Anti-interference low-power-consumption random number generation and real-time verification method and system

Abstract

The invention relates to the field of chip-level security technology and electric digital data processing, and discloses an anti-interference low-power consumption random number generation and real-time verification method and system, and the method comprises the steps: collecting multi-source entropy data and environment monitoring data in parallel, dynamically adjusting the weight of an entropy source according to the environment data through a cooperative scheduling mechanism, and carrying out the real-time verification of the entropy source; according to the method and the system, the real-time verification result and the task priority are linked to regulate and control the power consumption of the system, and hardware verification and exception interception are completed before the random number is output, so that the generation of the random number is converted from static preset configuration into a dynamic self-reconstruction process capable of actively adapting to the change of the external environment; the core security comes from the dynamic adaptation capability of the whole multi-source system instead of the robustness of a single entropy source, the matching between the system power consumption input and the security output is realized, and the stability and the reliability of the random number in a complex embedded scene are improved.

Description

Technical Field

[0001] This invention relates to a method and system for generating and verifying low-power random numbers with anti-interference capabilities in real time, belonging to the fields of chip-level security technology and electronic digital data processing technology. Background Technology

[0002] Currently, when providing hardware security support for embedded applications such as intelligent vehicle industrial control systems, true random number generators based on physical noise sources are commonly used as the basis for encryption operations and security protocols. The basic working principle of such generators is to collect the physical noise of semiconductor devices under specific operating conditions, such as thermal noise or clock jitter, and process it into a random number sequence. The statistical randomness of its output sequence is a prerequisite for the effectiveness of the entire security system.

[0003] When the aforementioned general technical solutions are applied to actual embedded working environments, there is an inherent technical mismatch between their static design parameters and dynamic operating conditions. Specifically, the physical processes upon which the entropy source depends are sensitive to changes in environmental temperature, power supply voltage, and electromagnetic interference. Fluctuations in these external conditions directly cause changes in the output characteristics of the entropy source, thereby affecting the quality of the final generated random numbers. This degradation in the quality of random numbers caused by environmental changes is usually difficult for upper-level applications of the system to monitor directly.

[0004] To address this issue, existing solutions in this field address different aspects, but fail to form a holistic solution. For example, while integrating multiple entropy sources provides redundancy, it lacks a mechanism for dynamically adjusting the entropy source combination based on the type of external interference. Introducing real-time random number verification focuses more on intercepting substandard output data, representing a post-processing quality control approach that cannot proactively intervene in and stabilize the entropy source generation process. Furthermore, the low-power control strategies applied in the system are typically adjusted based on the system's workload, not directly related to the real-time security status of the random number generation process. Under certain operating conditions, in order to reduce power consumption... The frequency or voltage reduction measures adopted may actually decrease the entropy source's tolerance to environmental changes, leading to a mutual constraint between safety and power consumption indicators. Analysis shows that existing technologies treat entropy source generation, power consumption control, and quality verification as three independently optimizable technical steps. This separate architecture results in the following technical problems: 1. The static configuration of the entropy source makes it unable to proactively adapt to dynamic changes in the external operating environment; 2. The adjustment of power consumption mode is decoupled from the safety and quality of random numbers, lacking a basis for power consumption adjustment to ensure core security; 3. The lack of closed-loop feedback between quality verification and entropy source generation makes it impossible to intervene in the stability of the entropy source. Therefore, how to establish an internal collaborative control mechanism so that the system power consumption mode and verification process of the random number entropy source can be dynamically adjusted in an integrated manner based on real-time monitoring results of the external environment and the quality verification results of the internal random numbers becomes the technical problem to be solved by this invention. Summary of the Invention

[0005] This invention provides a method and system for anti-interference, low-power random number generation and real-time verification. Its main purpose is to solve the problem in the prior art that random number generation, power consumption control and quality verification are separated and cannot be coordinated and adjusted according to environmental and self-quality feedback.

[0006] To achieve the above objectives, the present invention provides a method for anti-interference, low-power random number generation and real-time verification, the method comprising: Step 1: Data is collected in parallel from three independent physical processes—thermal noise, clock jitter, and power supply ripple—to form a multi-source bit stream. Electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data of the chip's operating environment are acquired in an independent monitoring channel. Step 2: Within the preset control period, based on the electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data, adjust the weight coefficients of the entropy sources corresponding to each physical process in the multi-source bitstream, and generate random number fragments based on the adjusted weight coefficients. Step 3: Before the random number fragment is output to the upper layer application, a health test is performed on the random number fragment through a hardware pipeline, and a verification result is generated to indicate whether the test passed or failed. Step 4: Based on the verification results and the externally set task priorities, switch the system power consumption mode of the random number generation process and the health test process in a coordinated manner. Step 5, and when the verification result indicates that the health test fails, execute the preset abnormal response sequence. The abnormal response sequence includes: pausing the generation of multi-source bitstreams, reconstructing the weight coefficients based on the verification result and monitoring data, and using the reconstructed weight coefficients to generate new random number fragments to perform re-verification.

[0007] Preferably, data is collected in parallel from three independent physical processes: thermal noise, clock jitter, and power supply ripple. Specifically, thermal noise is collected in a metal-oxide-semiconductor tube operating in the subthreshold region; clock signal jitter is collected through a cross-coupling structure; and the ripple signal of the system power supply is collected. The collected thermal noise, clock signal jitter, and ripple signal are then processed into three independent raw bit streams. The method is characterized in that, before step 2, it further includes: performing min-max normalization on the collected electromagnetic interference intensity data, temperature data, and fluctuation data, linearly scaling them to a preset interval of [0, 1] to form a standardized environmental state vector, and using the environmental state vector to adjust the weighting coefficients.

[0008] Preferably, the weighting coefficients are adjusted based on the monitoring data, including: comparing the electromagnetic interference intensity data, temperature data, and fluctuation data with their respective preset interference thresholds; when any monitoring data exceeds its corresponding interference threshold, it is determined that there is environmental interference, and for this environmental interference, the weighting coefficients corresponding to the entropy sources that are sensitive to the interference are selectively reduced, while the weighting coefficients corresponding to the entropy sources that are not sensitive to the interference are increased.

[0009] Preferably, the adjustment of the weighting coefficients is performed according to the following rules: ,in, For the first The weight coefficients updated for each entropy source; For the first The current weight coefficients of each entropy source; For the first Current values ​​of various monitoring data; In order to be with the first The preset interference threshold corresponding to the monitoring data; It is a symbolic function; Adjust the step size factor for the preset global weights; For the preset representation of the first The entropy source pairs with the first... Sensitivity coefficients for different types of interference.

[0010] Preferably, the system power consumption mode is switched in a coordinated manner, including: when the verification results of a preset number of consecutive tests indicate that the health test has passed, and the electromagnetic interference intensity data, temperature data and fluctuation data are all within their respective preset safe ranges, the random number generation process and the health test process are switched to the preset low power consumption working mode simultaneously.

[0011] Preferably, health testing is performed via a hardware pipeline, including: inputting a random number fragment into a dedicated hardware verification circuit; the hardware verification circuit runs synchronously with the random number fragment generation process, and performs at least one online statistical test within a preset time window after the random number fragment is generated to generate a verification result; when the verification result indicates that the health test fails, the hardware verification circuit drives an exception interception unit to cut off the output path of the random number fragment before it is called by the upper-layer application.

[0012] Preferably, the reconstructing of weight coefficients includes: establishing a preset fault rule base, which defines the correlation between different health test failure modes and the exceeding of limits of specific environmental interference data; when a health test fails, querying the fault rule base using the current health test failure mode and environmental monitoring data to identify the entropy source of performance degradation; adjusting the weight coefficients corresponding to the identified entropy sources of performance degradation to 0, and normalizing and increasing the weight coefficients of the remaining entropy sources according to preset rules.

[0013] Preferably, the low-power operating mode includes: simultaneously reducing the sampling clock frequency used for random number generation and the computation clock frequency used for health testing; and performing power gating operations on inactive hardware logic units during the random number generation process and the health testing process.

[0014] Preferably, the method further includes: recording the execution process of each adjustment action of the weight coefficient, each verification result, each switching action of the system power consumption mode, and each abnormal response sequence, forming an operation log containing timestamps for system status traceability and security auditing.

[0015] A system for interference-resistant, low-power random number generation and real-time verification, the system comprising: A multi-source dynamic entropy generation module is configured to collect data in parallel from three independent physical processes—thermal noise, clock jitter, and power supply ripple—to form a multi-source bit stream, and generate random number fragments based on weight coefficient adjustment instructions issued by a cooperative scheduling module. An environmental interference monitoring module is configured to acquire electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data of the chip's operating environment, and send the acquired data to the collaborative scheduling module; A real-time verification module is configured to perform health tests on random number fragments generated by the multi-source dynamic entropy generation module through a hardware pipeline, generate verification results, and send them to the collaborative scheduling module. The collaborative scheduling module is configured to: receive data sent by the environmental interference monitoring module and verification results sent by the real-time verification module; generate and send weight coefficient adjustment instructions to the multi-source dynamic entropy generation module based on the data sent by the environmental interference monitoring module; generate and send system power consumption mode switching instructions based on the verification results sent by the real-time verification module and the externally set task priorities; and execute a preset abnormal response sequence when the verification result indicates that the health test has failed. This sequence includes instructing the multi-source dynamic entropy generation module to pause generation and reconstruct the weight coefficients based on the verification results and monitoring data.

[0016] Compared with the prior art, the beneficial effects of the present invention are: 1. A dynamic response path between the composition of random number entropy sources and external environmental interference was established. The standardized monitoring results continuously collected by the environmental interference monitoring module are not used to trigger a general protection state, but are used as direct input of control commands by the collaborative scheduling module to adjust the contribution ratio of different types of entropy sources within the multi-source dynamic entropy generation module in real time. This transforms the core of the random number generation process, namely the composition of the entropy source, from a static, preset anti-interference configuration to a continuous self-reconstruction process that can actively track and adapt to environmental changes. As a result, when faced with complex electromagnetic temperature or voltage fluctuations, the output stability of the entropy source no longer depends solely on a single entropy source, but rather on the dynamic adaptability of the entire multi-source system.

[0017] 2. The real-time verification results of random numbers are directly linked to the power consumption mode switching of the system, forming a power consumption control closed loop based on output quality. The collaborative scheduling module receives the verification results from the real-time verification module. These results characterize the security level of the generated random numbers and, together with the priority of the upper-layer security tasks, serve as the basis for judging the switching of the system's power consumption mode. This approach avoids the inherent limitations of traditional power management, which only switches modes based on task load and ignores the stability of the security foundation. When the quality of random numbers is stable, the system can enter a lower power consumption state during the execution of regular tasks. However, once the verification results deteriorate, the system will be driven to switch to a high-security mode to strengthen the entropy source configuration, thus achieving a match between system power consumption input and safe random number output.

[0018] 3. Through the collaborative scheduling module, the external perception of environmental interference and the internal verification of random number quality are integrated, forming a dual-based entropy source reconstruction and anomaly response mechanism. When the real-time verification module reports an anomaly in random numbers, the collaborative scheduling module not only pauses generation and intercepts output, but also simultaneously retrieves the monitoring results from the environmental interference monitoring module at that point in time for correlation analysis. This helps determine the possible external causes of the anomaly and allows for more targeted adjustment of the entropy source weights of the multi-source dynamic entropy generation module, completing the entropy source reconstruction. This process couples the previously isolated environmental monitoring and quality verification stages into a unified fault handling process with preliminary diagnostic capabilities, improving the speed and accuracy of the system's recovery of core safety functions when encountering unknown interference. Furthermore, by utilizing hardware-based verification circuits and a parallel pipeline architecture, the generation and verification of random numbers are synchronized on the timeline. Each time the multi-source dynamic entropy generation module generates a random number fragment of a preset length, it is sent to the real-time verification module for synchronous processing. The verification result is then fed back to the collaborative scheduling module. This compact process of generation and verification transforms the confirmation of random number quality from the traditional offline, batch post-verification method into an inherent node embedded in the generation process. In this way, any abnormal random numbers caused by momentary interference can be detected and intercepted before they are called by upper-layer applications, thereby eliminating the time window for abnormal data to flow into the encryption task. This ensures that the security of random number usage no longer depends on subsequent software verification, but is guaranteed by the generation process itself. Attached Figure Description

[0019] Figure 1 This is a flowchart of the closed-loop adaptive random number generation and verification process of the present invention; Figure 2 This is a response curve of the entropy source weight coefficient of the present invention dynamically changing with environmental disturbances; Figure 3 This is a diagram illustrating the implementation architecture of the system of the present invention as an embedded hardware security module (HSM). Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the present invention clearer, the technical solutions of the present invention will be described in detail below. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The present invention discloses a method and system for anti-interference, low-power random number generation and real-time verification, which is configured as a closed-loop dynamic adaptive data processing system. It mainly consists of a multi-source dynamic entropy generation module, an environmental interference monitoring module, a real-time verification module, and a collaborative scheduling module. The environmental interference monitoring module is responsible for collecting external physical operating environment data and transmitting the quantized data to the collaborative scheduling module. Based on the received environmental data and internal quality feedback data from the real-time verification module, the collaborative scheduling module generates control commands to regulate the entropy source composition of the multi-source dynamic entropy generation module and the power consumption mode of the entire system. The multi-source dynamic entropy generation module generates random number data according to the control commands and submits it to the real-time verification module for synchronous quality verification. The real-time verification module then feeds back the verification results to the collaborative scheduling module, thereby completing the closure of the entire control loop.

[0022] In one specific implementation, for example, in a scenario where this system is deployed in an onboard hardware security module (HSM) chip, to cope with the complex electromagnetic environment, temperature changes, and power supply fluctuations during vehicle operation, the method of this invention first performs parallelized multidimensional data acquisition. This acquisition process is physically divided into an entropy source data channel and an environmental monitoring data channel. In the entropy source data channel, the multi-source dynamic entropy generation module uses three independent physical noise sources for data acquisition. Specifically, it acquires thermal noise signals through a metal-oxide-semiconductor (MOS) transistor operating in the subthreshold region, acquires jitter of the system clock signal using a cross-coupled inverter structure, and acquires the ripple signal of the system power supply. These three analog signals are converted into three independent raw ratio signals through their respective independent signal processing links. Special flow; Meanwhile, in the environmental monitoring data channel, the environmental interference monitoring module uses the on-chip integrated sensor unit to synchronously acquire real-time data of the chip's operating environment, including acquiring electromagnetic interference intensity data through a broadband antenna and detection circuit, acquiring chip temperature data through the temperature characteristics of the PN junction, and acquiring system power supply voltage fluctuation data through an analog-to-digital converter (ADC). This parallel acquisition mechanism provides data input for subsequent adaptive control; Considering that the electromagnetic interference intensity, temperature, and voltage fluctuation data acquired by the environmental interference monitoring module have different physical units and numerical ranges, the collaborative scheduling module performs min-max normalization on these raw monitoring data before weight adjustment, linearly scaling them to a preset [0, 1] interval to form a standardized environmental state vector; For any monitoring data raw measurement value Its normalized value pass Perform the calculation, where, and It is aimed at the first The monitoring data has a preset lower and upper limit of normal range. These two parameter values ​​are obtained by calibrating the reference environment during the chip factory testing phase and stored in the system's non-volatile memory. After this processing, the physical environment data of different dimensions are converted into a dimensionless environmental state vector for subsequent weight adjustment calculations.

[0023] To transform the environmental state vector into control over entropy sources, the collaborative scheduling module employs a set of weight adjustment procedures to regulate the contribution ratio of each entropy source in the multi-source bitstream. This procedure adjusts the weights of entropy sources sensitive to interference based on whether each monitored data point exceeds its corresponding interference threshold; the formula for weight updates is configured as follows: ;in, For the first The weight coefficients updated for each entropy source; For the first The current weight coefficients of each entropy source; The global weight adjustment step size factor is set to 0.1 to control the rate of weight adjustment. For the first The current value of the monitoring data after standardization; Is with the first The preset interference thresholds corresponding to various monitoring data, for example, the interference threshold for temperature can be set to 0.85; It is a symbolic function; To characterize the first The entropy source pairs with the first... The sensitivity coefficients for different types of interference are stored in a 3x3 matrix. For example, if a clock jitter entropy source is sensitive to electromagnetic interference, then its corresponding... The value is set to a positive value (e.g., 0.9), and its sensitivity to temperature changes is low, corresponding to... The value is then set to a negative value or close to 0; when an environmental disturbance, such as electromagnetic interference intensity... Exceeding its threshold hour, The output is 1. At this point, the weights of entropy sources sensitive to electromagnetic interference will be reduced, while the weights of entropy sources insensitive to it will be increased accordingly. The multi-source dynamic entropy generation module receives the updated weight coefficients... Then, a weighted mixer is used to combine the three original bit streams according to the new weight ratio to generate a random number segment with a preset length of 1024 bits.

[0024] To verify the quality of random number fragments before they are output to upper-layer applications, the system is configured with a real-time verification module that runs synchronously with the generation process. This module contains a dedicated hardware verification circuit and adopts a pipelined architecture design. Specifically, when the multi-source dynamic entropy generation module generates a 1024-bit random number fragment, the data fragment is simultaneously transmitted to the real-time verification module. The hardware verification circuit immediately starts a fixed online statistical test algorithm, such as performing a frequency test in the National Institute of Standards and Technology (NIST) SP800-22 test set. The decision logic of this test is implemented by hardware logic gates, which can complete the calculation within a preset time window and output a verification result indicating whether the test passes or fails. If the verification result is pass, the random number fragment can be called by the upper-layer application. If the verification result indicates that the test fails, the hardware verification circuit drives an anomaly interception unit. This unit physically acts as a gate on the data path. After receiving the interception signal, it cuts off the output path of the unqualified random number fragment and simultaneously feeds back the failed verification result to the cooperative scheduling module.

[0025] After receiving the verification result from the real-time verification module, the collaborative scheduling module, in conjunction with the externally set task priority, will execute system power mode switching or anomaly response in a coordinated manner. When N consecutive verification results (e.g., N=100) indicate that the test has passed, and environmental monitoring data shows that the current environment is within a safe range, and the task priority is normal or low, the collaborative scheduling module issues an instruction to synchronously reduce the sampling clock frequency of the multi-source dynamic entropy generation module and the operation clock frequency of the real-time verification module, and performs power gating operations on the inactive hardware logic units in both modules, so that the system enters a low-power operating mode. When the verification result indicates that the test has failed, an anomaly response sequence is triggered. This sequence first instructs the multi-source dynamic entropy generation module to pause bit stream generation. Subsequently, the collaborative scheduling module retrieves the current health test failure mode and environmental monitoring data, and queries a preset fault rule base. The library predefines the correlation between different test failure modes and exceeding the limits of specific environmental interference, which is established through previous chip-level fault injection tests. Based on the query results, the collaborative scheduling module identifies the entropy sources of performance degradation and performs weight coefficient reconstruction. For example, it adjusts the weight coefficients corresponding to the identified entropy sources to 0 and distributes their original weights proportionally to the remaining entropy sources. After generating new random number fragments using the reconstructed weight coefficients, it performs a second verification until the verification passes and normal output is restored. To ensure the traceability of the system's operating status, the collaborative scheduling module is also configured to record every critical operation it performs, including weight coefficient adjustments, health test verification results, system power consumption mode switching, and the execution process of abnormal response sequences. All these records are timestamped to form an operation log and are stored in a dedicated on-chip non-volatile memory for system status traceability and security auditing.

[0026] Example 1: In a continuously operating automotive chassis welding production line, a multi-axis robotic arm performs high-precision spot welding. The robot control command issuance and execution process is verified in real time through an encryption authentication mechanism based on hardware random number generation. When the robotic arm moves from the standby position to the welding station, its built-in hardware security module, which adopts the solution of this invention, is in normal power consumption mode. At this time, the electromagnetic interference intensity, temperature, and power supply voltage data collected by the environmental interference monitoring module are all within the preset safety range. Based on this environmental state, the collaborative scheduling module provides thermal noise entropy sources and time signals within the multi-source dynamic entropy generation module. The initial weighting coefficients {0.33, 0.34, 0.33} for the balanced allocation of clock jitter entropy source and power supply ripple entropy source are used. The multi-source dynamic entropy generation module continuously generates random number fragments and provides random numbers for the communication link authentication protocol in standby state through the health test of the real-time verification module. After the welding command is issued, the arc welding gun of the robot arm is activated, generating a high-intensity broadband electromagnetic field in the welding area. At the same time, the large current extraction causes a 150mV voltage drop in the system power supply bus. The environmental interference monitoring module processes the two monitoring data exceeding the preset interference threshold into standardized environmental state vectors and transmits them to the collaborative scheduling module in parallel.

[0027] Upon receiving the environmental state vector, the collaborative scheduling module immediately performs an entropy source weight reconstruction based on that specific vector, according to a preset sensitivity coefficient matrix. Based on the parameters, the collaborative scheduling module identifies the high sensitivity of clock jitter entropy sources to electromagnetic interference, as well as the direct correlation between power supply ripple entropy sources and voltage drops, and updates the calculation formula in real time through weights. The weighting coefficients of clock jitter entropy sources and power supply ripple entropy sources are lowered, while the weighting coefficients of thermal noise entropy sources, which are insensitive to these two types of interference, are raised. The updated weight vector is adjusted to {0.7, 0.15, 0.15}. This reconstruction process is completed within several clock cycles, ensuring that the entropy source composition of the multi-source dynamic entropy generation module is proactively adapted to the current operating conditions before the welding operation begins. Throughout the welding operation, the multi-source dynamic entropy generation module continuously generates random number fragments based on the adjusted weighting coefficients. Each fragment flows through the hardware pipeline of the real-time verification module for synchronous health testing before being output to the upper-layer encryption application. At this time, the verification results of the real-time verification module, along with the high-priority task status transmitted by the upper-layer application, are synchronized. Both serve as inputs for power consumption mode scheduling. Given that welding is a high-priority task, the collaborative scheduling module locks the system in a high-safety power consumption mode. Even if multiple consecutive verification results indicate that the health test has passed, the system maintains this power consumption mode until the high-priority task status is lifted. After the welding operation is completed, the arc is extinguished, and the environmental interference monitoring module reports that electromagnetic interference and voltage fluctuations have returned to a safe range. After confirming that the priority of subsequent tasks is normal and that the real-time verification module continues to provide feedback on the test results, the collaborative scheduling module generates and issues a power command to switch the system to a low-power operating mode. In this mode, the sampling clock frequency for random number generation and the operation clock frequency for health testing are both reduced until the next high-priority task is triggered.

[0028] Example 2: This example verifies the stability of random number generation and the effectiveness of power consumption control of the technical solution of this invention under different environmental interferences through a hardware test environment. The test platform consists of a system-on-a-chip (SoC) that implements the complete technical solution of this invention as the sample group, and a chip containing only a static multi-source entropy generation module (with fixed weights for each entropy source and no environmental feedback adjustment) as the control group. The external environment simulation is carried out through an electromagnetic compatibility test chamber, a high and low temperature chamber, and a programmable DC power supply. Their functional specifications are as follows: the electromagnetic compatibility test chamber can generate electromagnetic interference with a frequency range of 1MHz to 1GHz and a field strength of 0V / m to 20V / m; the temperature control range of the high and low temperature chamber is -50°C. Up to 150 The accuracy is ±0.5 The programmable DC power supply has an output voltage range of 0V to 5V and a ripple factor of less than 0.1%. Data acquisition and analysis are performed using a digital oscilloscope and a computer running the NIST SP800-22 and SP800-90B standard test suites.

[0029] During the experiment, the sample group of the present invention and the control group were first placed under the baseline operating conditions (25). The system was operated at 3.3V, 0V / m field strength, and 1Gbit random data was collected and average power consumption was measured. Subsequently, two sample groups were subjected to a single type of environmental stress, including high-intensity electromagnetic interference (400MHz, 10V / m) and high temperature (125℃). ), low temperature (-40) The data acquisition and power consumption measurement were repeated under each operating condition, including the supply voltage drop (2.9V). Finally, a high temperature (125℃) was applied. The test was conducted again under the combined pressure conditions of high-intensity electromagnetic interference (400MHz, 10V / m). For the sample group of this invention, the internal entropy source weight vector and the number of triggers of the abnormal interception unit reported in real time were also recorded synchronously. The core measurement indicators of the test were the pass rate of the random number sequence through the NISTSP800-22 test set, the average power consumption of the system, and the number of abnormal interceptions. Table 1 shows the performance data collected under different test conditions in this embodiment.

[0030] Table 1: Performance Data Comparison Table.

[0031] ; Referring to Table 1, the experimental data shows that under the baseline operating conditions, the NIST test pass rates of both sample groups were at a high level. The sample group of this invention entered a low-power operating mode because its collaborative scheduling module determined that the current environment and task priority were both low, with an average power consumption of 3.5mW, while the control group's power consumption was 8.0mW. When high-intensity electromagnetic interference was applied, the test pass rate of the control group dropped to 63.5%. After the environmental interference monitoring module of the sample group of this invention sensed the interference, its internal entropy source weight vector changed from {0.33, 0.34, 0.33}. The weights were dynamically adjusted to {0.70, 0.10, 0.20}, which reduced the weight of the clock jitter entropy source, which is sensitive to electromagnetic interference, and increased the weight of the thermal noise entropy source. This entropy source reconstruction ensured that the NIST test pass rate of the final output random number sequence remained at 100%. Under the condition of instantaneous pulse interference, the number of abnormal interceptions of the sample group of this invention was recorded as 3, indicating that its real-time verification module had captured and intercepted 3 random number segments that failed the health test through hardware pipeline before responding to the dynamic weight adjustment, thus preventing the output of unqualified data.

[0032] Example 3: This example combines Figures 1 to 3 This document describes a method and system for generating and verifying low-power, interference-resistant random numbers in real time. Figure 1As shown, the process begins with the parallel input of raw entropy signals and raw monitoring signals from two sources: the physical entropy source and the external physical environment, to the multi-dimensional data acquisition module. This module processes these signals and outputs raw bitstreams and raw monitoring data, respectively. The raw bitstream is input to the random number fragment generation stage, which generates random number fragments based on weight coefficients provided by the dynamic decision-making and scheduling module. Simultaneously, the raw monitoring data is sent to the normalized environmental data stage for processing into an environmental state vector, which is then transmitted to the dynamic decision-making and scheduling module. This scheduling module integrates the environmental state vector, verification results from health testing, task priorities from upper-layer applications, and thresholds / coefficients read from the calibration parameter library. It also obtains association rules by issuing fault mode queries to the fault rule library and makes decisions accordingly. The decision-making process is recorded in the operation log. The generated random number fragments then enter the health testing stage, producing qualified random numbers that are provided to the upper-layer application. Simultaneously, verification results are generated and fed back to the dynamic decision-making and scheduling module, thus forming a complete closed-loop adaptive control system.

[0033] like Figure 2 As shown, the horizontal axis represents the test conditions, which are, in order, the baseline condition, high electromagnetic interference, high temperature, low temperature, voltage drop, combined pressure, and instantaneous pulse. The vertical axis represents the weighting coefficients. The three curves in the figure represent the thermal noise entropy source, the clock jitter entropy source, and the power supply ripple entropy source, respectively. The data shows that the weights of the three are balanced under the baseline condition. However, under the high electromagnetic interference condition, the weighting coefficient of the thermal noise entropy source is increased to 0.7, while the weight of the clock jitter entropy source is suppressed to 0.1. Under the low temperature condition, the weight of the clock jitter entropy source is increased to 0.6, and under the combined pressure condition, the weight of the thermal noise entropy source is increased to 0.8. This intuitively presents the process of adaptive reconstruction of the entropy source composition based on the external environment.

[0034] like Figure 3 As shown in the figure, the system architecture depicted in the figure takes a hardware security module (HSM) physical entity as its main body. This module contains a system-on-a-chip (SoC) as its core. The SoC integrates functional circuit units that implement the method, including a cooperative scheduling module firmware, a multi-source dynamic entropy generation circuit, an environmental interference monitoring circuit, a real-time pipeline verification circuit, an anomaly interception and gating unit, and a built-in pseudo-random number generator. It also embeds the configuration and data required to support its operation, including operation log files, fault rule base, calibration parameter set interference threshold, sensitivity coefficient, and safe startup configuration. This hardware security module communicates with external entities through device interface buses such as CAN / SPI / I2C. On the one hand, it provides qualified random numbers to embedded application hosts such as automotive ECUs and robot controllers and receives their task priority instructions. On the other hand, it interacts with remote security audit terminals for traceability and auditing, and executes log retrieval and audit instructions, thus forming a complete and deployable hardware security solution.

[0035] Example 4: This example discloses a standardized engineering calibration procedure for determining key control parameters in the system of the present invention. This procedure is executed during the engineering verification phase of the system-on-a-chip (SoC) and aims to provide a set of control parameters that match the actual physical characteristics of the chip batch. The initial state of the procedure is defined as follows: a SoC device integrating the system of the present invention is placed on a test platform. This platform integrates an electromagnetic compatibility test chamber, a high and low temperature chamber, and a programmable DC power supply with corresponding functional specifications. The SoC's collaborative scheduling module is set to a calibration mode accessible to external test equipment. In this mode, the working state of any one or more entropy sources can be forcibly locked, and the internal environmental state vector, entropy source weight vector, and raw health test statistics output by the real-time verification module can be read back in real time. The procedure first sets the interference threshold corresponding to the environmental interference. The sensitivity coefficients of each entropy source to this interference Calibration is performed; taking the calibration of electromagnetic interference (EMI) related parameters as an example, the test procedure is as follows: under the reference operating condition (25... At 3.3V, the thermal noise entropy source and power supply ripple entropy source are forcibly disabled, so that the multi-source dynamic entropy generation module relies solely on the clock jitter entropy source to generate random number sequences. The electromagnetic compatibility test chamber is controlled to increase the electromagnetic field strength of 400MHz from 0V / m in increments of 0.5V / m. At each field strength level, 1Gbit of random number data is collected, and an external computer performs the full NIST SP800-22 test suite. When any item in the NIST test suite fails for the first time, the field strength value at that moment is recorded. In a specific test sequence, when the field strength reaches 12.5V / m, the random walk test fails for the first time; this field strength value is recorded as the physical failure point of the clock jitter entropy source for 400MHz electromagnetic interference. Subsequently, the above process is repeated to obtain the physical failure points of the thermal noise entropy source and the power supply ripple entropy source under the same interference (18.0V / m and 19.5V / m, respectively). The system's interference threshold... This refers to the minimum-maximum normalized value corresponding to the lowest physical failure point (12.5V / m) among all entropy sources; the sensitivity coefficient of each entropy source. Then, quantification is performed based on the difference between each of its physical failure points and the threshold point. For the entropy source that is least sensitive to interference, its... It is set to a large negative value, while the entropy source most sensitive to disturbances... It is then set to a large positive value.

[0036] After all parameters are calibrated, the procedure proceeds to the step of building a fault rule base. This process utilizes the data accumulated in the previous steps and, through data analysis software, seeks strong correlations between specific health test failure modes and specific environmental state vectors. By performing statistical regression analysis on thousands of failure samples collected under boundary conditions, a probability mapping table is established: (health test failure signature, environmental state vector) → suspected failure entropy source. The analysis process establishes a rule that a linear complexity test fails if and only if the component representing voltage fluctuations in the synchronously recorded environmental state vector exceeds its threshold. At that time, the probability of power supply ripple entropy source degradation is 95%. This statistical regularity is solidified into a rule in the fault rule base. Ultimately, the entire fault rule base consists of dozens of such association rules with pre-set confidence levels obtained through data mining. After completing all the above steps, the calibrated interference threshold is determined. Sensitivity coefficient matrix The completed fault rule base is burned as a configuration file into the on-chip non-volatile memory of the batch of SoC devices, enabling the SoC devices to adaptively adjust to their own physical characteristics.

[0037] Example 5: In a hardware security module deployed in a remote unattended base station, to meet the security requirements of compatibility with future-enabled post-quantum cryptography algorithms, the system of this invention is configured to execute periodic online self-testing and parameter adaptation procedures. This procedure, as a background task automatically triggered by the collaborative scheduling module during periods of low system load, fine-tunes the control parameters obtained through offline calibration based on the real-time hardware status. During the execution of this procedure, the collaborative scheduling module first retrieves historical environmental data from the operation log to determine the typical temperature and voltage fluctuation range at the device's location. Subsequently, without affecting upper-layer applications, it instructs the multi-source dynamic entropy generation module to sequentially perform small-scale stress tests on each entropy source. This test monitors the entropy source's activity as it approaches its interference threshold. Under normal operating conditions, the slope of the statistical characteristic change obtained from the entropy source output data after processing by the real-time verification module is compared with the initial baseline slope stored in the non-volatile memory during chip factory calibration. If the collaborative scheduling module calculates that the deviation between the current slope and the baseline slope exceeds a preset drift threshold, the system adjusts the associated sensitivity coefficient based on this deviation. and interference threshold Perform an iterative correction and record the correction action and related data in the operation log.

[0038] In this embodiment, when the hardware security module underwent a remote security audit, the auditor requested to trace a system anomaly that occurred at a specific point in the past, suspected to be caused by external instantaneous strong pulse interference. At this time, the system maintenance personnel retrieved the relevant operation logs before and after that point in time from the non-volatile memory of the hardware security module through an authorized encrypted channel. The logs, indexed by timestamps, clearly showed that at the exact moment the suspected anomaly occurred, the environmental interference monitoring module recorded an electromagnetic interference signal exceeding the normal threshold for a duration of 500 milliseconds. Immediately afterward, the collaborative scheduling module performed an emergency adjustment of the entropy source weight coefficient, reducing the weight of the clock cross-coupling jitter entropy source from 0.4 to 0.1. At the same time, the verification result fed back by the real-time verification module within this time window was always "pass," and the number of triggers of the anomaly interception unit was 0. After the interference disappeared, the log showed that the entropy source weight coefficient automatically recovered to the normal level within several control cycles. Through this immutable operation log containing a complete causal chain, the system not only proved that it did not generate any unqualified random number outputs, but also provided conclusive evidence of its adaptive response behavior under interference.

[0039] Example 6: The power consumption mode and task priority of the system of the present invention are managed by a 32-bit task priority and power consumption strategy control register set inside a collaborative scheduling module and readable and writable by the upper-layer application software; the bits [1:0] of the register are used to set the task priority, and its encoding is defined as binary 00 corresponding to low priority, 01 corresponding to normal priority, 10 corresponding to high priority, and 11 corresponding to emergency priority. Bit [2] is a power consumption mode lock flag. When this bit is set to 1, the system is forced to lock in the high-security power consumption mode; the collaborative scheduling module reads the current value of the register within a fixed control cycle and executes the power consumption mode decision logic in combination with the health test results from the real-time verification module. Only when the task priority is set to normal or low, the power consumption mode lock flag is 0, and the health test results of N consecutive times (N is a configurable parameter, and its default value is 100) are all passed, will the collaborative scheduling module generate and issue an instruction to enter the low-power working mode.

[0040] Upon receiving a reset signal, the system executes a power-on self-test and initialization sequence. In this sequence, the collaborative scheduling module first enters a preset safe startup state, and the multi-source dynamic entropy generation module is configured to a fixed working mode with preset parameters, that is, the weight coefficients of the three types of entropy sources are fixed to be equal and it runs at the highest sampling frequency. The initial random numbers generated in this mode are used partly for functional self-testing of the hardware logic of the real-time verification module, and partly as a seed input to a built-in pseudo-random number generator (PRNG) to provide random numbers for the early initialization of the operating system's bootloader and other core processes. This pseudo-random number generator complies with the relevant national cryptographic management standards. After the entire power-on self-test sequence passes and the environmental interference monitoring module confirms that all environmental parameters have entered a stable state, the collaborative scheduling module switches to the regular dynamic adaptive working mode and begins to execute closed-loop feedback adjustment based on the environment and verification results.

[0041] To further verify the superiority of the dynamic adaptive adjustment mechanism of the present invention, which is based on the collaborative scheduling module, over the prior art, the following comparative examples are provided.

[0042] Comparative Example 1: This comparative example aims to simulate a technical solution that integrates multiple entropy sources but lacks a dynamic adjustment mechanism. Therefore, a system-on-a-chip (SoC) hardware and external environment simulation test platform identical to that of Example 2 is used, but the internal collaborative scheduling module function is modified: its function of dynamically adjusting and reconstructing entropy source weights based on environmental data and verification results is forcibly disabled. Furthermore, the weight coefficients of the three entropy sources—thermal noise, clock jitter, and power ripple—in the multi-source dynamic entropy generation module are fixed to a balanced static configuration, i.e., {0.33, 0.34, 0.33}. Except for this single and essential difference, all other test conditions, data acquisition procedures, and performance evaluation standards are strictly consistent with Example 2. During the experiment, the statically configured test group is first placed under the baseline condition (25...). It operates at 3.3V, 0V / m field strength, and is subsequently subjected to high-intensity electromagnetic interference (400MHz, 10V / m) and high temperature (125℃) in sequence. ), low temperature (-40) ), supply voltage drop (2.9V) and combined stress (125) The test was conducted under five environmental pressure conditions: high temperature and 400MHz, 10V / m high-intensity electromagnetic interference. Under each condition, 1Gbit of random number data was collected, and its statistical randomness was evaluated. The core measurement index was the pass rate of the random number sequence through the NISTSP800-22 test set. The test results are recorded in Table 2.

[0043] Table 2: Performance data of static weight configuration scheme under different test conditions.

[0044] ; Experimental results show that, in the absence of a closed-loop feedback mechanism that dynamically adjusts the entropy source weights based on environmental monitoring data, the quality of the random number sequence generated by this system is sensitive to changes in the external environment. Especially under high-intensity electromagnetic interference and combined stress conditions, because the system cannot actively reduce the contribution of the significantly disturbed clock jitter entropy source, the pass rate of its output random number sequence in the NIST test has deteriorated to an unacceptable level. This directly proves that this static configuration scheme has serious safety and reliability defects in complex embedded working environments.

[0045] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0046] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for generating and verifying low-power random numbers with anti-interference capabilities in real time, characterized in that, The method includes: Step 1: Data is collected in parallel from three independent physical processes—thermal noise, clock jitter, and power supply ripple—to form a multi-source bit stream. Electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data of the chip's operating environment are acquired in an independent monitoring channel. Step 2: Within the preset control period, based on the electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data, adjust the weight coefficients of the entropy sources corresponding to each physical process in the multi-source bitstream, and generate random number fragments based on the adjusted weight coefficients. Step 3: Before the random number fragment is output to the upper layer application, a health test is performed on the random number fragment through a hardware pipeline, and a verification result is generated to indicate whether the test passed or failed. Step 4: Based on the verification results and the externally set task priorities, switch the system power consumption mode of the random number generation process and the health test process in a coordinated manner. Step 5, and when the verification result indicates that the health test fails, execute the preset abnormal response sequence. The abnormal response sequence includes: pausing the generation of multi-source bitstreams, reconstructing the weight coefficients based on the verification result and monitoring data, and using the reconstructed weight coefficients to generate new random number fragments to perform re-verification.

2. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 1, characterized in that, Data is collected in parallel from three independent physical processes: thermal noise, clock jitter, and power supply ripple. Specifically, thermal noise is collected in a metal-oxide-semiconductor tube operating in the subthreshold region; clock signal jitter is collected through a cross-coupling structure; and the ripple signal of the system power supply is collected. The collected thermal noise, clock signal jitter, and ripple signals are then processed into three independent raw bit streams. The method is characterized by the following step before step 2: performing min-max normalization on the collected electromagnetic interference intensity data, temperature data, and fluctuation data, linearly scaling them to a preset interval of [0, 1], and adjusting the weighting coefficients using the environmental state vector.

3. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 1, characterized in that, Based on the monitoring data, the weighting coefficients are adjusted, including comparing the electromagnetic interference intensity data, temperature data, and fluctuation data with their respective preset interference thresholds. When any monitoring data exceeds its corresponding interference threshold, it is determined that there is environmental interference. For this environmental interference, the weighting coefficients corresponding to entropy sources that are sensitive to the interference are selectively reduced, while the weighting coefficients corresponding to entropy sources that are not sensitive to the interference are increased.

4. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 3, characterized in that, The adjustment of the weighting coefficients is performed according to the following rules: ,in, For the first The weight coefficients updated for each entropy source; For the first The current weight coefficients of each entropy source; For the first Current values ​​of various monitoring data; In order to be with the first The preset interference threshold corresponding to the monitoring data; It is a symbolic function; Adjust the step size factor for the preset global weights; For the preset representation of the first The entropy source pairs with the first... Sensitivity coefficients for different types of interference.

5. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 1, characterized in that, The system power consumption mode is switched in a coordinated manner, including: when the verification results of a preset number of consecutive tests indicate that the health test has passed, and the electromagnetic interference intensity data, temperature data and fluctuation data are all within their respective preset safe ranges, the random number generation process and the health test process are switched to the preset low power consumption working mode simultaneously.

6. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 1, characterized in that, Health testing is performed via a hardware pipeline, including: inputting a random number fragment into a dedicated hardware verification circuit; the hardware verification circuit runs synchronously with the random number fragment generation process and performs at least one online statistical test within a preset time window after the random number fragment is generated to generate a verification result; when the verification result indicates that the health test fails, the hardware verification circuit drives an exception interception unit to cut off the output path of the random number fragment before it is called by the upper-layer application.

7. The method for generating and verifying low-power random numbers with anti-interference capabilities according to claim 1, characterized in that, The refactoring of weight coefficients includes: establishing a preset fault rule base, which defines the correlation between different health test failure modes and exceeding limits of specific environmental interference data; when a health test fails, querying the fault rule base using the current health test failure mode and environmental monitoring data to identify the entropy source of performance degradation; adjusting the weight coefficients corresponding to the identified entropy sources of performance degradation to 0, and normalizing and increasing the weight coefficients of the remaining entropy sources according to preset rules.

8. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 6, characterized in that, The low-power operating modes include: simultaneously reducing the sampling clock frequency used for random number generation and the computation clock frequency used for health testing; and performing power gating operations on inactive hardware logic units during the random number generation process and the health testing process.

9. The method for generating and verifying low-power random numbers with anti-interference capabilities in real time according to claim 1, characterized in that, The method also includes: recording the execution process of each adjustment of the weight coefficient, each verification result, each switching action of the system power consumption mode, and each abnormal response sequence to form an operation log with timestamps for system status traceability and security auditing.

10. A system for anti-interference, low-power random number generation and real-time verification, used to implement the method of any one of claims 1 to 10, characterized in that the system... include: A multi-source dynamic entropy generation module is configured to collect data in parallel from three independent physical processes—thermal noise, clock jitter, and power supply ripple—to form a multi-source bit stream, and generate random number fragments based on weight coefficient adjustment instructions issued by a cooperative scheduling module. An environmental interference monitoring module is configured to acquire electromagnetic interference intensity data, temperature data, and system power supply voltage fluctuation data of the chip's operating environment, and send the acquired data to the collaborative scheduling module; A real-time verification module is configured to perform health tests on random number fragments generated by the multi-source dynamic entropy generation module through a hardware pipeline, generate verification results, and send them to the collaborative scheduling module. The collaborative scheduling module is configured to: receive data sent by the environmental interference monitoring module and verification results sent by the real-time verification module; generate and send weight coefficient adjustment instructions to the multi-source dynamic entropy generation module based on the data sent by the environmental interference monitoring module; generate and send system power consumption mode switching instructions based on the verification results sent by the real-time verification module and the externally set task priorities; and execute a preset abnormal response sequence when the verification result indicates that the health test has failed. This sequence includes instructing the multi-source dynamic entropy generation module to pause generation and reconstruct the weight coefficients based on the verification results and monitoring data.

Images