Intelligent power grid terminal multi-factor identity authentication method and system based on edge calculation

Abstract

The invention discloses a smart grid terminal multi-factor identity authentication method and system based on edge computing, and the method employs an edge computing technology to deploy a pre-trained image recognition neural network model in an edge computing device, and the edge computing device stores a secret key and identity information of a user; an image recognition neural network model is adopted to detect a video stream acquired by the edge computing device, a template method is adopted to recognize a key of a user, then a feature point matching algorithm is adopted to perform attitude matching on the recognized key and a stored key, and identity authentication is performed on the user according to a matching result; according to the method, the system security and the authentication timeliness are optimized, the structure of a multi-factor authentication mechanism is improved, the possibility of leakage of personal privacy data of the user is greatly reduced, and the method is simple to implement and high in practicability.

Description

Technical Field

[0001] The present invention relates to the technical field of smart grid identity authentication, and specifically to a smart grid terminal multi-factor identity authentication method and system based on edge computing. Background Art

[0002] The power information system refers to a comprehensive system based on computer technology and communication technology for collecting, transmitting, storing, processing and analyzing power system-related data. The power industry's demand for network security continues to grow. In recent years, due to the increasingly complex structure of the power system, the expanding network boundaries, and the dynamic changes in the network security situation, relevant laws and regulations and network security level protection standards have been successively issued, which also put forward higher requirements for the dynamic authentication of the power information system.

[0003] At present, the power system adopts a node-based multi-level management framework to manage each power system terminal. In this structure, the commands received by the power system terminal and the requests generated by it must pass through nodes at all levels to transmit data and instructions layer by layer. In the management structure of the power system terminal, the number of terminal devices is increasing, and the data transmission channel in the network cannot meet the high-time information transmission and frequent data interaction between the terminal and the upper node server.

[0004] Based on the above problems, the use of edge computing architecture, deep learning network, computer vision technology and other technologies to construct a new smart grid identity authentication system has attracted more and more attention. However, in the existing traditional multi-factor authentication method, there is a problem of simple accumulation of multiple authentication factors, and the high security requirements of multi-factor authentication are not met. Some multi-factor identity authentication mechanisms have high requirements on the computing power of smart terminal devices and cannot adapt to different types of terminals. Summary of the invention

[0005] In view of the problems existing in the prior art, the present invention provides a multi-factor identity authentication method for smart grid terminals based on edge computing, which combines the concepts of graphical passwords and token-based identity authentication into a single-factor authentication (2FA) solution, and transfers the computing tasks generated in the identity authentication process from the power system main node to the smart terminal device, avoiding frequent data transmission between the terminal device and the upper node.

[0006] The present invention is achieved through the following technical solutions:

[0007] A multi-factor identity authentication method for a smart grid terminal based on edge computing comprises the following steps:

[0008] Use edge computing technology to deploy the pre-trained image recognition neural network model on the edge computing device, which stores the user's secret key and identity information;

[0009] An image recognition neural network model is used to detect the video stream obtained by the edge computing device, and a template method is used to identify the user's secret key. Then, a feature point matching algorithm is used to perform posture matching between the identified secret key and the stored secret key, and the user is authenticated based on the matching result.

[0010] Preferably, the image recognition neural network model is a YOLOv5 network model, which is trained using a key data set, wherein the key data set includes a template picture set and an interference picture set.

[0011] Preferably, the YOLOv5 network model includes:

[0012] Backbone module, used to aggregate and form image features at different image granularities;

[0013] Neck module, which is used to mix and combine image features and pass them to the prediction layer;

[0014] The Head module is used to generate bounding boxes for image features and distinguish categories.

[0015] Preferably, the training method of the YOLOv5 network model includes:

[0016] The weight parameters of the network model are adjusted through the loss function. The classification loss and confidence loss of the YOLOv5 network model are calculated using the BCEWithLogitsLoss loss function, and the bounding box loss is calculated using the SIOULoss loss function.

[0017] Preferably, the BCEWithLogitsLoss is a BCE Loss basis plus a sigmoid function;

[0018] The BCE Loss calculation formula is:

[0019]

[0020] The calculation formula of Sigmoid function is:

[0021]

[0022] Preferably, the SIOU Loss loss function includes angle loss, distance loss, shape loss and IoU loss; the expression of the SIOU Loss loss function is as follows:

[0023]

[0024] Among them, Δ is the distance loss, Ω is the shape loss, and IoU is the intersection over union loss.

[0025] Preferably, the edge computing device is an image sensing module, which is used to obtain a user's video stream, input the video stream into a pre-trained image recognition neural network model, and perform user identity authentication based on the recognition result.

[0026] Preferably, the image recognition neural network model detects the video stream as follows:

[0027] Perform frame detection and key matching on the video stream;

[0028] The image recognition neural network model obtains the secret key in the image frame and compares the recognition key with the secret key in the template. When the secret keys are consistent, the first factor authentication is completed;

[0029] The SIFT feature point matching algorithm is used to match the posture of the secret key with the posture of the secret key in the template. When the similarity reaches the preset level, the identity authentication is completed.

[0030] Preferably, if any link in the matching process fails, the exception handling path is entered and re-authentication is returned. If the authentication fails more than a set number of times, the edge computing device automatically locks the identity authentication system.

[0031] A multi-factor identity authentication system for smart grid terminals based on edge computing, comprising:

[0032] The initialization module uses edge computing technology to deploy the pre-trained image recognition neural network model to the edge computing device, which stores the user's secret key and identity information;

[0033] The authentication module uses an image recognition neural network model to detect the video stream obtained by the edge computing device, uses a template method to identify the user's secret key, and then uses a feature point matching algorithm to perform posture matching between the identified secret key and the stored secret key, and authenticates the user based on the matching results.

[0034] Compared with the prior art, the present invention has the following beneficial technical effects:

[0035] The multi-factor identity authentication method for smart grid terminals based on edge computing provided by the present invention adopts an edge computing architecture to transfer the computing tasks generated in the identity authentication process from the main node of the power system to the smart terminal device, thereby avoiding frequent data transmission between the terminal device and the upper node. Secondly, unlike general biometric identification technology, the present invention allows users to change the physical factors of identity authentication, just like they change the accessories they wear or carry. The present invention reduces the risk of lifelong impact on victims by opponents who obtain identity authentication secrets, thereby alleviating the need for biometric traceability and revocation. In addition, the present invention does not require common sensors or hardware, only a camera. The present invention is deployed in the identity authentication system of the terminal in a centralized training and offline deployment manner. The terminal device only needs a low-cost computing unit or multiple terminals share a computing power module. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a flow chart of a multi-factor identity authentication method for a smart grid terminal based on edge computing provided by an embodiment of the present invention;

[0037] Figure 2 It is the overall structure diagram of the YOLOv5 network model of the present invention.

[0038] Figure 3 It is a structural diagram of the identity authentication mechanism according to an embodiment of the present invention.

[0039] Figure 4 It is a schematic diagram of matching degree under different postures provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0040] The present invention will be further described in detail below in conjunction with the accompanying drawings, which are intended to explain the present invention rather than to limit it.

[0041] Example 1

[0042] See also Figure 1 , a multi-factor identity authentication method for smart grid terminals based on edge computing, the method is divided into four specific implementation processes, including model image data set processing process, model training process, model deployment process and identity authentication process, including the following steps:

[0043] Step 1: Collect the original authentication "key" item photo as the original image, pre-process the original image, and generate a data set;

[0044] The original authentication "key" item photo is a small part of the item worn by the user (e.g., a shirt, a shoe) or any available item around the user (e.g., a USB flash drive, a watch, etc.). The image is dynamically generated based on the physical world around the user. Since the user can freely present candidate passwords through photos of the physical world taken under different lighting, background and angle conditions, accurate matching of the items must be achieved.

[0045] The dataset is divided into a training set and a validation set in a ratio of 7:3. The training set is also divided into a template image set (correct answer) and an interference image set (authentication information close to the correct answer).

[0046] This embodiment is described using "U disk" as the authentication information template, so the template image set includes "U disk" images taken from various angles, and the interference image set includes several types of images of objects that are similar in shape and color to the "U disk".

[0047] Step 2: Input the training set into the YOLOv5 network model for training, perform “object posture” matching through the SIFT feature point matching algorithm, and obtain the trained YOLOv5 network model;

[0048] The YOLOv5 network model includes a backbone network, a neck network and a detection head network.

[0049] Backbone: The backbone network of the YOLOv5 network model, a convolutional neural network that aggregates and forms image features at different image granularities.

[0050] Neck: YOLOv5s network model neck network, a series of network layers that mix and combine image features and pass image features to the prediction layer.

[0051] Head: The YOLOv5s network model detects the head, generates bounding boxes and distinguishes categories.

[0052] Backbone is used to extract image features from images, neck is used to aggregate image features, and head is used for classification and positioning.

[0053] During model training, the weight parameters are adjusted through the loss function. The classification loss and confidence loss of the YOLOv5 network model are calculated using the BCEWithLogitsLoss loss function, and the bounding box loss is calculated using the SIOULoss loss function. BCEWithLogitsLoss is based on BCE Loss plus the sigmoid function. The BCE Loss calculation formula is:

[0054]

[0055] The calculation formula of Sigmoid function is:

[0056]

[0057] SIOU Loss consists of four parts: angle cost, distance cost, shape cost and IoU cost. Its calculation formula is:

[0058]

[0059] Among them, Δ is the distance loss, Ω is the shape loss, and IoU is the intersection over union loss. This loss function redefines the distance loss while taking into account the angle loss to make a more comprehensive evaluation of the position, size, and orientation of the object.

[0060] Step 4: Use the test set to verify the trained YOLOv5 network model, and deploy the verified YOLOv5 network model on the power system terminal device through the edge computing device;

[0061] Before deploying and applying the model to the intelligent terminal of the power system, the YOLO network model needs to be trained and data verified in advance.

[0062] The YOLO network model is verified using precision and recall.

[0063] The calculation formulas for Precision and Recall are as follows:

[0064]

[0065] TP: The label value is True and the model prediction is Positive;

[0066] FN: The label value is False and the model prediction is Negative;

[0067] FP: The label value is False and the model prediction is Positive;

[0068] TN: The label value is True and the model prediction is Negative.

[0069] After verification, the network model is deployed in an edge computing device, which is an OV5647 image sensing module (model: RPi Camera (M)) that can connect to multiple versions of Raspberry Pi. The module is used as the camera of the image sensor of the identity authentication system, and the focal length and size of the captured image make it similar to the image captured by a general surveillance camera.

[0070] The edge computing architecture migrates computing facilities to the terminal.

[0071] The edge computing architecture consists of three parts: the power information system sensor layer, the edge computing layer, and the cloud computing layer. The sensor layer continuously captures real-time images of the key "item" through surveillance cameras fixed to different power equipment in the power system. The neural network model deployed on the Raspberry Pi chip detects the type and posture of the object in the image and preliminarily determines the correctness of the key. When a difference is detected between the key "item" and the template, the authentication link fails, and the system immediately sends the authentication failure event to the central control terminal through the edge layer for early warning.

[0072] Next, the cloud layer provides reliable central storage, partially processed sensor quality is optimal, and image data is streamed between multiple edge computing devices with shared network bandwidth. The cloud layer then performs the next level of processing on the sensor and image data to analyze whether there is a mistake in the judgment of this authentication by referring to the template image set available in the repository, thereby monitoring the possibility of attacks on the power system.

[0073] Step 5: Use the YOLOv5 network model to detect and match the key in the video stream, and complete identity authentication based on the matching results.

[0074] Use the camera to collect images to form a video stream, input the video stream into the network model and match it frame by frame; if the match is successful, the authentication is completed. If any link in the matching fails, it enters the exception handling path and returns to re-authentication. If the authentication fails more than three times, the device automatically locks the identity authentication system.

[0075] For a detailed description of the identity authentication process, please refer to Figure 3 FIG. 1 is a schematic diagram of an identity authentication process provided by an embodiment of the present invention. Figure 3 As shown, in this embodiment, the process of the identity authentication model includes the following steps:

[0076] S5.1. “Possession-based” certification:

[0077] The multi-factor identity authentication mechanism proposed in the present invention first enters the normal channel, and the YOLO network model detects and identifies the types of "items" in the video stream input by the user, and authenticates the first factor of the identity information - "based on possession".

[0078] In this step, it is determined whether the "item" in the picture input by the user is consistent with the type of "item" in the template. If the calculation result output by the model shows that the "item" input by the user is consistent with the template, the first factor authentication of the identity information "based on possession" is successful, and the second factor authentication of the identity information is entered. If the calculation result output by the model shows that no "item" is detected or the type of "item" is different from the template, then this authentication link fails.

[0079] S5.2, "Knowledge-based" certification:

[0080] The SIFT feature point matching algorithm performs a knowledge-based authentication of the second factor of the identity information of the "object" in the picture input by the user. The focus is to measure the similarity index of the "object posture" by the matching degree of the SIFT feature points in the template picture and the picture information to be detected.

[0081] SIFT feature points are special points that remain unchanged under rotation, scaling, brightness changes, and maintain a certain degree of stability under perspective changes, affine transformations, and noise. The authentication mechanism sets a threshold for the matching degree of feature points between the template image and the image to be detected to distinguish whether the "object posture" is successfully authenticated.

[0082] In the authentication mechanism proposed in this embodiment, when any link in the normal path fails to authenticate, the abnormal processing path will be entered. The abnormal processing path leads to two results. If any link fails to authenticate within three times, it will return to the initial stage for secondary authentication. If any link fails to authenticate more than three times, an alarm will be triggered and the device will lock the identity authentication mode. The abnormal processing path allows users to have a certain number of authentication information input deviations when interfered by environmental factors, and also effectively defends against attackers' probing attacks.

[0083] It is worth noting that the authentication time of the multi-factor identity authentication system for smart grid terminals based on edge computing in the embodiment of the present invention is similar to that of the PIN text password, and it only takes 2-5 seconds to complete the two-step picture information authentication. When the authentication timeliness is similar, the security of the embodiment of the present invention is much greater than that of the PIN text password. For the SMS verification code login mechanism, since the waiting time for SMS is restricted by multiple parties such as communication operators, the authentication timeliness of the authentication mechanism in the embodiment of the present invention has obvious advantages, and the time consumed by SMS verification is 5.1 times that of the authentication mechanism in the embodiment of the present invention.

[0084] Example 2

[0085] A multi-factor identity authentication system for smart grid terminals based on edge computing, comprising:

[0086] The initialization module uses the edge computing algorithm to deploy the pre-trained image recognition neural network model to the edge computing device, which stores the user's secret key and identity information;

[0087] The authentication module uses an image recognition neural network model to detect the video stream obtained by the edge computing device, uses a template method to identify the user's secret key, and then uses a feature point matching algorithm to perform posture matching between the identified secret key and the stored secret key, and authenticates the user based on the matching results.

[0088] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants are intended to cover non-exclusive inclusion, so that an article or device including a series of elements includes not only those elements, but also other elements that are not explicitly listed. In the absence of more restrictions, the elements defined by the statement "including one..." do not exclude the existence of other identical elements in the article or device including the elements.

[0089] The above contents are only for explaining the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. Any changes made on the basis of the technical solution in accordance with the technical idea proposed by the present invention shall fall within the protection scope of the claims of the present invention.

Claims

1. A multi-factor identity authentication method for smart grid terminals based on edge computing, characterized in that: The following steps are involved: Use edge computing technology to deploy the pre-trained image recognition neural network model on the edge computing device, which stores the user's secret key and identity information; An image recognition neural network model is used to detect the video stream obtained by the edge computing device, and a template method is used to identify the user's secret key. Then, a feature point matching algorithm is used to perform posture matching between the identified secret key and the stored secret key, and the user is authenticated based on the matching result.

2. According to claim 1, a multi-factor identity authentication method for smart grid terminals based on edge computing is characterized in that: The image recognition neural network model is a YOLOv5 network model, which is trained using a secret key data set, wherein the secret key data set includes a template picture set and an interference picture set.

3. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 2, characterized in that: The YOLOv5 network model includes: Backbone module, used to aggregate and form image features at different image granularities; Neck module, which is used to mix and combine image features and pass them to the prediction layer; The Head module is used to generate bounding boxes for image features and distinguish categories.

4. According to claim 2, a multi-factor identity authentication method for smart grid terminals based on edge computing is characterized in that: The training method of the YOLOv5 network model includes: The weight parameters of the network model are adjusted through the loss function. The classification loss and confidence loss of the YOLOv5 network model are calculated using the BCEWithLogitsLoss loss function, and the bounding box loss is calculated using the SIOULoss loss function.

5. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 4, characterized in that: The BCEWithLogitsLoss is the BCE Loss basis plus the sigmoid function; The BCE Loss calculation formula is: The calculation formula of Sigmoid function is:

6. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 4, characterized in that: The SIOU Loss loss function includes angle loss, distance loss, shape loss and IoU loss; the expression of SIOU Loss loss function is as follows: Among them, Δ is the distance loss, Ω is the shape loss, and IoU is the intersection over union loss.

7. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 1, characterized in that: The edge computing device is an image sensing module, which is used to obtain the user's video stream, input the video stream into a pre-trained image recognition neural network model, and perform user identity authentication based on the recognition result.

8. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 1, characterized in that: The image recognition neural network model detects the video stream as follows: Perform frame detection and key matching on the video stream; The image recognition neural network model obtains the secret key in the image frame and compares the recognition key with the secret key in the template. When the secret keys are consistent, the first factor authentication is completed; The SIFT feature point matching algorithm is used to match the posture of the secret key with the posture of the secret key in the template. When the similarity reaches the preset level, the identity authentication is completed.

9. A multi-factor identity authentication method for smart grid terminals based on edge computing according to claim 8, characterized in that: If any link in the matching process fails, the system will enter the exception handling path and return to re-authentication. If the authentication fails more than the set number of times, the edge computing device will automatically lock the identity authentication system.

10. A system for executing the multi-factor identity authentication method for smart grid terminals based on edge computing as described in any one of claims 1 to 9, characterized in that: include: The initialization module uses edge computing technology to deploy the pre-trained image recognition neural network model to the edge computing device, which stores the user's secret key and identity information; The authentication module uses an image recognition neural network model to detect the video stream obtained by the edge computing device, uses a template method to identify the user's secret key, and then uses a feature point matching algorithm to perform posture matching between the identified secret key and the stored secret key, and authenticates the user based on the matching results.

Images