Gun pulling behavior automatic identification and platform early warning system based on internet of things

Abstract

The application discloses a gun pulling behavior automatic identification and platform early warning system based on Internet of Things, relates to the technical field of communication, and comprises a model construction module, a risk grading module, a signal identification module, an edge processing module and a linkage support module.The model construction module is used for constructing a dynamic trust evaluation model.The risk grading module is used for grading the risk of a gun pulling event.The signal identification module is used for acquiring a gun pulling signal of a law enforcement site, identifying a false touch of the gun pulling signal, performing edge computing real-time analysis, obtaining a high-confidence event, and sending the high-confidence event to a platform.The linkage support module is used for triggering a differentiated linkage strategy to support the platform according to the high-confidence event.The communication security module is used for dynamically encrypting a communication channel and dynamically changing a security protocol at random.The application has the effects of improving the accuracy and efficiency of situation identification and support force calling in a law enforcement site and the safety of Internet of Things communication.

Description

Technical Field

[0001] This application relates to the field of communication technology, and in particular to an IoT-based automatic identification and platform-based early warning system for gun-drawing behavior. Background Technology

[0002] During the course of their duties, law enforcement officers may encounter conflicts with those being enforced against. In cases where those being enforced against refuse to cooperate or violently resist, law enforcement officers may be required to draw their weapons to suppress or defend themselves.

[0003] Current technologies are insufficient for the automatic identification and early warning of law enforcement officers' drawing of weapons. They typically rely on a simple sensor on the holster to detect the action. We know that once a law enforcement officer draws their weapon, it indicates that the conflict at the scene has escalated to a very serious level. However, as demonstrated by the aforementioned traditional technologies, simply identifying the drawing action and uploading it to the backend is largely ineffective. After the drawing action, there is no further support to help mitigate the high risk at the scene. Furthermore, traditional weapon drawing detection methods are prone to inaccurate measurements, potentially leading to false alarms and negatively impacting the law enforcement situation. Summary of the Invention

[0004] The purpose of this invention is to provide an automatic identification and platform-based early warning system for gun-drawing behavior based on the Internet of Things, in order to solve the problems mentioned in the background art.

[0005] This application provides an IoT-based automatic gun-drawing behavior recognition and platform-based early warning system, the system comprising: Model building module: used to acquire law enforcement officers' daily physiological data, daily behavioral data, current task type and historical law enforcement data, and to build a dynamic trust assessment model based on the daily physiological data, daily behavioral data, current task type and historical law enforcement data; Risk grading module: used to obtain the on-site risk level of law enforcement scene, and to grade the risk of the gun-drawing incident according to the dynamic trust assessment model and the on-site risk level, to obtain high-level risk or low-level risk. Signal recognition module: used to acquire gun-drawing signals at the law enforcement scene, identify accidental gun-drawing signals, and obtain target gun-drawing signals; Edge processing module: used to perform real-time edge computing analysis based on the target's gun-drawing signal and the high-level risk or the low-level risk, to obtain high-confidence events, and send the high-confidence events to the platform; Linkage Support Module: This module is used by the platform to trigger differentiated linkage strategies based on the high-confidence events, generate target linkage strategies, and call IoT nodes connected to the platform for support based on the target linkage strategies. Communication security module: used to acquire the communication methods and security protocols between IoT nodes connected to the platform, dynamically encrypt the communication channel based on the communication methods, and randomly and dynamically change the security protocols based on the security protocols.

[0006] Preferably, the step of constructing a dynamic trust assessment model based on the daily physiological data, the daily behavioral data, the current task type, and the historical law enforcement data specifically includes: Based on the daily physiological data and the daily behavioral data, the physiological and behavioral patterns of law enforcement officers are obtained respectively. Based on the physiological and behavioral patterns, a baseline of law enforcement officers’ daily personal activities is constructed, and personal trust parameters of law enforcement officers are obtained based on the baseline of daily personal activities. Based on the current task type, the historical law enforcement data is filtered to obtain the target historical law enforcement data; Extract historical law enforcement scene information from the target historical law enforcement data, and obtain scene trust parameters for the historical law enforcement scene based on the historical law enforcement scene information; A dynamic trust assessment model is constructed by combining the personal trust parameters and the scenario trust parameters.

[0007] Preferably, the step of obtaining the on-site risk level of the law enforcement scene and classifying the risk of the gun-drawing incident based on the dynamic trust assessment model and the on-site risk level is as follows: Obtain on-site physiological data and body camera data of law enforcement officers at the law enforcement scene, and determine the on-site risk level based on the on-site physiological data and body camera data; Substituting the on-site physiological data and the law enforcement recorder data into the dynamic trust assessment model, the confidence level of the on-site information is obtained; The on-site risk level is identified based on the confidence level of the on-site information, and the identification result is obtained. The risk level is then classified according to the identification result to obtain a high-level risk or a low-level risk.

[0008] Preferably, the step of acquiring the weapon-drawing signal at the law enforcement scene, identifying accidental activation of the weapon-drawing signal, and obtaining the target weapon-drawing signal specifically includes: The triggering method and historical signal triggering records of the gun draw signal are obtained. The risk of accidental signal activation is assessed based on the triggering method to obtain a first accidental activation probability parameter. The historical signal trigger records are filtered for false touches to obtain historical false touch records, and the false touch methods and number of false touches in the historical false touch records are identified. Based on the accidental touch method and the number of accidental touches, the risk of accidental signal touch is assessed to obtain a second accidental touch probability parameter; Acquire the gun-drawing signal at the law enforcement scene, and evaluate the confidence level of the gun-drawing signal based on the first accidental trigger probability parameter and the second accidental trigger probability parameter to obtain the signal confidence level. Determine whether the credibility of the signal exceeds a preset credibility threshold. If the credibility of the signal exceeds the credibility threshold, then mark the gun-drawing signal as a target gun-drawing signal.

[0009] Preferably, the step of performing real-time edge computing analysis based on the target draw signal and the high-level risk or the low-level risk to obtain the high-confidence event specifically includes: Based on the high-level risk or the low-level risk, the on-site physiological data and the law enforcement recorder data at the law enforcement scene are preprocessed and packaged to generate an on-site law enforcement data package; The on-site law enforcement data package is sorted to obtain multiple on-site data information, and the confidence and urgency of each on-site data information are evaluated to obtain the data confidence and data urgency. Based on the data confidence level and the data urgency, multiple pieces of on-site data are filtered to obtain the target on-site data. The target's on-site data and the target's gun-drawing signal are merged and encapsulated to generate a high-confidence event.

[0010] Preferably, after the step of sending the high-confidence event to the platform, the method further includes: A local memory is set on the smart holster, and the high-confidence event is copied and stored in the local memory; During the transmission of the high-confidence event, key information is identified from the high-confidence event to obtain key information data from the law enforcement scene; Identify the network link between the smart holster and the platform, determine the blockchain storage location based on the network link, and store the key information in the blockchain storage location.

[0011] Preferably, the steps of the platform triggering a differentiated linkage strategy based on the high-confidence event, generating a target linkage strategy, and calling on IoT nodes connected to the platform for support according to the target linkage strategy are as follows: The platform identifies the high-confidence events, obtains the support index at the law enforcement scene, and determines whether the support index reaches a preset standard value. If it is determined that the supported index has reached the standard value, the differentiated linkage strategy is triggered, and a higher-level linkage strategy is selected in the differentiated linkage strategy. According to the high-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are filtered to obtain multiple advanced support nodes. All of the advanced support nodes are then called to the law enforcement scene. If it is determined that the support index has not reached the standard value, the differentiated linkage strategy is triggered, and a low-level linkage strategy is selected in the differentiated linkage strategy. According to the low-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are screened to obtain multiple low-level support nodes. Based on the high-confidence event, the type of support force required at the law enforcement site is obtained. The low-level support nodes are then filtered according to the type of support force to obtain the target support node, and the target support node is called to the law enforcement site.

[0012] Preferably, the step of dynamically encrypting the communication channel based on the communication method specifically includes: Based on the aforementioned communication method, the communication channel between the IoT node and the platform is identified, and the communication channel is subjected to vulnerability review to obtain channel vulnerabilities. Extract the vulnerability location of the channel vulnerability and the attack methods that can be used to break through the channel vulnerability, and set vulnerability decoys and countermeasures at the channel vulnerability location based on the vulnerability location and the attack methods; When the vulnerability decoy is triggered, the countermeasure mechanism identifies the external attack method and the external attack source, and performs countermeasure attacks based on the external attack method and the external attack source; Identify the current encryption method of the communication channel, decrypt the current encryption method, and obtain the channel encryption construction method; The channel encryption construction method is dynamically adjusted to obtain a dynamic construction method. A dynamic encryption method is generated based on the dynamic construction method, and the communication channel is updated based on the dynamic encryption method.

[0013] Preferably, the step of dynamically modifying the security protocol based on the security protocol includes: Based on the security protocol, the protocol framework of the security protocol is extracted, and multiple new security protocols are randomly generated according to the protocol framework; Based on the multiple new security protocols, a new security protocol table is generated, and the new security protocols in the table are randomly sorted to obtain a random sequence; Select the first new security protocol in the random sequence, update the security protocol, and then reorder the remaining new security protocols randomly. Within a preset time period, a time point is randomly selected as the update time point, and the security protocol is updated within the update time point.

[0014] In summary, this application includes at least one of the following beneficial technical effects: A dynamic trust assessment model is constructed by leveraging law enforcement officers' daily physiological and behavioral data, current task types, and historical law enforcement data. This model then assesses the risk level of on-site incidents and categorizes weapon-drawing events based on the risk level and the dynamic trust assessment model. Next, weapon-drawing signals are acquired at the scene and identified as accidental draws. If no accidental draw is confirmed, the signal is marked as a target weapon-drawing signal. Real-time edge computing analysis is then performed based on the target weapon-drawing signal and whether the risk level is high or low, identifying high-confidence events and sending them to the platform. The platform identifies high-confidence events and triggers differentiated linkage strategies, calling upon other IoT nodes connected to the platform, including drones and available law enforcement officers, to provide support at the scene. The communication methods and security protocols between the platform and the IoT nodes are then identified. The communication channels are dynamically encrypted based on the communication method, and the security protocols are dynamically modified and randomly updated. This improves the accuracy and efficiency of situation identification at the law enforcement scene, the mobilization of support forces, and the security of IoT communication. Attached Figure Description

[0015] Figure 1 This is a block diagram of a module for an IoT-based automatic gun-drawing behavior recognition and platform-based early warning system provided in an embodiment of this application.

[0016] Explanation of reference numerals in the attached diagram: 1. Model building module; 2. Risk classification module; 3. Signal recognition module; 4. Edge processing module; 5. Linkage support module; 6. Communication security module. Detailed Implementation

[0017] The following is in conjunction with the appendix Figure 1 This application will be described in further detail, but the embodiments of the present invention are not limited thereto.

[0018] This application discloses an Internet of Things-based automatic identification and platform-based early warning system for gun-drawing behavior.

[0019] In this embodiment, an IoT-based automatic gun-drawing behavior recognition and platform-based early warning system is provided, which includes: Model building module 1: Used to acquire law enforcement officers' daily physiological data, daily behavioral data, current task type and historical law enforcement data, and to build a dynamic trust assessment model based on daily physiological data, daily behavioral data, current task type and historical law enforcement data; Risk Classification Module 2: Used to obtain the on-site risk level of law enforcement scene. Based on the dynamic trust assessment model and the on-site risk level, the risk of the gun-drawing incident is classified into high-level risk or low-level risk. Signal recognition module 3: Used to acquire the gun-drawing signal at the law enforcement scene, identify accidental gun-drawing signals, and obtain the target gun-drawing signal; Edge processing module 4: It is used to perform real-time edge computing analysis based on the target's gun-drawing signal and high-level or low-level risk to obtain high-confidence events and send the high-confidence events to the platform; Linkage Support Module 5: This module is used by the platform to trigger differentiated linkage strategies based on high-confidence events, generate target linkage strategies, and call IoT nodes connected to the platform for support based on the target linkage strategies. Communication security module 6: Used to obtain the communication methods and security protocols between IoT nodes connected to the platform, dynamically encrypt the communication channel based on the communication method, and dynamically change the security protocol randomly based on the security protocol.

[0020] The steps for constructing a dynamic trust assessment model based on daily physiological data, daily behavioral data, current task type, and historical law enforcement data are as follows: Based on daily physiological and behavioral data, the physiological and behavioral patterns of law enforcement officers were obtained, respectively. Based on physiological and behavioral patterns, a baseline of law enforcement officers’ daily personal activities is constructed, and personal trust parameters of law enforcement officers are obtained based on the baseline of daily personal activities. Based on the current task type, historical law enforcement data is filtered to obtain the target historical law enforcement data; Extract historical law enforcement scene information from the target's historical law enforcement data, and obtain scene trust parameters for the historical law enforcement scene based on the historical law enforcement scene information; A dynamic trust assessment model is constructed by combining personal trust parameters and scenario-based trust parameters.

[0021] In practice, taking Officer A's law enforcement scene as an example, the first step is to obtain his daily physiological and behavioral data. His daily physiological data includes a heart rate typically around 70 beats per minute and stable blood pressure at 120 / 80 mmHg. His daily behavioral data includes waking up at 7 am and going to bed at 11 pm daily, walking approximately 10,000 steps daily. Based on this physiological and behavioral data, Officer A's physiological pattern is characterized by a stable heart rate and normal blood pressure, while his behavioral pattern is characterized by regular sleep patterns and moderate activity levels. Then, based on these physiological and behavioral patterns, a baseline for Officer A's daily personal activities is constructed. This baseline shows that his physiological indicators and behavioral habits are stable on normal workdays. Based on this baseline, Officer A's personal trust parameters are obtained; for example, his trust score is 85, indicating that he is a reliable law enforcement officer. Next, based on Officer A's current task type, such as handling a street dispute, his historical law enforcement data is filtered to obtain target historical law enforcement data similar to street disputes, such as five similar disputes he has handled in the past. Historical law enforcement scene information is extracted from these target historical law enforcement data, such as the number of people present, the level of conflict, and the outcome of the handling. Based on this historical law enforcement scene information, scene trust parameters for the historical law enforcement scene are obtained. For example, in a similar scenario, A's on-site handling score is 90 points.

[0022] The steps for obtaining the on-site risk level of law enforcement incidents and classifying the risk of gun-drawing incidents based on the dynamic trust assessment model and the on-site risk level are as follows: Obtain on-site physiological data and body camera data from law enforcement officers at the scene of law enforcement, and determine the on-site risk level based on the on-site physiological data and body camera data; By substituting on-site physiological data and law enforcement recorder data into a dynamic trust assessment model, the confidence level of on-site information is obtained; The risk level at the scene is identified based on the confidence level of the information at the scene, and the identification results are then classified into high-level risk or low-level risk based on the identification results.

[0023] In practice, taking Officer A's law enforcement scene as an example, the system acquires on-site physiological data and body camera data. Officer A's on-site physiological data includes a heart rate rising to 100 beats per minute and blood pressure rising to 140 / 90 mmHg. Body camera data shows there are 5 people at the scene, 2 of whom are agitated and speaking loudly. Based on this on-site physiological data and body camera data, by analyzing the changes in heart rate and blood pressure and the conflict situation involving the number of people at the scene, the on-site risk level is determined to be medium risk. Then, Officer A's on-site physiological data, such as a heart rate of 100 beats per minute and blood pressure of 140 / 90, and the body camera data, such as 2 of the 5 people being agitated, are substituted into the previously constructed dynamic trust assessment model. The model calculates based on this data, obtaining an 80% confidence level for the on-site information, indicating that the current on-site information is highly credible. Next, based on this 80% confidence level, the medium risk level is identified, and the result is that the current on-site situation does indeed have some risk, but it is still within a controllable range. Finally, based on this identification result, the risk level was classified as low risk because the confidence level was high and the risk level was medium. This means that although there is a risk, the highest level of alert is not required for the time being.

[0024] The steps for obtaining the weapon-drawing signal at the law enforcement scene, identifying false triggers of the weapon-drawing signal, and obtaining the target's weapon-drawing signal are as follows: The triggering method and historical signal triggering records of the draw signal are obtained. The risk of accidental signal activation is assessed based on the triggering method to obtain the first accidental activation probability parameter. Filter historical signal trigger records to obtain historical false touch records, and identify the false touch method and number of false touches in the historical false touch records; The risk of accidental touch is assessed based on the method and number of accidental touches, resulting in a second accidental touch probability parameter; Acquire the gun-drawing signal at the law enforcement scene, and evaluate the confidence of the gun-drawing signal based on the first and second accidental trigger probability parameters to obtain the signal credibility. Determine whether the signal credibility exceeds a preset credibility threshold. If the signal credibility exceeds the credibility threshold, mark the gun-drawing signal as the target gun-drawing signal.

[0025] In application, taking Officer A's smart holster as an example, the system first acquires the trigger method and historical signal trigger records for the draw signal. The trigger method is a push-button action, and the historical signal trigger records show that it has been triggered 10 times in the past year. Based on this push-button trigger method, the risk of accidental activation is assessed, as the button can be accidentally pressed, resulting in a first accidental activation probability parameter of 15%. Then, the historical signal trigger records are filtered for accidental activations, revealing that 2 out of the 10 triggers were accidental activations, such as one when the holster was caught in a car door and another when it bumped into a wall during patrol. The accidental activation methods for these two instances were identified as squeezing and collision, with two instances of accidental activation. Based on the squeezing and collision activation methods and the two instances of accidental activation, the risk of accidental activation is assessed, resulting in a second accidental activation probability parameter of 20%. Next, the draw signal at the current law enforcement scene is acquired, showing that the button has been pressed. Based on the first accidental activation probability parameter of 15% and the second accidental activation probability parameter of 20%, the confidence level of the draw signal is assessed, resulting in a signal confidence level of 65%. Finally, it was determined whether the credibility of this 65% signal exceeded the preset credibility threshold of 70%. Since 65% did not exceed 70%, this gun-drawing signal was not marked as a target gun-drawing signal. The system considered it to be a false trigger and further observation was required.

[0026] Based on the target's draw signal and whether the risk level is high or low, the steps for performing real-time edge computing analysis to obtain high-confidence events are as follows: Based on high or low risk levels, on-site physiological data and law enforcement recorder data are preprocessed and packaged to generate on-site law enforcement data packages. The on-site law enforcement data packets were sorted to obtain multiple on-site data information. The confidence and urgency of each on-site data information were assessed to obtain the data confidence and data urgency. Multiple pieces of on-site data are filtered based on data confidence level and data urgency to obtain the target on-site data. The target's on-site data and the target's gun-drawing signal are merged and encapsulated to generate a high-confidence event.

[0027] In practice, taking Officer A's law enforcement scene as an example, based on the previously obtained low-risk assessment, the on-site physiological data and body camera data are preprocessed and packaged. The on-site physiological data includes a heart rate of 100 beats per minute and blood pressure of 140 / 90. The body camera data includes video and audio of two out of five people being agitated. This data is packaged into a single on-site law enforcement data package, approximately 50MB in size. This data package is then processed to obtain multiple pieces of on-site data information. For example, the first piece of information is "5 people on-site," the second is "2 people are emotionally agitated," and the third is "the officer's heart rate is elevated." Each piece of on-site data is assessed for confidence and urgency. For instance, the confidence level for "5 people on-site" is 95%, and the urgency level is 30%; the confidence level for "2 people are emotionally agitated" is 80%, and the urgency level is 70%; the confidence level for "the officer's heart rate is elevated" is 90%, and the urgency level is 60%. Based on the data confidence and urgency, these on-site data are filtered, retaining only information with a confidence level higher than 80% and an urgency level higher than 50%. After filtering, target scene data information was obtained, including "two people are emotionally agitated" and "the police officer's heart rate is elevated." Finally, these two target scene data pieces were merged and encapsulated with previously unmarked draw signals to generate a high-confidence event. This event contains key risk and status information of the current scene.

[0028] After sending high-confidence events to the platform, the following steps are also included: Set up local memory on the smart holster and copy high-confidence events and store them in the local memory; During the transmission of high-confidence events, key information is identified to obtain key information data from the law enforcement scene; Identify the network link between the smart holster and the platform, determine the blockchain storage location based on the network link, and store the key information in the blockchain storage location.

[0029] In practice, taking Officer A's law enforcement scene as an example, a local storage device with a capacity of 32GB is installed on Officer A's smart holster. When a high-confidence event is generated, this event is copied and stored in this local storage device, ensuring that data is not lost even if the network is interrupted. During the process of sending the high-confidence event to the platform, key information is identified from the event content. For example, key information data of the law enforcement scene is identified from the event, including "Officer A, badge number 12345", "Location: Renmin Road intersection", "Risk level: low", and "Number of people involved: 5". Then, the network link between the smart holster and the platform is identified, and it is found that the current 4G mobile network is used with good signal strength. Based on this network link, a blockchain storage location is determined, such as the address of a distributed storage node. Finally, this key information data, such as officer information and location information, is stored in this blockchain storage location. Utilizing the immutability of blockchain, the security and traceability of key information are guaranteed, preventing data from being maliciously modified or deleted.

[0030] The platform triggers differentiated linkage strategies based on high-confidence events, generates target linkage strategies, and calls upon IoT nodes connected to the platform for support according to the target linkage strategies. The specific steps are as follows: The platform identifies high-confidence events, obtains the support index at the law enforcement scene, and determines whether the support index reaches the preset standard value. If the supported index is determined to have reached the standard value, a differentiated linkage strategy is triggered, and a higher-level linkage strategy is selected from the differentiated linkage strategies. Based on the high-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are screened to obtain multiple high-level support nodes. All high-level support nodes are then called to the law enforcement scene. If it is determined that the support index has not reached the standard value, a differentiated linkage strategy is triggered, and a low-level linkage strategy is selected in the differentiated linkage strategy. Based on the low-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are screened to obtain multiple low-level support nodes. Based on high-confidence events, the type of support force needed at the law enforcement site is obtained. Low-level support nodes are then filtered according to the type of support force to obtain the target support node, and the target support node is called to the law enforcement site.

[0031] In practice, taking Officer A's law enforcement scene as an example, after receiving a high-confidence event, the platform identifies the event and obtains the support index for the law enforcement scene. By analyzing information such as the risk level of the event, the number of people on site, and the officer's status, the support index is calculated to be 60 points. Then, it is determined whether this score of 60 points reaches the preset standard value of 80 points. Since the score of 60 points does not reach 80 points, a differentiated linkage strategy is triggered, and a low-level linkage strategy is selected. According to the low-level linkage strategy, all IoT nodes connected to the platform are identified, including 10 patrol cars, 5 drones, and 20 idle law enforcement officers. These IoT nodes are then filtered, selecting nodes that are idle and close to the scene, resulting in multiple low-level support nodes, such as 3 patrol cars and 2 idle law enforcement officers. Further, based on the high-confidence event, the type of support force needed at the law enforcement scene is analyzed. For example, if the current situation is a street dispute, police support and on-site recording support may be needed. Based on the type of support force, low-level support nodes are filtered, such as selecting 2 idle law enforcement officers for police support and 1 drone for aerial monitoring and recording. Finally, these target support nodes, namely two law enforcement officers and one drone, are called upon to go to the law enforcement site at the intersection of Renmin Road to provide support. The platform sends instructions and on-site location information to these nodes.

[0032] The steps for dynamically encrypting a communication channel based on the communication method are as follows: Based on the communication method, the communication channel between the IoT node and the platform is identified, and the vulnerability of the communication channel is examined to obtain the channel vulnerability. Extract the location of channel vulnerabilities and the attack methods that can be used to exploit them, and set up vulnerability decoys and countermeasures at the channel vulnerabilities based on the vulnerability location and attack methods. When the vulnerability decoy is triggered, the countermeasure mechanism identifies the external attack method and the source of the external attack, and carries out countermeasure attacks based on the external attack method and the source of the external attack. Identify the current encryption method of the communication channel, decrypt the current encryption method, and obtain the channel encryption construction method; The channel encryption construction method is dynamically adjusted to obtain a dynamic construction method. A dynamic encryption method is generated based on the dynamic construction method, and the communication channel is updated based on the dynamic encryption method.

[0033] In practice, taking Officer A's law enforcement scene as an example, based on the communication method between IoT nodes and the platform, such as a 4G mobile network, the communication channel between them is identified. This communication channel undergoes vulnerability review, and scanning tools are used to discover a vulnerability, such as insufficient encryption strength at a certain stage of data transmission. The location of this vulnerability is extracted, such as in the data encapsulation layer, and the attack methods that can be used to bypass it, such as replay attacks. Based on the vulnerability location and attack methods, a vulnerability decoy is set at the vulnerability site, such as a piece of disguised critical data, and a countermeasure mechanism is set up, which is activated once the decoy is triggered. When the vulnerability decoy is triggered by an external attack, the countermeasure mechanism immediately identifies the external attack method, such as identifying a replay attack, and identifies the IP address of the attack source. Based on the attack method and source, the countermeasure mechanism performs a counterattack, such as sending a large number of invalid data packets to the attack source, blocking its network. Simultaneously, the current encryption method of the communication channel is identified, such as AES-128 encryption, and the current encryption method is decrypted and analyzed to obtain the encryption construction method of the channel, including the key generation method and data block processing flow. This encryption method can be dynamically adjusted, for example, by increasing the key length to 256 bits or changing the data block order, resulting in a new dynamic encryption method. Based on this dynamic method, a new dynamic encryption method, such as AES-256-CBC, can be generated, and this new encryption method can be used to update the entire communication channel, improving channel security.

[0034] The steps for dynamically modifying a security protocol based on a random mechanism are as follows: Based on the security protocol, the protocol framework of the security protocol is extracted, and multiple new security protocols are randomly generated according to the protocol framework; Based on multiple new security protocols, a new security protocol table is generated, and the new security protocols in the table are randomly sorted to obtain a random sequence; Select the first new security protocol in the random sequence, update the security protocol, and then reorder the remaining new security protocols randomly. Within a preset time period, a time point is randomly selected as the update time point, and the security protocol is updated within the update time point.

[0035] In practice, taking Officer A's law enforcement scene as an example, based on the security protocol used between the platform and IoT nodes, such as the TLS 1.3 protocol, the protocol framework is first extracted, including key components like the handshake protocol and record protocol. Then, based on this framework, multiple new security protocols are randomly generated, for example, five new protocols, each with different modifications to the handshake process and encryption algorithm. A new security protocol table is generated based on these five new protocols, listing the ID and characteristics of each new protocol. These new security protocols in the table are randomly sorted to obtain a random sequence, for example, [protocol C, protocol A, protocol E, protocol B, protocol D]. The first new security protocol in the random sequence, protocol C, is selected and used to update the currently used TLS 1.3 security protocol, replacing the old protocol. Then, the remaining new security protocols, namely protocols A, E, B, and D, are re-randomly sorted to obtain a new sequence, for example, [protocol E, protocol B, protocol A, protocol D], preparing for the next update. Within a preset time period, such as the next 24 hours, several time points are randomly selected as update times, for example, 3 AM and 4 PM. At these randomly selected update times, the security protocol is updated again; for example, at 3 AM, the protocol is updated to the next protocol in the sequence, i.e., protocol E. This achieves random and dynamic changes to the security protocol, increasing the difficulty for attackers to crack it.

[0036] The above are all preferred embodiments of this application, and are not intended to limit the scope of protection of this application. Therefore, all equivalent changes made in accordance with the structure, shape and principle of this application should be covered within the scope of protection of this application.

Claims

1. An IoT-based automatic gun-drawing behavior recognition and platform-based early warning system, characterized in that, include: Model building module: used to acquire law enforcement officers' daily physiological data, daily behavioral data, current task type and historical law enforcement data, and to build a dynamic trust assessment model based on the daily physiological data, daily behavioral data, current task type and historical law enforcement data; Risk grading module: used to obtain the on-site risk level of law enforcement scene, and to grade the risk of the gun-drawing incident according to the dynamic trust assessment model and the on-site risk level, to obtain high-level risk or low-level risk. Signal recognition module: used to acquire gun-drawing signals at the law enforcement scene, identify accidental gun-drawing signals, and obtain target gun-drawing signals; Edge processing module: used to perform real-time edge computing analysis based on the target's gun-drawing signal and the high-level risk or the low-level risk, to obtain high-confidence events, and send the high-confidence events to the platform; Linkage Support Module: This module is used by the platform to trigger differentiated linkage strategies based on the high-confidence events, generate target linkage strategies, and call IoT nodes connected to the platform for support based on the target linkage strategies. Communication security module: used to acquire the communication methods and security protocols between IoT nodes connected to the platform, dynamically encrypt the communication channel based on the communication methods, and randomly and dynamically change the security protocols based on the security protocols.

2. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 1, characterized in that, The steps for constructing a dynamic trust assessment model based on the daily physiological data, the daily behavioral data, the current task type, and the historical law enforcement data are as follows: Based on the daily physiological data and the daily behavioral data, the physiological and behavioral patterns of law enforcement officers are obtained respectively. Based on the physiological and behavioral patterns, a baseline of law enforcement officers’ daily personal activities is constructed, and personal trust parameters of law enforcement officers are obtained based on the baseline of daily personal activities. Based on the current task type, the historical law enforcement data is filtered to obtain the target historical law enforcement data; Extract historical law enforcement scene information from the target historical law enforcement data, and obtain scene trust parameters for the historical law enforcement scene based on the historical law enforcement scene information; A dynamic trust assessment model is constructed by combining the personal trust parameters and the scenario trust parameters.

3. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 2, characterized in that, The steps for obtaining the on-site risk level of law enforcement incidents and classifying the risk of gun-drawing incidents based on the dynamic trust assessment model and the on-site risk level are as follows: Obtain on-site physiological data and body camera data of law enforcement officers at the law enforcement scene, and determine the on-site risk level based on the on-site physiological data and body camera data; Substituting the on-site physiological data and the law enforcement recorder data into the dynamic trust assessment model, the confidence level of the on-site information is obtained; The on-site risk level is identified based on the confidence level of the on-site information, and the identification result is obtained. The risk level is then classified according to the identification result to obtain a high-level risk or a low-level risk.

4. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 3, characterized in that, The steps of obtaining the weapon-drawing signal at the law enforcement scene, identifying accidental activation of the weapon-drawing signal, and obtaining the target weapon-drawing signal are as follows: The triggering method and historical signal triggering records of the gun draw signal are obtained. The risk of accidental signal activation is assessed based on the triggering method to obtain a first accidental activation probability parameter. The historical signal trigger records are filtered for false touches to obtain historical false touch records, and the false touch methods and number of false touches in the historical false touch records are identified. Based on the accidental touch method and the number of accidental touches, the risk of accidental signal touch is assessed to obtain a second accidental touch probability parameter; Acquire the gun-drawing signal at the law enforcement scene, and evaluate the confidence level of the gun-drawing signal based on the first accidental trigger probability parameter and the second accidental trigger probability parameter to obtain the signal confidence level. Determine whether the credibility of the signal exceeds a preset credibility threshold. If the credibility of the signal exceeds the credibility threshold, then mark the gun-drawing signal as a target gun-drawing signal.

5. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 4, characterized in that, The steps for performing real-time edge computing analysis to obtain high-confidence events based on the target draw signal and the high-level or low-level risk are as follows: Based on the high-level risk or the low-level risk, the on-site physiological data and the law enforcement recorder data at the law enforcement scene are preprocessed and packaged to generate an on-site law enforcement data package; The on-site law enforcement data package is sorted to obtain multiple on-site data information, and the confidence and urgency of each on-site data information are evaluated to obtain the data confidence and data urgency. Based on the data confidence level and the data urgency, multiple pieces of on-site data are filtered to obtain the target on-site data. The target's on-site data and the target's gun-drawing signal are merged and encapsulated to generate a high-confidence event.

6. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 5, characterized in that, After the step of sending the high-confidence event to the platform, the method further includes: A local memory is set on the smart holster, and the high-confidence event is copied and stored in the local memory; During the transmission of the high-confidence event, key information is identified from the high-confidence event to obtain key information data from the law enforcement scene; Identify the network link between the smart holster and the platform, determine the blockchain storage location based on the network link, and store the key information in the blockchain storage location.

7. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 6, characterized in that, The platform triggers a differentiated linkage strategy based on the high-confidence event, generates a target linkage strategy, and calls on IoT nodes connected to the platform for support according to the target linkage strategy. The specific steps are as follows: The platform identifies the high-confidence events, obtains the support index at the law enforcement scene, and determines whether the support index reaches a preset standard value. If it is determined that the supported index has reached the standard value, the differentiated linkage strategy is triggered, and a higher-level linkage strategy is selected in the differentiated linkage strategy. According to the high-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are filtered to obtain multiple advanced support nodes. All of the advanced support nodes are then called to the law enforcement scene. If it is determined that the support index has not reached the standard value, the differentiated linkage strategy is triggered, and a low-level linkage strategy is selected in the differentiated linkage strategy. According to the low-level linkage strategy, all IoT nodes connected to the platform are identified, and all IoT nodes are screened to obtain multiple low-level support nodes. Based on the high-confidence event, the type of support force required at the law enforcement site is obtained. The low-level support nodes are then filtered according to the type of support force to obtain the target support node, and the target support node is called to the law enforcement site.

8. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 7, characterized in that, The steps for dynamically encrypting the communication channel based on the aforementioned communication method are as follows: Based on the aforementioned communication method, the communication channel between the IoT node and the platform is identified, and the communication channel is subjected to vulnerability review to obtain channel vulnerabilities. Extract the vulnerability location of the channel vulnerability and the attack methods that can be used to break through the channel vulnerability, and set vulnerability decoys and countermeasures at the channel vulnerability location based on the vulnerability location and the attack methods; When the vulnerability decoy is triggered, the countermeasure mechanism identifies the external attack method and the external attack source, and performs countermeasure attacks based on the external attack method and the external attack source; Identify the current encryption method of the communication channel, decrypt the current encryption method, and obtain the channel encryption construction method; The channel encryption construction method is dynamically adjusted to obtain a dynamic construction method. A dynamic encryption method is generated based on the dynamic construction method, and the communication channel is updated based on the dynamic encryption method.

9. The IoT-based automatic gun-drawing behavior recognition and platform-based early warning system according to claim 8, characterized in that, The steps for randomly and dynamically modifying the security protocol based on the aforementioned security protocol are as follows: Based on the security protocol, the protocol framework of the security protocol is extracted, and multiple new security protocols are randomly generated according to the protocol framework; Based on the multiple new security protocols, a new security protocol table is generated, and the new security protocols in the table are randomly sorted to obtain a random sequence; Select the first new security protocol in the random sequence, update the security protocol, and then reorder the remaining new security protocols randomly. Within a preset time period, a time point is randomly selected as the update time point, and the security protocol is updated within the update time point.

Images