor unit, explicitly parallel instruction computing microprocessor unit, graphics processing unit, digital signal processing unit, or any other type of processing circuit. The one or more hardware processors 202 may also include embedded controllers, such as generic or programmable logic devices or arrays, application specific integrated circuits, single-chip computers, and the like.
[028] The memory 204 may be non-transitory volatile memory and non-volatile memory. The memory 204 may be coupled for communication with the one or more hardware processors 202, such as being a computer-readable storage medium. The one or more hardware processors 202 may execute machine-readable instructions and / or source code stored in the memory 204. A variety of machine-readable instructions may be stored in and accessed from the memory 204. The memory 204 may include any suitable elements for storing data and machine-readable instructions, such as read only memory, random access memory, erasable programmable read only memory, electrically erasable programmable read only memory, a hard drive, a removable media drive for handling compact disks, digital video disks, diskettes, magnetic tape cartridges, memory cards, and the like. In the present embodiment, the memory 204 includes the plurality of modules 210 stored in the form of machine-readable instructions on any of the above-mentioned storage media and may be in communication with and executed by the one or more hardware processors 202.
[029] The storage unit 206 may be a cloud storage or a local file directory within a remote server. The storage unit 206 may store the registration details, the one or more policy parameters and the merchandise information.
[030] In an embodiment of the present disclosure, the system 102 may be an application for the secure multi-party machine learning.
[031] Embodiments of the application further provide a non-transitory computer-readable storage medium, in which a computer program is stored. When executed by the processor, the computer program may prevent misdemeanor behavior during the events.
[032] The non-transitory computer-readable storage medium may be any combination of one or more computer-readable media. The computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium may be, but is not limited to, for example, an electrical, magnetic, optical, electromagnetic, infrared or semiconductor system, device or apparatus or any combination thereof. More specific examples (non-exhaustive list) of the computer-readable storage medium include an electrical connector with one or more wires, a portable computer disk, a hard disk, a RAM, a ROM, an Erasable Programmable ROM (EPROM) or a flash memory, an optical fiber, a portable CD-ROM, an optical storage device, a magnetic storage device, or any proper combination thereof. In the application, the computer-readable storage medium may be any tangible medium including or storing a program that may be used by or in combination with an instruction execution system, device, or apparatus.
[033] FIG. 3 is a process flow diagram 300 illustrating an exemplary method for secure multi-party machine learning, in accordance with an embodiment of the present disclosure.
[034] At step 302, the method 300 may include establishing secure communication channels between each of the plurality of participating nodes and the central aggregator 108.
[035] At step 304, the method 300 may include performing local training at the each of the plurality of participating nodes 104a, 104b, 104c…104n.
[036] At step 306, the method 300 may include encrypting one or more model updates before transmission over the communication network from the each of the plurality of participating nodes 104a, 104b, 104c…104n and the central aggregator 108.
[037] At step 308, the method 300 may include aggregating the one or more encrypted updates using the model update encryption module 212.
[038] At step 310, the method 300 may include verifying an integrity of the aggregated model upon aggregating the one or more encrypted updates.
[039] The method 300 may include deploying the aggregated model for inference tasks with end-to-end encryption.
[040] The method 300 may include the plurality of participating nodes 104a, 104b, 104c…104n add differential privacy noise to the one or more model updates.
[041] The method 300 may include validating the one or more model updates using one or more cryptographic hash functions and digital signatures. Further, the method 300 may include upon validating the one or more model updates, detecting malicious activities using the integrity verification module 216. Further, the method 300 may include mitigating the malicious activities during the transmission of the one or more model updates to the central aggregator 108 from the each of the plurality of participating nodes 104a, 104b, 104c…104n.
[042] The methods may be implemented in any suitable hardware, software, firmware, or combination thereof.
[043] Thus, various embodiments of the present invention provide the system for secure multi-party machine learning. The system enables secure collaboration between multiple nodes participating in machine learning without compromising the privacy of their local datasets. The system performs local training and generating model updates securely, . The system ensures that sensitive data is never shared directly, thereby maintaining privacy.
[044] The use of encrypted aggregation by the central aggregator ensures that even the aggregated model updates are protected during transmission, preventing exposure of sensitive model data during the aggregation process. The incorporation of a differential privacy module introduces noise into the model updates, reducing the risk of data reconstruction attacks and ensuring that individual data points cannot be extracted from the aggregated models. The integrity verification module uses cryptographic hash functions and digital signatures to validate the authenticity and integrity of the model updates. This reduces the risk of malicious activities, ensuring that only legitimate model updates are considered during aggregation. The system supports a wide variety of participating nodes, which can be added or removed as needed, offering scalability for larger networks while maintaining secure, privacy-preserving machine learning practices.
[045] The system leverages federated learning techniques, the system ensures that training occurs locally on each node, allowing the system to operate efficiently across distributed environments while minimizing the data exchanged between nodes. The model update encryption module, supporting both symmetric and asymmetric encryption, ensures secure communication between nodes and the central aggregator, protecting the system from potential eavesdropping and tampering during data exchange. The ability to adjust the level of noise based on a privacy budget allows for fine-grained control over the trade-off between privacy and model accuracy, catering to different use cases and privacy requirements.
[046] The system deploys the aggregated model with end-to-end encryption, the system extends security to the inference phase, ensuring that even when the model is used for prediction tasks, the data remains encrypted and secure. The integrity verification module’s ability to detect and mitigate malicious activities during the transmission of model updates enhances overall system security, ensuring that only trustworthy and valid model updates are incorporated into the final model.
[047] The following features have been described in accordance with an embodiment of the present disclosure. A system (102) for secure multi-party machine learning includes,each of a plurality of participating nodes (104a, 104b, 104c…104n) configured to perform local training on one or more private datasets, wherein each of the plurality of participating nodes (104a, 104b, 104c…104n) configured to generate one or more model updates;a central aggregator (108) configured to receive the one or more model updates from the plurality of participating nodes (104a, 104b, 104c…104n) and perform encrypted aggregation;a model update encryption module (212) configured to aggregate one or more encrypted updates;a differential privacy module (214) configured to inject noise into the one or more model updates to prevent one or more data reconstruction attacks; andan integrity verification module (216) configured to detect malicious activities using one or more cryptographic hash functions and digital signatures.
[048] The system as described in paragraph
[047] , the plurality of participating nodes (104a, 104b, 104c…104n) comprises a federated learning module for the local training.
[049] The system as described in paragraphs [047-048], the central aggregator (108) is configured to perform aggregation using a homomorphic encryption.
[050] The system as described in paragraphs [047-049], the model update encryption module (212) is configured to symmetric and asymmetric encryption for secure communication.
[051] The system as described in paragraphs [047-050], the differential privacy module (214) is configured to adjust a noise level based on a privacy budget.
[052] The system as described in paragraphs [047-051], the integrity verification module (216) is configured to use one or more hash trees to validate the one or more model updates.
[053] A method for secure multi-party machine learning, the method includes,establishing, by a plurality of participating nodes (104a, 104b, 104c…104n), one or more secure communication channels between each of the plurality of participating nodes (104a, 104b, 104c…104n) and a central aggregator (108);performing, by each of the plurality of participating nodes (104a, 104b, 104c…104n), local training on one or more private datasets based on the established one or more secure communication channels;encrypting, by the plurality of participating nodes (104a, 104b, 104c…104n), one or more model updates before transmission over a communication network from the each of the plurality of participating nodes (104a, 104b, 104c…104n) to the central aggregator (108);aggregating, by the plurality of participating nodes (104a, 104b, 104c…104n), the one or more encrypted model updates; andverifying, by the plurality of participating nodes (104a, 104b, 104c…104n), an integrity of an aggregated model upon aggregating the one or more encrypted updates.
[054] The method as described in paragraph
[053] , deploying the aggregated model for inference tasks with end-to-end encryption.
[055] The method as described in paragraphs [053-054], the plurality of participating nodes (104a, 104b, 104c…104n) configured to add differential privacy noise to the one or more model updates.
[056] The method as described in paragraphs [053-055], verifying the integrity of the aggregated model, includes,validating the one or more model updates using one or more cryptographic hash functions and digital signatures; andupon validating the one or more model updates, detecting malicious activities using an integrity verification module (216) and mitigating the malicious activities during the transmission of the one or more model updates to the central aggregator (108) from the each of the plurality of participating nodes (104a, 104b, 104c…104n).
[057] The written description describes the subject matter herein to enable any person skilled in the art to make and use the embodiments. The scope of the subject matter embodiments is defined by the claims and may include other modifications that occur to those skilled in the art. Such other modifications are intended to be within the scope of the claims if they have similar elements that do not differ from the literal language of the claims or if they include equivalent elements with insubstantial differences from the literal language of the claims.
[058] The embodiments herein can comprise hardware and software elements. The embodiments that are implemented in software include but are not limited to, firmware, resident software, microcode, etc. The functions performed by various modules described herein may be implemented in other modules or combinations of other modules. For the purposes of this description, a computer-usable or computer readable medium can be any apparatus that can comprise, store, communicate, propagate, or transport the program for use by or in connection with the instruction execution system, apparatus, or device.
[059] The medium can be an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system (or apparatus or device) or a propagation medium. Examples of a computer-readable medium include a semiconductor or solid-state memory, magnetic tape, a removable computer diskette, a random-access memory (RAM), a read-only memory (ROM), a rigid magnetic disk and an optical disk. Current examples of optical disks include compact disk-read only memory (CD-ROM), compact disk-read / write (CD-R / W) and DVD.
[060] Input / output (I / O) devices (including but not limited to keyboards, displays, pointing devices, etc.) can be coupled to the system either directly or through intervening I / O controllers. Network adapters may also be coupled to the system to enable the data processing system to become coupled to other data processing systems or remote printers or storage devices through intervening private or public networks. Modems, cable modem and Ethernet cards are just a few of the currently available types of network adapters.
[061] A representative hardware environment for practicing the embodiments may include a hardware configuration of an information handling / computer system in accordance with the embodiments herein. The system herein comprises at least one processor or central processing unit (CPU). The CPUs are interconnected via system bus 208 to various devices such as a random-access memory (RAM), read-only memory (ROM), and an input / output (I / O) adapter. The I / O adapter can connect to peripheral devices, such as disk units and tape drives, or other program storage devices that are readable by the system. The system can read the inventive instructions on the program storage devices and follow these instructions to execute the methodology of the embodiments herein.
[062] The system further includes a user interface adapter that connects a keyboard, mouse, speaker, microphone, and / or other user interface devices such as a touch screen device (not shown) to the bus to gather user input. Additionally, a communication adapter connects the bus to a data processing network, and a display adapter connects the bus to a display device which may be embodied as an output device such as a monitor, printer, or transmitter, for example.
[063] A description of an embodiment with several components in communication with each other does not imply that all such components are required. On the contrary, a variety of optional components are described to illustrate the wide variety of possible embodiments of the invention. When a single device or article is described herein, it will be apparent that more than one device / article (whether or not they cooperate) may be used in place of a single device / article. Similarly, where more than one device or article is described herein (whether or not they cooperate), it will be apparent that a single device / article may be used in place of the more than one device or article, or a different number of devices / articles may be used instead of the shown number of devices or programs. The functionality and / or the features of a device may be alternatively embodied by one or more other devices which are not explicitly described as having such functionality / features. Thus, other embodiments of the invention need not include the device itself.
[064] The illustrated steps are set out to explain the exemplary embodiments shown, and it should be anticipated that ongoing technological development will change the manner in which particular functions are performed. These examples are presented herein for purposes of illustration, and not limitation. Further, the boundaries of the functional building blocks have been arbitrarily defined herein for the convenience of the description. Alternative boundaries can be defined so long as the specified functions and relationships thereof are appropriately performed. Alternatives (including equivalents, extensions, variations, deviations, etc., of those described herein) will be apparent to persons skilled in the relevant art(s) based on the teachings contained herein. Such alternatives fall within the scope and spirit of the disclosed embodiments. Also, the words "comprising," "having," "containing," and "including," and other similar forms are intended to be equivalent in meaning and be open-ended in that an item or items following any one of these words is not meant to be an exhaustive listing of such item or items or meant to be limited to only the listed item or items. It must also be noted that as used herein and in the appended claims, the singular forms “a,” “an,” and “the” include plural references unless the context clearly dictates otherwise.
[065] Finally, the language used in the specification has been principally selected for readability and instructional purposes, and it may not have been selected to delineate or circumscribe the inventive subject matter. It is therefore intended that the scope of the invention be limited not by this detailed description, but rather by any claims that issue on an application based here on. Accordingly, the embodiments of the present invention are intended to be illustrative, but not limiting, of the scope of the invention, which is set forth in the following claims.
Claims
, Claims:WE CLAIM:
1. A method for detecting malicious activities using one or more cryptographic hash functions and digital signatures in secure multi-party machine learning, comprising:establishing, by a plurality of participating nodes (104a, 104b, 104c…104n), one or more secure communication channels between each of the plurality of participating nodes (104a, 104b, 104c…104n) and a central aggregator (108);performing, by each of the plurality of participating nodes (104a, 104b, 104c…104n), local training on one or more private datasets based on the established one or more secure communication channels;encrypting, by the plurality of participating nodes (104a, 104b, 104c…104n), one or more model updates before transmission over a communication network from the each of the plurality of participating nodes (104a, 104b, 104c…104n) to the central aggregator (108);aggregating, by the plurality of participating nodes (104a, 104b, 104c…104n), the one or more encrypted model updates; andverifying, by the plurality of participating nodes (104a, 104b, 104c…104n), an integrity of an aggregated model upon aggregating the one or more encrypted updates.
2. The method as claimed in claim 1, comprising:validating, by the plurality of participating nodes (104a, 104b, 104c…104n), the one or more model updates using one or more cryptographic hash functions and digital signatures; andupon validating the one or more model updates, detecting, by the plurality of participating nodes (104a, 104b, 104c…104n), the malicious activities and mitigating the malicious activities during the transmission of the one or more encrypted model updates to the central aggregator (108) from the each of the plurality of participating nodes (104a, 104b, 104c…104n).
3. The method as claimed in claim 1, comprising: deploying, by the plurality of participating nodes (104a, 104b, 104c…104n), the aggregated model for inference tasks with end-to-end encryption.
4. The method as claimed in claim 1, comprising:adding, by the plurality of participating nodes (104a, 104b, 104c…104n), differential privacy noise to the one or more model updates.
5. A system (102) for detecting malicious activities using one or more cryptographic hash functions and digital signatures in secure multi-party machine learning, comprising:each of a plurality of participating nodes (104a, 104b, 104c…104n) configured to perform local training on one or more private datasets, wherein each of the plurality of participating nodes (104a, 104b, 104c…104n) configured to generate one or more model updates;a central aggregator (108) configured to receive the one or more model updates from the plurality of participating nodes (104a, 104b, 104c…104n) and perform encrypted aggregation;a model update encryption module (212) configured to aggregate one or more encrypted model updates;a differential privacy module (214) configured to inject noise into the one or more model updates to prevent one or more data reconstruction attacks; andan integrity verification module (216) configured to detect malicious activities using the one or more cryptographic hash functions and the digital signatures.
6. The system as claimed in claim 5, wherein the plurality of participating nodes (104a, 104b, 104c…104n) comprises a federated learning module for the local training.
7. The system as claimed in claim 5, wherein the plurality of participating nodes (104a, 104b, 104c…104n) is configured to deploy an aggregated model for inference tasks with end-to-end encryption.
8. The system as claimed in claim 5, wherein the model update encryption module (212) is configured to symmetric and asymmetric encryption for secure communication.
9. The system as claimed in claim 5, wherein the differential privacy module (214) is configured to adjust a noise level based on a privacy budget.
10. The system as claimed in claim 5, wherein the integrity verification module (216) is configured to use one or more hash trees to validate the one or more model updates. Dated this Signature: Name: Dr Priti Tayade (IN / PA No. ) 4704 Authorized Patent Agent for Applicant
