Screen anti-stealing method and system based on permission control

Abstract

This application relates to the field of screen privacy and access control technology, and discloses a screen anti-spy camera method and system based on access control. The method includes: first, obtaining the access status of the screen user and mapping it to access level parameters; then, monitoring the surrounding area of ​​the screen to identify shooting behavior and extracting behavioral features; subsequently, inputting the access-related information and behavioral features into a deep learning risk assessment model to obtain a spy camera risk assessment value. Based on the access status, whether the shooting behavior is valid, and the risk assessment value, it is determined whether to issue a power-on command to the anti-spy camera screen. After power-on, the screen changes from a high-transparency state to a uniformly pixelated state, and the pixelation density is determined by combining the access level parameters and the risk assessment value. This invention achieves differentiated and refined screen anti-spy camera protection, balancing information security and ease of use, and is suitable for highly sensitive information scenarios.

Description

Technical Field

[0001] This application relates to the field of screen privacy and access control technology, specifically a screen anti-spy camera method and system based on access control. Background Technology

[0002] In scenarios with high sensitivity to screen information security, such as confidential meetings, financial transaction demonstrations, and internal government business presentations, screen anti-spy camera protection is a crucial step in ensuring information is not leaked. Existing screen anti-spy camera technologies mostly use cameras to capture images of the screen's perimeter, relying on basic motion recognition models to identify filming behavior and trigger audio-visual alerts or fixed protective actions.

[0003] However, these technical solutions have significant shortcomings. On the one hand, the protection triggering logic is too simplistic, relying solely on whether a shooting action is detected to activate protection, without fully considering the different permission statuses of the screen user. This makes it easy for legitimate high-privilege users to be interfered with by accidental triggering of protection when performing critical operations, and the security strategy lacks fine-grained control. On the other hand, the protection execution method lacks a hierarchical design. After powering on, privacy screens often use a uniform blocking or masking mode, failing to dynamically adjust the protection strength based on the content sensitivity level and permission level, often resulting in over-protection or under-protection.

[0004] Furthermore, in existing technologies, access control is only used during the login or access phase and does not extend throughout the entire protection process. Moreover, the shooting behavior recognition mostly outputs binary judgment results, which is difficult to deal with grayscale scenes between shooting and non-shooting. Its adaptability to complex environments is limited, and it cannot meet the needs of accurate and flexible anti-spy camera protection in highly sensitive scenarios. Summary of the Invention

[0005] The purpose of this application is to provide a screen anti-spy camera method and system based on access control to solve the problems mentioned in the background art.

[0006] According to the first aspect of this application, a screen anti-spy photography method based on access control is provided, comprising the following steps:

[0007] Obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter;

[0008] The system monitors the area surrounding the current screen, identifies whether a shooting action is valid, and extracts behavioral features related to the shooting action.

[0009] The permission status, the permission level parameter, and the behavioral characteristics are input into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk.

[0010] Based on the permission status, whether the shooting behavior is valid, and the risk assessment value, determine whether the conditions for issuing a power-on command to the privacy screen are met;

[0011] If the conditions for issuing a power-on command are met, a power-on command is issued to the privacy screen, causing the privacy screen to change from a high-transparency state to a uniform coding state.

[0012] The coding density under the uniform coding state is determined based on the permission level parameter and the risk assessment value.

[0013] Preferably, the permission status includes at least administrator permission status, ordinary user permission status, and public display status;

[0014] Based on the permission status, whether the shooting behavior is valid, and the risk assessment value, the conditions for determining whether a power-on command to be issued to the privacy screen are met include:

[0015] When the permission status is administrator permission status, if the shooting behavior is established, the risk assessment value is greater than or equal to the first risk threshold, and the duration of the shooting behavior exceeds the first time threshold, then it is determined that the conditions for issuing the power-on command are met.

[0016] When the permission status is that of a normal user, if the shooting behavior is successful and the risk assessment value is greater than or equal to the second risk threshold, then it is determined that the conditions for issuing a power-on command are met, and the second risk threshold is less than the first risk threshold.

[0017] Preferably, the input to the deep learning risk assessment model further includes the current state of the privacy screen and the historical trigger records of shooting behavior;

[0018] The behavioral features include intermediate features generated during the shooting behavior recognition process, confidence information, and shooting behavior stability features extracted from a continuous time window.

[0019] Preferably, the deep learning risk assessment model includes a convolutional neural network layer and a temporal modeling layer;

[0020] The convolutional neural network layer is used to process the spatial features in the behavioral features;

[0021] The time-series modeling layer is used to process the time-series features in the behavioral characteristics and output the risk assessment value.

[0022] Preferably, the coding density under the uniform coding state is determined specifically based on the permission level parameter and the risk assessment value as follows:

[0023] Get the base CAPTCHA density parameter corresponding to the current permission level parameter;

[0024] Based on the preset adjustment formula, the final coding density is calculated according to the basic coding density parameter and the risk assessment value.

[0025] The adjustment formula is: coding density = basic coding density parameter × [1 + adjustment coefficient × (risk assessment value - risk benchmark value)].

[0026] Preferably, after the privacy screen is in a powered-on, uniformly coded state, the method further includes dynamic adjustment of the protection:

[0027] Continuously monitor the shooting behavior status and the risk assessment value;

[0028] If the permission status changes, the masking density is re-determined and the masking intensity of the privacy screen is adjusted based on the permission level parameter corresponding to the changed permission status and the current risk assessment value.

[0029] Preferably, after the privacy screen is in a powered, uniformly coded state, the method further includes protection restoration:

[0030] If the detected shooting behavior disappears, and the risk assessment value drops below the preset recovery threshold and continues to exceed the recovery time threshold, then the privacy screen is powered off to restore it to a high-transparency state.

[0031] A second aspect of this application also provides a screen anti-spy camera system based on access control, comprising:

[0032] It includes a permission management module, a behavior recognition module, a risk assessment module, and a protection and control module;

[0033] The permission management module is used to obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter;

[0034] The behavior recognition module is used to monitor the area around the current screen, identify whether the shooting behavior is valid, and extract the behavioral features related to the shooting behavior.

[0035] The risk assessment module is used to input the permission status, the permission level parameter, and the behavioral characteristics into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk.

[0036] The protection control module is communicatively connected to the permission management module, the behavior recognition module, the risk assessment module, and a privacy screen, and is used to determine whether the conditions for issuing a power-on command to the privacy screen are met based on the permission status, whether the shooting behavior is valid, and the risk assessment value.

[0037] If the conditions for issuing a power-on command are met, a power-on command is issued to the privacy screen, causing the privacy screen to change from a high-transparency state to a uniform coding state.

[0038] The coding density under the uniform coding state is determined based on the permission level parameter and the risk assessment value.

[0039] Preferably, the permission status includes at least administrator permission status, ordinary user permission status, and public display status;

[0040] The protection control module is specifically used to: when the permission status is administrator permission status, if the shooting behavior is established, and the risk assessment value is greater than or equal to the first risk threshold, and the duration of the shooting behavior exceeds the first time threshold, then it is determined that the conditions for issuing a power-on command are met.

[0041] When the permission status is that of a normal user, if the shooting behavior is successful and the risk assessment value is greater than or equal to the second risk threshold, then it is determined that the conditions for issuing a power-on command are met, and the second risk threshold is less than the first risk threshold.

[0042] Preferably, the deep learning risk assessment model includes a convolutional neural network layer and a temporal modeling layer;

[0043] The convolutional neural network layer is used to process the spatial features in the behavioral features;

[0044] The time-series modeling layer is used to process the time-series features in the behavioral characteristics and output the risk assessment value.

[0045] This application constructs a complete, sophisticated, and adaptive screen anti-spy camera protection system through a series of steps, including permission status acquisition and level mapping, shooting behavior monitoring and feature extraction, deep learning risk assessment, differentiated power-on decision-making, dynamic masking density adjustment, and linked protection. This system integrates permission control throughout the entire protection process and, combined with a deep learning model, achieves accurate risk assessment, solving problems in existing technologies such as single protection trigger logic, lack of hierarchical control in protection execution, and mismatch between permissions and risks. The intelligent power-on anti-spy camera screen switches from a high-transparency standby state to uniform masking when powered on, forming multiple protective barriers in conjunction with audible and visual alarms and signal shielding, ensuring comprehensive protection and rapid, efficient response. Simultaneously, through a dynamic adjustment and recovery mechanism, it minimizes interference with normal operation while ensuring information security, balancing security and practicality. Attached Figure Description

[0046] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0047] Other features, objects, and advantages of this application will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings:

[0048] Figure 1 A schematic diagram of a screen anti-spy photography method based on access control provided in an embodiment of this application;

[0049] Figure 2 This is a schematic diagram of the risk assessment model structure provided in the embodiments of this application;

[0050] Figure 3 A schematic diagram of the coding density determination and dynamic protection adjustment process provided in the embodiments of this application;

[0051] Figure 4 This is a schematic diagram of a screen anti-spy camera system based on access control, provided as an embodiment of this application. Detailed Implementation

[0052] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0053] It should be noted that all user information (including but not limited to user device information, user personal information, object information corresponding to device usage data, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, device usage data, etc.) involved in all embodiments of this application are information and data authorized by the user or fully authorized by all parties.

[0054] This method is applicable to scenarios with high sensitivity to screen information security, such as confidential meetings, financial transaction demonstrations, internal government business presentations, and core enterprise data operations. It can effectively prevent various types of unauthorized filming of screens and ensure that sensitive information is not leaked. Implementing this method typically requires meeting the following basic conditions: The system needs to deploy hardware devices with video capture and behavior recognition capabilities, including high-definition analysis cameras, computing servers, intelligent power-on anti-spy screens, sound and light alarms, signal jammers, etc.; At the software level, it needs to be equipped with a permission management module, a shooting behavior recognition model, a deep learning risk assessment model, and corresponding underlying algorithm libraries, such as computer vision processing libraries, deep learning acceleration libraries, and inference engines, to ensure real-time acquisition of user permission status, accurate identification of shooting behavior, risk assessment, and execution of protective operations. The entire process requires no new hardware and does not change the working mechanism of the original core equipment.

[0055] The implementation process of the screen anti-spy camera method based on access control described in this application will be described in detail below with reference to specific embodiments. It should be noted that this embodiment is only used to explain this application and is not intended to limit the scope of protection of this application. Conventional adjustments or substitutions of each step by those skilled in the art without departing from the concept of this application should be included in the protection scope of this application.

[0056] like Figure 1 As shown in the diagram, this application discloses a screen anti-spy photography method based on access control, which includes the following method steps:

[0057] S1, obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter;

[0058] S2, monitor the area around the current screen, identify whether the shooting behavior is valid, and extract behavioral features related to the shooting behavior;

[0059] S3, input the permission status, the permission level parameter and the behavioral characteristics into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk;

[0060] S4. Based on the permission status, whether the shooting behavior is established, and the risk assessment value, determine whether the conditions for sending a power-on command to the privacy screen are met.

[0061] S5. If the conditions for issuing a power-on command are met, then issue a power-on command to the privacy screen to change the privacy screen from a high-transparency state to a uniform coding state.

[0062] S6. Determine the coding density under the uniform coding state based on the permission level parameter and the risk assessment value.

[0063] In some embodiments, permission status acquisition is mapped to permission levels. After the system starts and completes initialization, the permission management module acquires the permission status of the current screen user in real time. This permission status, as a continuously effective runtime parameter, is dynamically updated and stored in the system cache, providing a basis for subsequent protection decisions. Permission status includes at least administrator permission status, ordinary user permission status, and public display status. The administrator permission status corresponds to user scenarios with full system operation permissions. These users typically perform critical business operations and have high requirements for business continuity. The ordinary user permission status corresponds to user scenarios with only basic operation and viewing permissions. The sensitivity of their operations is relatively controllable. The public display status corresponds to scenarios where no specific user is logged in and the screen displays information to the public. The sensitivity of the displayed content can be preset according to the actual situation.

[0064] After acquiring the permission status, the system maps different permission statuses to corresponding permission level parameters according to preset mapping rules. These mapping rules can be configured and adjusted by the administrator based on the security requirements of the actual application scenario. For example, permission levels can be divided into three levels: administrator permission status is mapped to level one, ordinary user permission status to level two, and public display status to level three. Each permission level corresponds to a unique identifier parameter, stored in the system's permission configuration database for quick retrieval during subsequent protection execution. This process solves the problem of existing technologies failing to consider permission differences, leading to a single protection strategy. By associating permission status with levels, it provides a foundation for differentiated protection, ensuring that protection strategies for users with different permissions are more closely aligned with actual needs.

[0065] In some embodiments, step S2 involves monitoring and extracting behavioral features during shooting. Utilizing a dual-camera setup with high-definition analysis cameras, full coverage monitoring of a 180° area around the protected screen can be achieved. The cameras' high pixel count and autofocus capabilities accurately capture dynamic targets within a set distance range, ensuring no blind spots. The cameras continuously capture real-time images of the area surrounding the screen. The captured images are transmitted to the system's image processing module, where they undergo preprocessing using a computer vision processing library. This preprocessing includes image denoising, image enhancement, and size normalization to eliminate the impact of ambient light, shooting angle, and other factors on subsequent recognition accuracy, thereby improving the accuracy of shooting behavior recognition.

[0066] The pre-processed footage is transmitted to the inference engine, where a pre-trained shooting behavior recognition model identifies action features in the footage. The core features for valid shooting behaviors include "the mobile terminal is raised and pointed at the screen" and "the lens maintains a stable alignment with the screen." When the model recognizes these features, it determines that the shooting behavior is valid and outputs a binary judgment result indicating whether the shooting behavior is valid. During the recognition process, the system not only outputs this binary judgment result but also retains intermediate features and confidence information. Intermediate features include the mobile terminal's contour features, motion trajectory features, and lens area features. Confidence information characterizes the model's trustworthiness of the current recognition result; higher confidence indicates a more reliable recognition result.

[0067] In one embodiment, the pre-trained shooting behavior recognition model adopts a hybrid architecture of "CNN and LSTM" and inputs a pre-processed fixed-size RGB image tensor. Spatial features such as the mobile terminal contour and lens area are extracted through three convolutional and pooling layers, and a global spatial feature vector is obtained through global average pooling. The spatial feature vectors within a continuous time window are concatenated temporally and input into an LSTM layer with 128 hidden units to capture time-series features such as action continuity and orientation stability. Temporal global pooling is then used to obtain a temporal feature vector. The two types of features are concatenated to form a fused feature vector, which is then transformed nonlinearly through two fully connected layers. Finally, the confidence score in the [0,1] interval is output by the sigmoid activation function. Combined with a preset threshold, a binary result indicating whether the shooting behavior is valid is output, while retaining spatial, temporal features and confidence scores as intermediate information.

[0068] The model training process includes constructing a video dataset covering shooting behaviors and normal scenes from different environments, angles, and terminal types. This dataset is divided into training, validation, and test sets proportionally. After data augmentation processing such as annotation, image flipping, and temporal cropping, the dataset is used for training. The Adam optimizer (initial learning rate 0.001, decaying as needed) and binary cross-entropy loss function are employed. A maximum of 50 training epochs are set, and an early stopping mechanism is introduced to prevent overfitting. The batch size is configured to 32 based on computing power. After forward propagation to output prediction results and backpropagation to update parameters, the model's hyperparameters are fine-tuned using the validation set. Finally, after validation on the test set (accuracy ≥ 95%, recall ≥ 93%, recognition time ≤ 0.3 seconds), the model is deployed to the inference engine.

[0069] Furthermore, the system extracts stability features of shooting behavior within consecutive time windows using a sliding time window method. The length of the sliding time window is set, and parameters such as the duration of the shooting behavior, the amplitude of action changes, and the fluctuation of the lens angle pointing at the screen are calculated within each time window. These parameters are then used as stability features. For example, within a certain sliding time window, the longer the duration of the shooting behavior, the smaller the amplitude of action changes, and the smaller the fluctuation of the lens angle pointing at the screen, the higher the stability of the shooting behavior within that window, and the larger the corresponding stability feature value. These intermediate features, confidence information, and stability features together constitute the behavioral features related to the shooting behavior, providing comprehensive and accurate input data for subsequent deep learning risk assessment. This solves the problem of inaccurate risk assessment caused by relying solely on binary judgment results in existing technologies, enabling risk assessment to more comprehensively reflect the actual shooting scenario.

[0070] In some embodiments, for step S3, the deep learning risk assessment model is trained and risk assessment is performed. The deep learning risk assessment model is the core of achieving accurate risk assessment. It does not exist as an independent algorithm, but as a risk assessment step module in the access control anti-spy camera method. The output result does not directly control the smart power-on anti-spy screen, but serves as an adjustment factor for access control decision and execution parameters. It is used to solve the technical problems that access control and risk are not always in a one-to-one correspondence, that shooting behavior has a "grayscale range", and that traditional rule-based decision-making systems have limited ability to adapt to complex scenarios.

[0071] The model's input includes permission status, permission level parameters, behavioral features, the current state of the smart power-on privacy screen, and historical trigger records of shooting behavior. The current state of the smart power-on privacy screen includes high-transparency and power-on masked states, converted into numerical data recognizable by the model using binary encoding. High-transparency is encoded as 0, and power-on masked state is encoded as 1. Historical trigger records of shooting behavior require extracting information such as the trigger time, duration, corresponding permission status, and risk assessment value of the most recent preset number of triggers to construct a historical feature vector, reflecting the historical patterns of shooting behavior and assisting in current risk assessment. All input data must be standardized before being input into the model, for example, using the Min-Max standardization method to map different types of features to the same numerical range, eliminating dimensional differences between different features, and improving model training efficiency and evaluation accuracy. The standardization formula is:

[0072]

[0073] in, These are the original eigenvalues. This is the minimum value of the feature in the training dataset. This is the maximum value of the feature in the training dataset. These are the standardized feature values.

[0074] The standardized features can be fused. Specifically, the feature vector representing the spatiotemporal characteristics of behavior extracted by the temporal modeling layer of the convolutional neural network is concatenated with the permission status features after one-hot encoding, the normalized permission level parameters, the current status features of the privacy screen using binary encoding, and the historical feature vector constructed from the historical trigger records of shooting behavior to form a fused feature vector.

[0075] In one embodiment, such as Figure 2 As shown, Figure 2 This is a schematic diagram of the risk assessment model structure provided in an embodiment of this application. The model structure includes a convolutional neural network layer, a pooling layer, a temporal modeling layer, and a fully connected layer. The convolutional neural network layer is used to process the spatial features in the behavioral features, and the temporal modeling layer is used to process the time series features in the behavioral features and output a risk assessment value.

[0076] Specifically, a convolutional neural network layer includes multiple convolutional operations. The first convolutional layer can be configured with 32 kernels, a kernel size of 3×3, a stride of 1, and same padding. The activation function used is the ReLU function, and its formula is:

[0077]

[0078] The first layer performs preliminary spatial feature extraction from the input features, enhancing their expressive power. The second convolutional layer can be configured with 64 kernels, a kernel size of 3×3, a stride of 1, and same padding. The ReLU activation function is also used to further extract more complex spatial features. Following the convolutional layers is a pooling layer using max pooling with a kernel size of 2×2 and a stride of 2. This reduces the dimensionality of the features extracted by the convolutional layers, decreasing the number of model parameters, improving model computation speed, while preserving key feature information and avoiding overfitting.

[0079] The temporal modeling layer employs a Long Short-Term Memory (LSTM) network from a recurrent neural network to capture the temporal series features of the shooting behavior and characterize its continuity. The LTM layer can be configured with 128 hidden units. This layer receives the feature sequence output from the pooling layer and uses a gating mechanism to selectively remember and forget time-series information, effectively handling long-sequence dependencies and accurately capturing the changing patterns of the shooting behavior within a continuous time window. For example, when the shooting behavior gradually changes from "the phone is raised but not yet stably aimed at the screen" to "the lens is stably aimed at the screen," the LTM layer can capture the feature differences in this continuous change process, providing temporal support for risk assessment. Following the LTM layer is a fully connected layer, consisting of two fully connected operations. The first fully connected layer can be configured with 64 neurons, using the ReLU activation function to map the feature vector output from the LTM layer to a 64-dimensional feature space. The second fully connected layer has only one neuron, using the Sigmoid activation function, with the following formula:

[0080]

[0081] The output is mapped to the range [0,1]. This output value is the risk assessment value of being secretly filmed in the current time period. It is used to characterize the overall risk level of the screen being secretly filmed. The closer the value is to 1, the higher the risk of being secretly filmed. The closer the value is to 0, the lower the risk of being secretly filmed.

[0082] Before model training, a training dataset can be constructed. The dataset includes samples with different permission states, different shooting scenarios, and different environmental conditions, along with corresponding label values. Label values ​​are manually annotated, set to continuous values ​​within the range [0,1] based on the level of risk of being secretly filmed in the actual scenario. For example, higher label values ​​are set when the filming behavior is clear and continuous, and the environment is crowded; lower label values ​​are set when the phone is briefly raised without being pointed at the screen, and the environment is a single-person enclosed space. During training, mean squared error is used as the loss function, with the following formula:

[0083]

[0084] in, The number of training samples. For the first The true label value of each sample For the first The model predicts values ​​for each sample. The optimizer uses the Adam optimizer, with a configurable learning rate of 0.001 and configurable iterations of 100 rounds. During each iteration, the training dataset is input into the model in batches, and the model parameters are updated through backpropagation to minimize the loss function. An early stopping mechanism is introduced during training. A validation set is set, and the model's loss value on the validation set is calculated after each iteration. Training stops when the validation set loss value no longer decreases after a preset number of iterations, and the current optimal model parameters are saved to avoid overfitting. Simultaneously, the model is periodically tested during training to evaluate metrics such as accuracy and recall, ensuring the model meets the requirements of practical applications.

[0085] After training, the model is deployed to the system's inference engine for real-time risk assessment. The system feeds processed input data into the trained deep learning risk assessment model in real time. The model outputs the current risk assessment value for surreptitious filming through forward propagation. This risk assessment value comprehensively reflects the level of surreptitious filming risk under the influence of multiple factors, including behavioral stability, access sensitivity, and environmental complexity. For example, when an administrator-privileged user is in a crowded environment with frequent phone use, and the model detects the phone being raised and gradually pointed at the screen, the model's risk assessment value will be higher. Conversely, when an ordinary user is in a secluded, single-person environment and the phone is briefly raised without being pointed at the screen, the model's risk assessment value will be lower, providing an accurate and reliable basis for subsequent protection decisions.

[0086] In some embodiments, for step S4, the power-on command issuance condition is determined. After obtaining the shooting behavior recognition result, the risk assessment value output by the deep learning model, and the current permission status, the system will perform a joint judgment to determine whether the conditions for issuing a power-on command to the smart power-on privacy screen are met. This process constitutes the first layer of protection decision logic, which solves the technical problems of the existing technology's single protection trigger logic and insufficient consideration of permission status and actual risks, and realizes differentiated control in the protection trigger stage.

[0087] For administrator privileges, the system presets a first risk threshold and a first time threshold. The first risk threshold determines whether the current risk assessment value reaches a level requiring protection, and the first time threshold determines whether the filming behavior is a continuous high-risk behavior. Once filming is confirmed, the system continuously monitors the risk assessment value and the duration of the filming behavior. Only when the risk assessment value is greater than or equal to the first risk threshold, and the duration of the filming behavior exceeds the first time threshold, will the system send a power-on command to the smart power-on privacy screen. For example, the first risk threshold can be configured to 0.7, and the first time threshold can be configured to 3 seconds. If, under administrator privileges, filming is confirmed, the risk assessment value is 0.8, and the filming behavior lasts for more than 3 seconds, the system triggers a power-on command. If the risk assessment value is 0.6, even if the filming behavior lasts for more than 3 seconds, no power-on command is triggered. If the risk assessment value is 0.8, but the filming behavior terminates after 2 seconds, no power-on command is triggered either. This setting prevents the system from triggering mandatory protection when administrators are performing critical business operations due to accidental or low-risk filming by external personnel, thus disrupting business continuity and ensuring smooth operation for high-privilege users.

[0088] For ordinary users with access privileges, the system presets a second risk threshold, which can be lower than the first risk threshold, for example, configured as 0.5. Since the sensitivity of information accessed by ordinary users is relatively controllable, but the risk of being secretly filmed still needs to be prevented, when filming is confirmed and the risk assessment value is greater than or equal to the second risk threshold, the system immediately sends a power-on command to the smart privacy screen, without waiting for the filming to continue for a certain period. This ensures that high-risk filming behaviors in ordinary user scenarios can be quickly blocked, protecting screen information security.

[0089] For public display scenarios, the system presets a third risk threshold, which can be configured based on the sensitivity of the displayed information. For example, if the information is highly sensitive, the third risk threshold is configured to 0.4; if the information is less sensitive, the third risk threshold is configured to 0.6. When a recording is confirmed and the risk assessment value is greater than or equal to the third risk threshold, the system triggers the intelligent power-on privacy screen to power on, achieving flexible protection in public display scenarios. This ensures that sensitive information is not leaked while avoiding excessive protection that could disrupt normal information display.

[0090] In some embodiments, step S5 involves the switching of the intelligent power-on privacy screen state and the linkage protection. If the system determines that the conditions for issuing a power-on command are met, it immediately issues a power-on command to the intelligent power-on privacy screen. Upon receiving the power-on command, the intelligent power-on privacy screen switches from a high-transparency standby state to a uniformly shaded state. This uniform shading physically blocks screen information, preventing the camera from obtaining clear and valid content. Compared to traditional software shading, there is no risk of partial omissions, making the protection more comprehensive and reliable. The intelligent power-on privacy screen switches from a high-transparency standby state to uniform shading when powered on. This state switching process is fast and efficient, and the switching time can be controlled within a short range, ensuring that the act of secretly filming can be blocked in a timely manner.

[0091] Optionally, when a power-on command is sent to the intelligent power-on privacy screen, the system simultaneously executes linked protection steps, triggering an audible and visual alarm and activating a signal jammer to interfere with signals in a specific frequency band, forming a triple protection system of "physical privacy, on-site alarm, and signal jamming." The LED warning light of the audible and visual alarm flashes and emits an alarm sound, providing both visual and auditory prompts to stop unauthorized filming activities, serving as a deterrent. The signal jammer activates specific frequency band signal interference, preventing the filming device from transmitting the captured content to the outside via the network, further reducing the risk of information leakage and ensuring the comprehensiveness and effectiveness of the protection.

[0092] Furthermore, upon determining that the conditions for issuing a power-on command are met, the system simultaneously executes a data recording step, taking a screenshot of the scene at the time of the protection trigger. The screenshot includes key information such as the person taking the picture, the camera, and the scene environment, providing a complete chain of evidence for subsequent event tracing. The system associates this screenshot with the corresponding trigger time, permission status, and risk assessment value, storing it in a designated directory on the local solid-state drive of the core computing unit. This local storage mode eliminates the data upload to the cloud, ensuring the security of the screenshot data itself and preventing secondary leakage. Screenshots are categorized and archived according to a first-level directory of "shooting date" and a second-level directory of "monitoring camera number," facilitating rapid location of unauthorized screenshots from specific times and locations, thus improving the efficiency of event tracing.

[0093] In some embodiments, for step S6, the coding density is determined and the protection is dynamically adjusted. After the smart power-on privacy screen is powered on, the system determines the coding density under uniform coding state based on the current permission level parameters and risk assessment value, thereby achieving fine-tuning of the protection strength and solving the technical problems of lack of hierarchical control and fixed protection strength in the existing protection execution methods.

[0094] Please see Figure 3 , Figure 3This is a schematic diagram illustrating the process of determining and dynamically adjusting masking density according to an embodiment of this application. In S301, the system obtains the basic masking density parameters corresponding to each permission level. The system pre-configures the corresponding basic masking density parameters for each permission level. The masking density parameters are related to the size and number of masking blocks; the higher the masking density, the greater the degree of screen information obstruction and the stronger the protection. For example, the basic masking density parameter corresponding to the first-level permission level is 0.3, the basic masking density parameter corresponding to the second-level permission level is 0.7, and the basic masking density parameter corresponding to the third-level permission level is 0.9. These basic masking density parameters can be adjusted by the administrator according to the security needs of the actual scenario.

[0095] In step S302, based on a preset adjustment formula, the final coding density is calculated according to the basic coding density parameters and the risk assessment value. After determining the basic coding density parameters, the system calculates the final coding density based on the preset adjustment formula, which is:

[0096]

[0097] in, For the final coding density, This is the base hashing density parameter corresponding to the current permission level. This is an adjustment coefficient, configurable to a value between 0.5 and 1.5, used to adjust the degree of influence of the risk assessment value on the coding density. Based on the current risk assessment value, The risk baseline value can be configured to 0.5. Through this adjustment formula, the masking density can be dynamically adjusted according to changes in the risk assessment value, ensuring a precise match between the protection strength and the actual risk level. For example, under administrator privileges, the basic masking density parameter is 0.3, the adjustment coefficient is 0.8, and the risk baseline value is 0.5. When the risk assessment value is 0.8, the final masking density is 0.3 × [1 + 0.8 × (0.8 - 0.5)] = 0.372; when the risk assessment value is 0.6, the final masking density is 0.3 × [1 + 0.8 × (0.6 - 0.5)] = 0.324, achieving differentiated protection for different risk levels under the same privilege level.

[0098] Optionally, in S303, dynamic adjustments to protection are performed based on continuous monitoring of shooting behavior and risk assessment values. After the smart power-on privacy screen is in a uniformly masked state, the system continuously monitors the shooting behavior and risk assessment values, and performs dynamic adjustments to protection. If the permission status changes, the system recalculates the masking density using the aforementioned adjustment formula based on the permission level parameters corresponding to the changed permission status and the current risk assessment value. It then sends an adjustment command to the smart power-on privacy screen to redetermine the masking density and adjust the masking intensity without needing to re-trigger power, ensuring that the protection intensity can quickly adapt to the new permission status and risk level. For example, in a normal user permission state, protection is activated with a masking density of 0.77. If the user's permission is switched to administrator permission, the system immediately updates the basic masking density parameter to 0.3, recalculates the final masking density based on the current risk assessment value, and performs the adjustment, avoiding over- or under-protection due to permission changes.

[0099] In some embodiments, the system also includes protection recovery and anomaly handling. After the smart power-on privacy screen is in a uniformly masked state, the system continuously monitors the shooting behavior status and risk assessment value, and executes protection recovery steps. If the system detects that the shooting behavior has disappeared, and the risk assessment value drops below a preset recovery threshold and continues to exceed a recovery time threshold, the system controls the smart power-on privacy screen to power off, restoring it to a high-transparency state without affecting normal meetings or workflows. For example, the recovery threshold can be configured to 0.3, and the recovery time threshold can be configured to 5 seconds. When the shooting behavior ends, the risk assessment value drops to 0.2, and no new shooting behavior is triggered for 5 seconds, the screen automatically restores to a high-transparency state, ensuring that the screen can be quickly restored to normal use after the risk is eliminated.

[0100] During system operation, the system monitors its hardware and software status in real time. When hardware failures or software anomalies occur, multi-level prompts enable rapid troubleshooting. If issues such as poor camera contact, insufficient storage space, delayed protection response, or power failure / abnormal power supply to the smart privacy screen are detected, corresponding anomaly prompts will pop up in the system status display area of ​​the main interface. Examples include "Camera 1 has no signal," "Storage space less than 10% remaining," and "Smart privacy screen communication failure," facilitating timely problem identification by on-site personnel. Administrators can access the "System Log" section in the backend to view the time of occurrence, type of anomaly, and detailed causes, enabling rapid equipment maintenance or parameter adjustments to ensure stable system operation and reduce the risk of protection failure due to system anomalies.

[0101] This method constructs a complete, sophisticated, and adaptive screen anti-spy camera protection system through a series of steps, including permission status acquisition and level mapping, shooting behavior monitoring and feature extraction, deep learning risk assessment, differentiated power-on decision-making, dynamic masking density adjustment, and linked protection. This system integrates permission control throughout the entire protection process and, combined with a deep learning model, achieves accurate risk assessment, solving problems in existing technologies such as single protection trigger logic, lack of hierarchical control in protection execution, and mismatch between permissions and risks. The intelligent power-on anti-spy camera screen switches from a high-transparency standby state to uniform masking when powered on, forming multiple protective barriers with audible and visual alarms and signal shielding, ensuring comprehensive protection and rapid, efficient response. Simultaneously, through a dynamic adjustment and recovery mechanism, it minimizes interference with normal operation while ensuring information security, balancing security and practicality.

[0102] It should be noted that although the operations of the method of this application are described in a specific order in the accompanying drawings, this does not require or imply that these operations must be performed in that specific order, or that all the operations shown must be performed to achieve the desired result. On the contrary, the steps depicted in the flowchart can be performed in a different order. Additionally or alternatively, certain steps may be omitted, multiple steps may be combined into one step, and / or one step may be broken down into multiple steps.

[0103] Please see Figure 4 , Figure 4 This application provides a block diagram of a screen anti-spy camera system based on access control. The system specifically includes: an access control module 401, a behavior recognition module 402, a risk assessment module 403, and a protection control module 404.

[0104] The permission management module 401 is used to obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter;

[0105] The behavior recognition module 402 is used to monitor the area around the current screen, identify whether the shooting behavior is valid, and extract the behavioral features related to the shooting behavior.

[0106] The risk assessment module 403 is used to input the permission status, the permission level parameter and the behavioral characteristics into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk.

[0107] The protection control module 404 is communicatively connected to the permission management module 401, the behavior recognition module 402, the risk assessment module 403, and a privacy screen, and is used to determine whether the conditions for issuing a power-on command to the privacy screen are met based on the permission status, whether the shooting behavior is established, and the risk assessment value.

[0108] If the conditions for issuing a power-on command are met, a power-on command is issued to the privacy screen, causing the privacy screen to change from a high-transparency state to a uniform coding state.

[0109] The coding density under the uniform coding state is determined based on the permission level parameter and the risk assessment value.

[0110] Preferably, the permission status includes at least administrator permission status, ordinary user permission status, and public display status;

[0111] The protection control module is specifically used to: when the permission status is administrator permission status, if the shooting behavior is established, and the risk assessment value is greater than or equal to the first risk threshold, and the duration of the shooting behavior exceeds the first time threshold, then it is determined that the conditions for issuing a power-on command are met.

[0112] When the permission status is that of a normal user, if the shooting behavior is successful and the risk assessment value is greater than or equal to the second risk threshold, then it is determined that the conditions for issuing a power-on command are met, and the second risk threshold is less than the first risk threshold.

[0113] Preferably, the deep learning risk assessment model includes a convolutional neural network layer and a temporal modeling layer;

[0114] The convolutional neural network layer is used to process the spatial features in the behavioral features;

[0115] The time-series modeling layer is used to process the time-series features in the behavioral characteristics and output the risk assessment value.

[0116] It should be noted that the working process of each module in the access control-based screen anti-spy camera system described in this embodiment can refer to the working process of the access control-based screen anti-spy camera method described in the above embodiments, and the technical effect achieved is the same as that of the access control-based screen anti-spy camera method described in the above embodiments, so it will not be repeated here.

[0117] The above description represents the preferred embodiments of the present invention. It should be noted that, for those skilled in the art, various improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications are also considered to be within the scope of protection of the present invention.

Claims

1. A screen anti-spy photography method based on access control, characterized in that, Includes the following steps: Obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter; The system monitors the area surrounding the current screen, identifies whether a shooting action is valid, and extracts behavioral features related to the shooting action. The permission status, the permission level parameter, and the behavioral characteristics are input into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk. Based on the permission status, whether the shooting behavior is valid, and the risk assessment value, determine whether the conditions for issuing a power-on command to the privacy screen are met; If the conditions for issuing a power-on command are met, a power-on command is issued to the privacy screen, causing the privacy screen to change from a high-transparency state to a uniform coding state. The coding density under the uniform coding state is determined based on the permission level parameter and the risk assessment value.

2. The screen anti-spy photography method based on access control according to claim 1, characterized in that, The permission status includes at least administrator permission status, ordinary user permission status, and public display status; Based on the permission status, whether the shooting behavior is valid, and the risk assessment value, the conditions for determining whether a power-on command to be issued to the privacy screen are met include: When the permission status is administrator permission status, if the shooting behavior is established, the risk assessment value is greater than or equal to the first risk threshold, and the duration of the shooting behavior exceeds the first time threshold, then it is determined that the conditions for issuing the power-on command are met. When the permission status is that of a normal user, if the shooting behavior is successful and the risk assessment value is greater than or equal to the second risk threshold, then it is determined that the conditions for issuing a power-on command are met, and the second risk threshold is less than the first risk threshold.

3. The screen anti-spy photography method based on access control according to claim 2, characterized in that, The input to the deep learning risk assessment model also includes the current state of the privacy screen and the historical trigger records of shooting behavior; The behavioral features include intermediate features generated during the shooting behavior recognition process, confidence information, and shooting behavior stability features extracted from a continuous time window.

4. The screen anti-spy photography method based on access control according to claim 3, characterized in that, The deep learning risk assessment model includes a convolutional neural network layer and a temporal modeling layer; The convolutional neural network layer is used to process the spatial features in the behavioral features; The time-series modeling layer is used to process the time-series features in the behavioral characteristics and output the risk assessment value.

5. A screen anti-spy camera method based on access control according to claim 1, characterized in that, Based on the permission level parameter and the risk assessment value, the specific coding density under the uniform coding state is determined as follows: Get the base CAPTCHA density parameter corresponding to the current permission level parameter; Based on the preset adjustment formula, the final coding density is calculated according to the basic coding density parameter and the risk assessment value. The adjustment formula is: coding density = basic coding density parameter × [1 + adjustment coefficient × (risk assessment value - risk benchmark value)].

6. A screen anti-spy photography method based on access control according to claim 5, characterized in that, After the privacy screen is in a powered-on, uniformly coded state, the method further includes dynamic adjustment of the protection: Continuously monitor the shooting behavior status and the risk assessment value; If the permission status changes, the masking density is re-determined and the masking intensity of the privacy screen is adjusted based on the permission level parameter corresponding to the changed permission status and the current risk assessment value.

7. A screen anti-spy camera method based on access control according to claim 6, characterized in that, After the privacy screen is in a powered, uniformly coded state, the method further includes protection restoration: If the detected shooting behavior disappears, and the risk assessment value drops below the preset recovery threshold and continues to exceed the recovery time threshold, then the privacy screen is powered off to restore it to a high-transparency state.

8. A screen anti-spy camera system based on access control, characterized in that, It includes a permission management module, a behavior recognition module, a risk assessment module, and a protection and control module; The permission management module is used to obtain the permission status of the current screen user and map the permission status to the corresponding permission level parameter; The behavior recognition module is used to monitor the area around the current screen, identify whether the shooting behavior is valid, and extract the behavioral features related to the shooting behavior. The risk assessment module is used to input the permission status, the permission level parameter, and the behavioral characteristics into a preset deep learning risk assessment model to obtain a risk assessment value that represents the current level of surreptitious filming risk. The protection control module is communicatively connected to the permission management module, the behavior recognition module, the risk assessment module, and a privacy screen, and is used to determine whether the conditions for issuing a power-on command to the privacy screen are met based on the permission status, whether the shooting behavior is valid, and the risk assessment value. If the conditions for issuing a power-on command are met, a power-on command is issued to the privacy screen, causing the privacy screen to change from a high-transparency state to a uniform coding state. The coding density under the uniform coding state is determined based on the permission level parameter and the risk assessment value.

9. A screen anti-spy camera system based on access control according to claim 8, characterized in that, The permission status includes at least administrator permission status, ordinary user permission status, and public display status; The protection control module is specifically used for: When the permission status is administrator permission status, if the shooting behavior is established, the risk assessment value is greater than or equal to the first risk threshold, and the duration of the shooting behavior exceeds the first time threshold, then it is determined that the conditions for issuing the power-on command are met. When the permission status is that of a normal user, if the shooting behavior is successful and the risk assessment value is greater than or equal to the second risk threshold, then it is determined that the conditions for issuing a power-on command are met, and the second risk threshold is less than the first risk threshold.

10. A screen anti-spy camera system based on access control according to claim 9, characterized in that, The deep learning risk assessment model includes a convolutional neural network layer and a temporal modeling layer; The convolutional neural network layer is used to process the spatial features in the behavioral features; The time-series modeling layer is used to process the time-series features in the behavioral characteristics and output the risk assessment value.

Images