Information protection system and method of smart home based on federated learning

Abstract

The present invention relates to an information protection system and method for a federated learning-based smart home, and more specifically, to an information protection system and method for a federated learning-based smart home that can enhance personal information protection by utilizing federated learning technology and differential privacy technology. To this end, the present invention provides a smart home information protection system for protecting personal information in a smart home system comprising: a smart home device including a light sensor, an air quality sensor, a security sensor, a motion detection sensor, and a home appliance provided in the home; a user terminal communicating for the control of the smart home device; and a server receiving data through the smart home device and the user terminal and controlling the smart home device with optimized data. The smart home device is operated by locally training collected data based on Federated Learning technology to generate a local model, transmitting encrypted data to the server, and updating the model parameters of a global model received from the server. The server is characterized by generating a global model based on the encrypted data received from the smart home device using Differential Privacy technology and an aggregation algorithm, and transmitting the generated global model to the smart home device so that the local model is updated.

Description

Technology Field

[0001] The present invention relates to an information protection system and method for a federated learning-based smart home, and more specifically, to an information protection system and method for a federated learning-based smart home that can enhance personal information protection by utilizing federated learning technology and differential privacy technology. Background Technology

[0002] Recently, the rapid development of artificial intelligence and smart home technology has made people's lifestyles much more convenient.

[0003] However, along with these advancements, the risk of personal information infringement has also increased significantly, causing serious problems regarding information protection in smart homes.

[0004] Smart home devices collect extensive user data, including behavioral habits, lifestyle patterns, daily activities, and device usage; if this data is maliciously exploited, it can pose a serious threat to property security and other safety issues.

[0005] Meanwhile, smart home devices often rely on statistical analysis of extensive user behavior data to optimize functions and improve the user experience.

[0006] For example, a smart thermostat can dynamically adjust the energy-saving mode based on the household temperature usage pattern.

[0007] However, reliance on data sharing and centralized data processing of smart devices significantly increases the risk of data leakage and misuse.

[0008] As a technology to address these issues and achieve a balance between privacy and service quality, there is Federated Learning (FL), an emerging distributed machine learning paradigm.

[0009] Federated learning is a distributed machine learning framework that allows models to be trained locally on multiple devices without uploading user data to a central server.

[0010] Each device completes model training locally using its own data and sends model updates to a central server for aggregation into a global model.

[0011] This distributed approach significantly reduces privacy risks by keeping user data local, while improving model performance by leveraging distributed device resources.

[0012] Therefore, it is expected that by utilizing such federated learning technology, personal information can be protected and smart homes can be operated efficiently. Prior art literature

[0013] Registered Patent Publication No. 10-2715322 (October 4, 2024) The problem to be solved

[0014] The present invention was devised to solve the aforementioned problems, and aims to provide an information protection system and method for a federated learning-based smart home that can protect personal information by utilizing federated learning technology instead of centralized data storage and transmission.

[0015] The purpose is to provide an information protection system and method for a federated learning-based smart home that utilizes differential privacy technology capable of preventing the inference of personal data.

[0016] In addition, the purpose is to provide an information protection system and method for a federated learning-based smart home that can enhance security during data transmission and reception.

[0018] In addition, other objects and advantages of the present invention will be described below, and it should be noted that they will be encompassed to a broader extent by means and combinations within the scope that can be easily derived from the matters described in the claims of the present invention and the disclosure of the embodiments thereof. means of solving the problem

[0019] The present invention, for achieving the above objective, is an information protection system for a smart home for protecting personal information in a smart home system comprising: a smart home device including a light sensor, an air quality sensor, a security sensor, a motion detection sensor, and a home appliance provided in the home; a user terminal communicating for the control of the smart home device; and a server receiving data through the smart home device and the user terminal and controlling the smart home device with optimized data. The smart home device is characterized by generating a local model by locally training collected data based on Federated Learning technology, transmitting encrypted data to the server, and updating the model parameters of a global model received from the server. The server is characterized by generating a global model based on the encrypted data received from the smart home device using Differential Privacy technology and an aggregation algorithm, and transmitting the generated global model to the smart home device so that the local model is updated.

[0020] And according to a preferred embodiment of the present invention, the local model according to the federated learning technology of the smart home device is characterized by being composed of an input layer that processes sensor data detected from each sensor, a convolutional layer that extracts local features from the sensor data using a convolutional kernel, a recurrent layer that captures the temporal dependency of the sensor data, a fully connected layer that abstracts the extracted local features and performs classification or regression, and an output layer that aggregates and transmits the local model to the server.

[0021] In addition, according to a preferred embodiment of the present invention, the server is characterized by adding a security layer using the differential privacy technology that adds the global model update or noise to the federated learning framework.

[0022] The present invention for achieving the above objective is an information protection method using the aforementioned federated learning-based smart home information protection system, characterized by comprising: a step of collecting data from each sensor of the smart home device; a step of locally training the collected data based on the federated learning technology; a step of encrypting the local data generated by the local training and transmitting it to the server; a step of aggregating the encrypted data based on differential privacy technology at the server and generating a global model; and a step of transmitting the generated global model to the smart home device and feeding back data using the received global model at the smart home device. Effects of the invention

[0023] As described above, according to the present invention, the following effects can be expected.

[0024] Through federated learning, the centralized storage and transmission of raw data can be eliminated, which is effective in fundamentally reducing privacy risks in smart home data processing.

[0025] In addition, through differential privacy technology, noise can be added to model updates to effectively prevent the inference of personal data.

[0026] In addition, encrypted communication has the effect of guaranteeing the integrity and security of data during transmission.

[0027] In particular, by implementing robust multi-layered privacy protection on smart home devices, this technology enables efficient collaborative learning and personalized services across multiple devices while protecting user privacy.

[0029] In addition, it should be noted that other effects of the present invention will be encompassed to a broader extent by the embodiments described above and the matters described in the claims of the present invention, as well as by effects that can be easily derived from them and potential advantages that contribute to industrial development. Brief explanation of the drawing

[0030] FIG. 1 is a simplified diagram illustrating an information protection system for a federated learning-based smart home according to the present invention. FIG. 2 is a diagram showing the information protection system of a federated learning-based smart home according to the present invention of FIG. 1 in more detail. Figure 3 is a diagram showing the structure of a local model of a smart home device in an information protection system of a federated learning-based smart home according to the present invention. FIG. 4 is a diagram illustrating an information protection method for a smart home based on federated learning according to the present invention. Specific details for implementing the invention

[0031] Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the attached drawings. Prior to the description, the advantages and features of the present invention and the methods for achieving them will become clear by referring to the embodiments described below in detail together with the attached drawings. Furthermore, it should be noted that the terms used in this specification are for describing the embodiments and are not intended to limit the present invention; that singular forms of such terms include plural forms unless specifically stated otherwise in the text, and that words indicating direction in the description are intended to aid in understanding the description and may change depending on the context.

[0033] Hereinafter, an information protection system and method for a federated learning-based smart home according to a preferred embodiment of the present invention will be described in detail with reference to the attached drawings. FIG. 1 is a simplified drawing of an information protection system for a federated learning-based smart home according to the present invention, and FIG. 2 is a more detailed drawing of FIG. 1.

[0034] Referring to FIGS. 1 and 2, the information protection system of a federated learning-based smart home according to the present invention includes a smart home device (10), a user terminal (20), and a server (30).

[0035] The smart home device (10) is a sensor and home appliance (15) provided in the home, including a light sensor, an air quality sensor, a security sensor, a motion detection sensor, and a home appliance.

[0036] These smart home devices (10) collect data in real time through each sensor and home appliance, and this collected data is transmitted to a server (30).

[0037] At this time, when transmitting data, the smart home device (10) according to the present invention locally trains the collected data based on federated learning (FL) technology to create a local model and transmits the encrypted data to the server (30).

[0038] And the smart home device (10) updates the model using the model parameters of the global model generated and received from the server (30), and each sensor and home appliance is operated.

[0039] Meanwhile, to achieve efficient information protection and behavior prediction in such smart home environments, a lightweight CNN (Convolutional Neural Network) and RNN (Recurrent Neural Network) architecture is combined.

[0040] In the present invention, all data processing is performed locally on the device by federated learning technology, so there is no need to upload data to the server (30), and the risk of personal information infringement can be greatly reduced.

[0041] The local model according to the federated learning technology of such smart home devices (10) is composed of multiple layers. FIG. 3 is a diagram showing the structure of the local model of a smart home device in an information protection system of a federated learning-based smart home according to the present invention.

[0042] Referring to FIG. 3, the local model consists of an input layer (11a), a convolutional layer (11b), a recurrent layer (11c), a fully connected layer (11d), and an output layer (11e).

[0043] The input layer (11a) processes sensor data detected from each sensor (15).

[0044] The input layer (11a) standardizes the data into a normalized range to facilitate effective model training.

[0045] The convolutional layer (11b) extracts local features from sensor data using a convolutional kernel.

[0046] At this time, the convolutional layer (11b) can capture subtle variation of the data.

[0047] The circular layer (11c) captures the temporal dependency of the sensor data.

[0048] In the present invention, the recurring layer (11c) uses a single-layer LSTM (Long Short-Term Memory) network with 32 hidden units to capture the time dependency of smart home data.

[0049] The gate mechanism of the LSTM can effectively maintain the user's smart home device (10) usage patterns, such as specific daytime and nighttime behaviors.

[0050] The fully connected layer (11d) abstracts the extracted local features and performs classification or regression work.

[0051] In the present invention, the fully connected layer (11d) is composed of two layers with 64 neurons and 32 neurons, respectively, using a ReLU activation function.

[0052] These layers further abstract the extracted features and map them to the target space for classification or regression tasks.

[0053] The output layer (11e) aggregates the local model and sends it to the server (30).

[0054] In the present invention, the output layer (11e) converts the output into a probability by utilizing a sigmoid activation function for binary classification tasks and utilizes a softmax activation function for multi-class classification tasks to predict various states of operation of the smart home device (10).

[0056] The user terminal (20) transmits and receives data for the control of the smart home device (10) and is connected via wireless communication.

[0057] The user terminal (20) may be an electronic device such as a smartphone or PC used by the user.

[0059] The server (30) receives data through the smart home device (10) and the user terminal (20) and controls the smart home device (10) with optimized data.

[0060] The server (30) according to the present invention generates a global model based on differential privacy technology and an aggregation algorithm for encrypted data received from a smart home device (10).

[0061] Then, the server (30) transmits the generated global model to the smart home device (10) so that the local model of the smart home device (10) is updated.

[0062] That is, the server (30) adds a security layer by using differential privacy technology that adds global model updates or noise to the federated learning framework.

[0063] In this way, the server (30) can ensure that, through differential privacy technology, even if an attacker accesses aggregated data through global model updates or the introduction of noise, sensitive personal information about individual users cannot be inferred.

[0064] In other words, differential privacy technology can effectively obscure specific individual data by introducing a certain amount of random noise into the data distribution, thereby enhancing personal information protection while maintaining the overall distribution trend, which ensures data usability.

[0066] Hereinafter, we examine an information protection method using an information protection system of a federated learning-based smart home according to the present invention. Figure 4 is a diagram illustrating an information protection method of a federated learning-based smart home according to the present invention.

[0067] Referring to FIG. 4, data is first collected from each sensor (15) of the smart home device (10) (S1).

[0068] The smart home device (10) interacts with the user through each sensor (15) to capture user behavior data in real time and store it in an encrypted format.

[0069] After that, the smart home device (10) locally trains the collected data based on federated learning technology (S3).

[0070] At this time, the smart home device (10) uses the collected data to train a model and updates the model parameters, thereby keeping the data locally and protecting the user's personal information.

[0071] Next, the smart home device (10) encrypts the local data generated by local training and transmits it to the server (30) (S5).

[0072] That is, the smart home device (10) applies differential privacy to the updated model parameters and uploads them to the server (30) via a secure network protocol.

[0073] After that, the server (30) aggregates encrypted data based on differential privacy technology to create a global model (S7).

[0074] That is, when the server (30) receives encrypted model parameters from the smart home device (10), it generates a global model using a federated averaging algorithm.

[0075] At this time, the server (30) can further protect personal information by directly aggregating encrypted data without decoding or storing the data of individual smart home devices (10).

[0076] This global model is then redistributed to smart home devices (10) through a secure channel to update the local model.

[0077] And the server (30) analyzes the balance between the model's performance and privacy protection, and adjusts the privacy budget accordingly.

[0078] Next, the server (30) transmits the generated global model to the smart home device (10) (S9), and the smart home device (10) updates the model with the received global model (S11) and feeds back the data (S13).

[0079] When the smart home device (10) receives a model from the server (30), it decodes and loads the new global model.

[0080] Through this, the smart home device (10) can provide optimized services, and new user data is generated during use and becomes input data for the next training cycle, making continuous model iteration and optimization easier.

[0082] The foregoing description is merely an illustrative explanation of the technical concept of the present invention, and those skilled in the art to which the present invention pertains will be able to make various modifications, changes, and substitutions within the scope of the essential characteristics of the present invention. Furthermore, as described above, the embodiments disclosed in the present invention and the accompanying drawings are intended to explain, not limit, the technical concept of the present invention, and the scope of the technical concept of the present invention is not limited by such embodiments and accompanying drawings. The scope of protection of the present invention shall be interpreted by the claims below, and all technical concepts within an equivalent scope shall be interpreted as being included within the scope of rights of the present invention. Explanation of the symbols

[0083] 10; Smart home devices 11; Associated learning technology 11a; Input layer 11b; Convolutional layer 11c; Circular hierarchy 11d; Fully connected layer 11e; output layer 15; Sensors, home appliances 20; User terminal 30; server 31; Differential privacy technology 35; Storage section

Claims

Claim 1 A smart home information protection system for protecting personal information in a smart home system comprising: a smart home device including a light sensor, an air quality sensor, a security sensor, a motion detection sensor, and a home appliance provided within the home; a user terminal communicating for the control of the smart home device; and a server receiving data through the smart home device and the user terminal and controlling the smart home device with optimized data; wherein the smart home device is operated by locally training collected data based on federated learning technology to generate a local model, transmitting encrypted data to the server, and updating the model parameters of a global model received from the server; and the server generates a global model based on the encrypted data received from the smart home device using differential privacy technology and an aggregation algorithm, and transmits the generated global model to the smart home device so that the local model is updated. Claim 2 An information protection system for a federated learning-based smart home according to claim 1, wherein the local model according to the federated learning technology of the smart home device comprises an input layer for processing sensor data detected by each sensor, a convolutional layer for extracting local features from the sensor data using a convolutional kernel, a recurrent layer for capturing the temporal dependency of the sensor data, a fully connected layer for abstracting the extracted local features and performing classification or regression work, and an output layer for aggregating and transmitting the local model to the server. Claim 3 An information protection system for a federated learning-based smart home according to claim 1, wherein the server adds a security layer using the differential privacy technology that adds the global model update or noise to the federated learning framework. Claim 4 A method for protecting information using an information protection system of a federated learning-based smart home as described in claim 1 above, comprising: a step of collecting data from each sensor of the smart home device; a step of locally training the collected data based on the federated learning technology; a step of encrypting the local data generated by the local training and transmitting it to the server; a step of aggregating the encrypted data based on differential privacy technology at the server and generating a global model; and a step of transmitting the generated global model to the smart home device and feeding back data using the received global model at the smart home device.

Images