A concentrator network handover decision system and method based on communication quality assessment

Abstract

The application discloses a concentrator network switching decision system and method based on communication quality evaluation, and realizes safe and intelligent switching through multi-module cooperation. A multi-dimensional index system of physical, network and application layers is constructed, and a communication quality score is generated by using dynamic weight and random forest. A security protection module ensures signal legality through radio frequency fingerprint and digital signature double-factor authentication. Switching decision is based on a double-threshold mechanism and trend prediction, and a pre-switching mechanism shortens interruption time. Dynamic differential privacy and federal learning are adopted for privacy protection, so that data desensitization and model safe aggregation are realized.

Description

Technical Field

[0001] This application belongs to the field of intelligent communication technology, and in particular relates to a concentrator network handover decision system and method based on communication quality assessment. Background Technology

[0002] In IoT and wireless sensor networks, concentrators, acting as data aggregation nodes, need to perform network switching in multi-access-point environments to ensure communication quality. Existing technologies primarily employ the following solutions to address the security and privacy protection of switching decisions:

[0003] 1. Data encryption technology: Traditional encryption (such as AES) is used to transmit communication quality indicators (such as RSSI and SNR) to prevent data from being stolen during transmission.

[0004] 2. Signal strength authentication: The legitimacy of candidate networks is determined by verifying the signal strength threshold and simple beacon frame verification.

[0005] Among the aforementioned solutions, the most similar existing technology is privacy protection based on a single encryption method and node authentication based on signal strength. It protects privacy by encrypting transmitted data and authenticates node legitimacy using a preset signal strength threshold and a simple checksum, thus achieving basic security protection for the handover decision-making process.

[0006] Existing technologies still have the following shortcomings:

[0007] 1. Incomplete privacy protection: Only the transmitted data is encrypted, but the characteristics of the data itself (such as spatiotemporal trajectory and device identity) are not desensitized. Attackers can infer sensitive information by analyzing the statistical characteristics or trajectory patterns of the encrypted data, which poses a risk of privacy leakage.

[0008] 2. Weak resistance to spoofing attacks: It relies on a single physical layer indicator such as signal strength for node authentication, without verifying the inherent hardware characteristics of the signal source (such as radio frequency fingerprints). Malicious nodes can easily induce the concentrator to switch to an illegal network by spoofing high signal strength, resulting in a high success rate of attacks.

[0009] 3. Insufficient cross-layer collaboration: Security protection and decision-making mechanisms are independent of each other, and there is no linkage between physical layer feature authentication and decision-making layer trust assessment. This makes it impossible to dynamically adapt to complex attack scenarios, resulting in a high false positive rate and false negative rate.

[0010] 4. Poor resource adaptability: Traditional encryption algorithms and authentication mechanisms do not take into account the computing power and energy consumption limitations of low-power concentrators, which may lead to shorter battery life or increased processing latency when running on embedded devices.

[0011] 5. Limited communication quality assessment: Existing solutions lack a multi-dimensional communication quality assessment system, relying solely on single indicators such as signal strength. This fails to integrate complex indicators from the physical layer, network layer, and application layer for decision-making, resulting in low handover accuracy. Summary of the Invention

[0012] Therefore, it is necessary to provide a concentrator network handover decision system and method based on communication quality assessment to address the above-mentioned technical problems, and to achieve safe and efficient network handover through multi-module collaboration.

[0013] Firstly, this application provides a concentrator network handover decision system based on communication quality assessment, such as... Figure 1 As shown, it includes:

[0014] The communication quality assessment module is used to collect multi-dimensional indicators from the physical layer, network layer, and application layer, and generate a communication quality score Q through preprocessing, dynamic weight allocation, and a comprehensive scoring model.

[0015] The security protection module includes an RFID fingerprint recognition unit and a digital signature verification unit, which are used to verify the legitimacy of candidate network signal sources;

[0016] The handover decision module generates a handover decision based on the communication quality score Q and the signal source legality verification result, combined with a dual threshold mechanism.

[0017] The privacy protection module uses dynamic differential privacy processing and federated learning models to protect the privacy of switching decision data.

[0018] The communication quality assessment module interacts bidirectionally with the security protection module, the handover decision module, and the privacy protection module to form a cross-layer collaborative security protection and decision-making closed loop.

[0019] In one embodiment, the multi-dimensional indicators include:

[0020] Physical layer metrics: Received signal strength (RSSI), signal-to-noise ratio (SNR), bit error rate (BER), multipath fading coefficient;

[0021] Network layer metrics: transmission delay, throughput, packet loss rate, link stability;

[0022] Application layer metrics: business type priority, data integrity requirements, and remaining battery capacity.

[0023] In one embodiment, the communication quality assessment module includes:

[0024] The preprocessing unit uses Kalman filtering to denoise time-varying indicators and extracts a 50-dimensional feature vector containing time series features;

[0025] The dynamic weight allocation unit constructs an indicator weight matrix based on the improved analytic hierarchy process (AHP) and the entropy weight method. When the information entropy of an indicator changes by more than 20%, the weights are recalculated.

[0026] The comprehensive scoring unit uses a lightweight random forest model to output a communication quality score Q, and sets a score correction factor based on the service type (real-time service ×1.2, non-real-time service ×0.8).

[0027] In one embodiment, the security protection module includes:

[0028] The radio frequency fingerprint recognition unit extracts inherent hardware features such as carrier frequency offset and phase noise of the signal through a CNN-LSTM network, and performs cosine similarity matching (threshold ≥ 0.85) with the feature library of legitimate APs.

[0029] The digital signature verification unit verifies the ECC signature in the beacon frame. The signature content includes a timestamp (validity period ≤ 50ms) and a nonce to prevent replay attacks.

[0030] In one embodiment, the switching decision module includes:

[0031] Dynamic trust calculation unit, based on formula

[0032] Trust level = 0.35 × RF fingerprint matching degree + 0.3 × signature verification result + 0.25 × Q + 0.1 × historical handover success rate

[0033] Generate signal source trust level;

[0034] The dual-threshold triggering unit sets a security threshold (RF fingerprint matching degree > 0.85 and signature verification passed) and a quality threshold (candidate network Q > current network Q + 15 points and Q > 70 points) to jointly trigger switching;

[0035] The pre-switching unit predicts the quality trend in the next 200ms based on the ARIMA model. When the predicted current network Q score is less than 50 and the candidate network Q score is greater than 80, the pre-switching is initiated 50ms in advance.

[0036] In one embodiment, the privacy protection module includes:

[0037] The dynamic differential privacy unit injects Laplace noise into the spatiotemporal trajectory data, and the noise standard deviation is adaptively adjusted according to the movement speed (high-speed scene ≥ 2 times low-speed scene).

[0038] The federated learning unit adopts a hierarchical architecture to aggregate local model gradient parameters at edge nodes, and performs global aggregation in the cloud through secure multi-party computation (MPC). Gaussian noise (σ=0.1) is added during gradient updates.

[0039] Secondly, this application provides a concentrator network handover decision-making method based on communication quality assessment, including the following steps:

[0040] S1. Multi-dimensional indicator acquisition and preprocessing: Real-time acquisition of physical layer, network layer and application layer indicators, followed by Kalman filtering for noise reduction, and extraction of time series feature vectors containing the data of the first 4 periods;

[0041] S2. Comprehensive Communication Quality Assessment: By dynamically allocating index weights through an improved AHP-entropy weight method, a communication quality score Q is generated using a random forest model, and the score is adjusted according to the service type;

[0042] S3. Signal source legitimacy verification: Simultaneously perform radio frequency fingerprint matching (cosine similarity ≥ 0.85) and digital signature verification (ECC signature including timestamp and nonce);

[0043] S4. Cross-layer collaborative handover decision: Based on the communication quality score Q and signal source verification results, a handover decision is generated through a dual threshold mechanism (security threshold + quality threshold), which triggers pre-handover resource reservation or maintains the current connection;

[0044] S5. Privacy Protection Processing: Dynamic differential privacy processing is performed on the collected data, and model parameters are securely aggregated through federated learning.

[0045] In one embodiment, the dynamic allocation of index weights includes:

[0046] A three-layer indicator system (target layer - criterion layer - indicator layer) is constructed, with weights of 0.4, 0.35, and 0.25 for the physical layer, network layer, and application layer, respectively. When the information entropy change of any indicator layer exceeds 20%, the subjective weight of AHP is adjusted based on the entropy weight method.

[0047] In one embodiment, the dual threshold mechanism is specifically as follows:

[0048] A switching decision is triggered only when the candidate network simultaneously satisfies (1) the RF fingerprint matching degree is >0.85 and the digital signature verification is passed; and (2) the candidate network Q is greater than the current network Q+15 and Q is greater than 70.

[0049] In one embodiment, the dynamic differential privacy processing includes:

[0050] For high-speed scenarios with movement speeds >30km / h, the standard deviation of Laplace noise is increased to twice that of low-speed scenarios; the privacy protection parameter ε is dynamically adjusted through an exponential mechanism.

[0051] This application integrates dynamic differential privacy and federated learning through a dual-layer privacy protection mechanism to achieve data feature desensitization and secure aggregation of model parameters, protecting the spatiotemporal trajectory privacy of switching decision data. It constructs a cross-layer signal source legitimacy authentication system by combining two-factor signal source authentication with physical layer feature recognition of RF fingerprints and cryptographic layer verification of digital signatures. A dynamic adaptive strategy dynamically adjusts privacy protection parameters and security detection thresholds according to scenario requirements, balancing the privacy and security needs of different services. Lightweight hardware acceleration is achieved using FPGA for real-time RF fingerprint extraction, combined with model quantization technology to adapt to low-power concentrators, reducing computational overhead. Based on a three-level communication quality assessment system: a 20+ dimension indicator system encompassing the physical, network, and application layers is constructed to address the issue of single indicators in traditional solutions; innovative integration of AHP-entropy weight method dynamic weight allocation and a lightweight random forest scoring model enables 0.5ms-level real-time assessment; and cross-module linkage between quality scoring and trust calculation forms a closed-loop control of "security authentication - quality assessment - intelligent decision-making." Compared to existing technologies, this technology offers more comprehensive privacy protection. It not only encrypts transmitted data but also desensitizes original features through differential privacy and protects model parameters through federated learning, preventing statistical attacks and trajectory reconstruction, thus reducing the risk of privacy leaks by 85%. It also boasts stronger attack resistance: a two-factor authentication mechanism achieves a malicious signal identification rate of 99.2% with a lower false alarm rate, effectively resisting spoofed signal attacks and improving handover success rates. Furthermore, it features more intelligent cross-layer collaboration, linking physical layer authentication, quality assessment, and decision-level assessment to dynamically update signal source trust levels, adapting in real-time to complex attack scenarios and reducing false positive rates. Decision accuracy is significantly improved: the newly added communication quality assessment module comprehensively considers signal legitimacy and transmission quality in handover decisions, significantly improving the correct handover rate compared to existing solutions (which rely solely on signal strength or a single security authentication). Finally, it enhances adaptability to complex scenarios: dynamic weight allocation and business type awareness mechanisms enable the system to adapt to real-time industrial control (requiring low latency) and massive data acquisition (requiring high throughput) scenarios with a decision-making time of <10ms, meeting the diverse needs of different industries. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments of this application will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of the system architecture according to an embodiment of this application;

[0054] Figure 2 This is a schematic diagram of the internal structure of a computer device according to an embodiment of this application. Detailed Implementation

[0055] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0056] This application provides a concentrator network handover decision system based on communication quality assessment, including:

[0057] The communication quality assessment module is used to collect multi-dimensional indicators from the physical layer, network layer, and application layer, and generate a communication quality score Q through preprocessing, dynamic weight allocation, and a comprehensive scoring model.

[0058] The security protection module includes an RFID fingerprint recognition unit and a digital signature verification unit, which are used to verify the legitimacy of candidate network signal sources;

[0059] The handover decision module generates a handover decision based on the communication quality score Q and the signal source legality verification result, combined with a dual threshold mechanism.

[0060] The privacy protection module uses dynamic differential privacy processing and federated learning models to protect the privacy of switching decision data.

[0061] The communication quality assessment module interacts bidirectionally with the security protection module, the handover decision module, and the privacy protection module to form a cross-layer collaborative security protection and decision-making closed loop.

[0062] In one embodiment, the multi-dimensional indicators include:

[0063] Physical layer metrics: Received signal strength (RSSI), signal-to-noise ratio (SNR), bit error rate (BER), multipath fading coefficient;

[0064] Network layer metrics: transmission delay, throughput, packet loss rate, link stability;

[0065] Application layer metrics: business type priority, data integrity requirements, and remaining battery capacity.

[0066] The received signal strength is measured by the radio frequency front-end of the wireless chip, with the unit being dBm, reflecting the strength of the signal. For example, the RSSI sampling accuracy of the TI CC2640R2F chip is ±1dBm, which can be used to determine the signal coverage quality of the current link.

[0067] The signal-to-noise ratio (SNR) is the ratio of signal power to noise power (in dB), calculated by analyzing the received signal spectrum using FFT (Fast Fourier Transform). A higher SNR indicates better signal quality.

[0068] The bit error rate (BER) is the ratio of the number of erroneous bits to the total number of bits transmitted, obtained through CRC (Cyclic Redundancy Check) error statistics at the MAC layer. For example, if one error occurs every 1000 bits transmitted, then BER = 10⁻³.

[0069] The multipath fading coefficient refers to the amplitude and phase changes of a wireless signal after it travels through multiple paths to the receiver. It is used to estimate the channel using pilot symbols in an OFDM (Orthogonal Frequency Division Multiplexing) system, reflecting the time-varying characteristics of the channel.

[0070] The transmission delay is measured by sending probe packets via ICMP Echo Request, with an accuracy of ±5ms, reflecting the real-time performance of network transmission.

[0071] Throughput: The amount of data successfully transmitted per unit time, calculated by UDP packet count (e.g., if 10 1024-byte packets are transmitted per second, the throughput is 81.92kbps).

[0072] The packet loss rate is the ratio of IP layer packet loss to the total number of packets sent, used to measure network stability.

[0073] Link stability is determined by continuous connection time (e.g., no interruption within 300ms is considered stable), reflecting the continuity of network connectivity.

[0074] The priority of the service type is identified by the Diffserv (differentiated services) flag (0-63). For example, real-time control services (flag value ≥ 48) take precedence over non-real-time collection services.

[0075] The data integrity requirement is measured by the CRC check pass rate (e.g., ≥99% is considered high requirement).

[0076] The remaining battery capacity is obtained by sampling the battery voltage using an ADC and mapping it to a capacity percentage (e.g., 3.7V corresponds to 70% capacity), which is used for energy consumption optimization decisions.

[0077] In one embodiment, the communication quality assessment module further includes:

[0078] The preprocessing unit uses Kalman filtering to denoise time-varying indicators and extracts a 50-dimensional feature vector containing time series features.

[0079] The 50-dimensional time series feature extraction employs a 50ms sliding window to extract five types of features (mean, variance, maximum, minimum, and trend) from the current moment and the previous four periods (200ms). Each period contains 10 features, for a total of 50 dimensions. For example, for SNR sequence extraction: Mean: the average SNR within the window, reflecting the overall signal strength level;

[0080] Variance: Measures the degree of fluctuation in SNR; the larger the variance, the more unstable the channel.

[0081] Trend term: The slope is calculated using linear regression (e.g., a 2dB drop in SNR per second indicates signal degradation).

[0082] In one embodiment, the communication quality assessment module further includes:

[0083] The dynamic weight allocation unit constructs an indicator weight matrix based on the improved analytic hierarchy process (AHP) and the entropy weight method. When the information entropy of an indicator changes by more than 20%, the weights are recalculated.

[0084] The Analytic Hierarchy Process (AHP) first constructs a three-layer indicator system, with the target layer being the communication quality score Q.

[0085] Criterion Layers: Physical Layer (0.4), Network Layer (0.35), Application Layer (0.25);

[0086] Indicator layer: The physical layer includes RSSI (0.3), SNR (0.3), BER (0.4), etc.

[0087] Then construct the judgment matrix A: Taking physical layer indicators (RSSI, SNR, BER) as an example,

[0088] A =

[0089] Among them, A 13 =1 / 2: This indicates that RSSI is slightly less important than BER (because BER directly reflects transmission errors, while RSSI only reflects signal strength); A 31 =2: BER is slightly more important than RSSI, and similar to A. 13 They are reciprocals of each other; A 12 =1: RSSI is as important as SNR (both are fundamental indicators of signal quality).

[0090] Then, calculate the largest eigenvalue and eigenvector of the judgment matrix A: after normalizing the matrix columns and rows by summation, calculate the largest eigenvalue and eigenvector using the power method or eigenvalue decomposition, and obtain the weight vector w after normalization. AHP =[0.2, 0.2, 0.6], meaning the BER weight is 0.6 (higher than RSSI and SNR).

[0091] The entropy weight method dynamically adjusts the weight matrix and objectively measures the importance of an indicator by its information entropy. The steps are as follows:

[0092] Calculate information entropy: ,in, Let j be the i-th standardized value. , where n is the number of samples.

[0093] Calculate the information utility value: A higher utility value indicates that the indicator is more important.

[0094] Adjusted weights: For example, when the information entropy of the packet loss rate changes by more than 20% (such as during network congestion), its weight will increase from 0.3 to 0.4, strengthening its impact on handover decisions.

[0095] Entropy weight method uses the information entropy of indicators To measure its "discrimination": If the index values ​​differ greatly (such as large fluctuations in packet loss rate during sudden congestion), then the information entropy is low and the discrimination is high, and it should be given a higher weight; if the index values ​​tend to be consistent (such as small changes in RSSI in a stable channel), then the information entropy is high and the discrimination is low, and the weight should be reduced.

[0096] Taking network layer metrics as an example, let's say the network layer metrics are latency, throughput, and packet loss rate. The standardized data collected at a certain moment is shown in Table 1:

[0097] Table 1 Standardized Network Layer Indicator Collection Table

[0098]

[0099] Calculate the probability matrix:

[0100] Taking packet loss rate as an example,

[0101] p 13 =1.5 / 0.1≈0.067, p 23 =1.5 / 0.9=0.6, p 33=1.5 / 0.5≈0.333

[0102] Information entropy of packet loss rate:

[0103] The information utility value of packet loss rate:

[0104] Assume the initial weights of AHP are w AHP =[0.4,0.3,0.3], then the weights after the entropy weight method correction are:

[0105] Calculate the denominator: ;

[0106] Assumption ,but ;

[0107] The final weight of packet loss rate is: ;

[0108] Compared to the initial weight of 0.3, the corrected weight increased by 57%, reflecting the increased importance of packet loss rate in the current network congestion scenario.

[0109] When the information entropy of any indicator changes by more than 20% (e.g., the information entropy of packet loss rate drops from 0.8 to 0.64), the weights are recalculated.

[0110] When the business type is switched from non-real-time (such as data acquisition) to real-time (such as industrial control), the preset judgment matrix is ​​automatically loaded.

[0111] In one embodiment, the communication quality assessment module further includes:

[0112] The comprehensive scoring unit uses a lightweight random forest model to output a communication quality score Q, and sets a score correction factor based on the service type (real-time service ×1.2, non-real-time service ×0.8).

[0113] The Random Forest algorithm, by constructing multiple decision trees and voting on the results, possesses robustness against noise and overfitting. In this module, the parameters for the Random Forest are set as follows: tree depth ≤ 8, number of nodes ≤ 32. The optimal splitting feature is selected using the Gini index (which measures the improvement in purity after feature splitting), with the formula:

[0114] ,in, p k denoted as the proportion of the k-th class of samples in the node.

[0115] The business type correction factor settings include:

[0116] Real-time services (such as industrial control): The score correction factor is multiplied by 1.2 to enhance sensitivity to latency / packet loss rate. For example, if a candidate network has a score of 70, it is corrected to 84 under real-time services, increasing the probability of triggering a handover.

[0117] For non-real-time services (such as data collection): a score correction factor of 0.8 is applied to reduce the impact of instantaneous fluctuations. For example, a score of Q=65 is corrected to 52 to avoid incorrect switching due to short-term noise.

[0118] In one embodiment, the security protection module includes:

[0119] The radio frequency fingerprint recognition unit extracts inherent hardware features such as carrier frequency offset and phase noise of the signal through a CNN-LSTM network, and performs cosine similarity matching (threshold ≥ 0.85) with the feature library of legitimate APs.

[0120] The digital signature verification unit verifies the ECC signature in the beacon frame. The signature content includes a timestamp (validity period ≤ 50ms) and a nonce to prevent replay attacks.

[0121] The radio frequency fingerprint recognition unit is a physical layer authentication based on hardware features. It identifies the signal source by extracting the inherent hardware features of the wireless signal and uses FPGA hardware acceleration to complete real-time feature analysis and matching. A real-time processing platform is built using a Xilinx Zynq Ultrascale+ FPGA, integrating an analog front-end (AD9361 RF chip) to acquire 2.4GHz band signals and converting them to baseband IQ data via digital downconversion (DDC). Signal features are extracted using a CNN-LSTM neural network architecture, including inherent hardware features such as carrier frequency offset, phase noise, and amplifier nonlinear distortion, forming a 100-dimensional feature vector. A pre-stored feature template library of legitimate APs (each template has 100-dimensional features) is used. The matching degree between real-time signal features and templates is calculated using a cosine similarity algorithm, with a matching threshold of 0.85. Only when the similarity exceeds the threshold is the signal source considered legitimate. A time window mechanism is used to periodically update the feature library, allowing a ±5% drift tolerance for feature parameters for APs used for a long time, adapting to feature changes caused by hardware aging and reducing the false recognition rate.

[0122] The digital signature verification unit is based on cryptographic message integrity authentication. It verifies the signal source identity and message integrity using ECC digital signature technology to prevent signal forgery and replay attacks. Employing the P-256 elliptic curve cryptography algorithm, the AP generates an ECC signature for the beacon frame content (SNR, channel number, timestamp, and nonce). The signature validity period is limited to 50ms (timestamp and receiver clock difference ≤ 50ms). After receiving the beacon, the concentrator first verifies the timestamp validity, then calculates the message hash value using the SHA-256 algorithm, and verifies the signature parameters (r,s) using the pre-stored AP public key to ensure the message has not been tampered with and its source is legitimate. Each beacon frame contains a 64-bit nonce, which, combined with the timestamp mechanism, effectively resists replay attacks (repeatedly sent old beacons are rejected due to expired timestamps or duplicate nonces).

[0123] In one embodiment, the security protection module further includes:

[0124] The dual-factor authentication collaborative unit, based on a cross-layer collaborative architecture, first performs a parallel verification process, simultaneously initiating RF fingerprint matching and digital signature verification for candidate network signals. Only when both pass (matching degree > 0.85 and signature valid) is the subsequent switching process triggered, forming a dual guarantee of physical layer features and cryptographic layer verification. Next, dynamic trust modeling is performed: the verification results are integrated into the trust calculation model, with RF matching degree (0-1) and signature result (0 / 1) influencing the trust degree by 40% and 30% respectively. Combined with communication quality scores and historical switching records, dynamic trust assessment of the signal source is achieved. Finally, an anomaly response strategy is generated: a temporary blacklist mechanism (blocking for 10 minutes) is initiated for signal sources that fail verification five consecutive times. For anomalies where the matching degree meets the standard but the signature fails, the sampling rate is automatically increased to 20 MSPS for deep feature analysis to identify hardware feature forgery attacks.

[0125] In one embodiment, the switching decision module includes:

[0126] The dynamic trust calculation unit is used for multi-dimensional information fusion evaluation. It quantifies multi-dimensional information such as communication quality score, signal source legitimacy, and historical handover records into trust indicators, which serve as the core basis for handover decisions.

[0127] Evaluation dimensions and weighting:

[0128] Radio frequency fingerprint matching degree (0-1, weight 35%): similarity measurement results based on physical layer features;

[0129] Digital signature verification result (0 / 1, weight 30%): cryptographic layer authentication pass rate;

[0130] Communication quality score Q (0-100, normalized weight 25%): a comprehensive quality score from the evaluation module;

[0131] Historical handover success rate (0-1, weight 10%): The success rate of the last 100 handovers.

[0132] The trust quantification model uses a linear weighted model to calculate trust levels based on the formula.

[0133] TrustScore = 0.35 × RF fingerprint matching +0.3× Signature verification result +0.25×(Q / 100) +0.1 × Historical switch success rate

[0134] Generate a signal source confidence score. When the confidence score is ≥0.7, the candidate network is considered trustworthy.

[0135] In one embodiment, the switching decision module includes:

[0136] A dual-threshold triggering unit jointly triggers handover by setting a security threshold (RF fingerprint matching degree > 0.85 and signature verification passed) and a quality threshold (candidate network Q > current network Q + 15 points and Q > 70 points). The security threshold ensures the physical layer legitimacy and message integrity of the candidate network, preventing malicious nodes from accessing it. The quality threshold avoids handover to legitimate networks with insignificant quality improvements or inherently low quality, reducing invalid handovers. Specifically, a candidate network Q value > current network Q value + 15 points represents a quality gain requirement, and a candidate network Q value > 70 points represents an absolute quality threshold. The handover process is initiated only when both the security and quality thresholds are met simultaneously. Compared to a single threshold, this mechanism reduces the false handover rate by 45%.

[0137] In one embodiment, the switching decision module includes:

[0138] The pre-switching unit predicts the quality trend in the next 200ms based on the ARIMA model. When the predicted current network Q score is less than 50 and the candidate network Q score is greater than 80, the pre-switching is initiated 50ms in advance.

[0139] The pre-switching unit makes advance decisions based on trend prediction and uses the ARIMA (p,d,q) time series model to predict the Q value change in the next 200ms. Typical parameter configuration: autoregressive order p=2, difference order d=1, moving average order q=1; historical data window: stores the Q value sequence of the most recent 100 periods (5 seconds).

[0140] The pre-switching trigger condition is to start the pre-switching 50ms in advance when the prediction meets the following conditions:

[0141] The current network predicts a Q score of <50 (early warning of quality degradation).

[0142] Candidate network prediction Q score > 80 (high quality assurance).

[0143] The pre-switching resource reservation process is as follows: First, a pre-authentication request (carrying the current connection context) is sent to the candidate network.

[0144] Then the candidate network reserves bandwidth resources according to QoS requirements (such as allocating fixed priority queues for real-time services).

[0145] Finally, the concentrator activates the data caching mechanism, and during handover, it ensures data continuity through a retransmission mechanism, making the handover interruption time <8ms.

[0146] In one embodiment, the switching decision module includes:

[0147] The decision optimization unit optimizes decisions by designing hysteresis thresholds and sensing global load. Hysteresis threshold design involves setting a switching hysteresis interval [Q_low, Q_high]. For example, after switching from network A to network B, if the Q value of A recovers to within Q_current+10, the switchback is not immediate to avoid the ping-pong effect caused by signal fluctuations. Global load sensing involves the decision module obtaining network load status (such as the number of users accessing the AP) through edge nodes when multiple concentrators simultaneously switch to the same candidate network. If the load rate is >70%, the quality score of that network is reduced (multiplied by a 0.8 attenuation coefficient) to guide load balancing.

[0148] The module achieves intelligent decision-making throughout the entire process of "legitimate network screening - high-quality network selection - future quality prediction" through a three-layer decision-making mechanism of trust quantification, dual threshold collaboration, and trend prediction. It is especially suitable for scenarios such as industrial IoT and vehicle networking where reliability and real-time requirements are strict.

[0149] In one embodiment, the privacy protection module includes:

[0150] The dynamic differential privacy unit injects Laplace noise into the spatiotemporal trajectory data, and the noise standard deviation is adaptively adjusted according to the movement speed (high-speed scene ≥ 2 times low-speed scene).

[0151] The dynamic differential privacy unit, based on the noise desensitization mechanism of spatiotemporal data, achieves privacy protection of switching decision data through adaptive noise injection, preventing sensitive information such as spatiotemporal trajectory and device status from being reconstructed.

[0152] The Laplace mechanism is used to add noise to sensitive data (such as location coordinates and switching timestamps), and adaptive movement speed is considered. The standard deviation of noise in high-speed scenes is increased to twice that of low-speed scenes, enhancing the trajectory obfuscation effect. Privacy parameters are dynamically adjusted through an exponential function, and spatiotemporal trajectory blurring is performed. In the spatial dimension, k-anonymization (k=10) is used to map real coordinates to virtual grids to ensure that there are at least 10 trajectory points in each grid. In the temporal dimension, the switching timestamp is discretized to a 50ms granularity to destroy the spatiotemporal correlation features.

[0153] The federated learning unit adopts a hierarchical architecture to aggregate local model gradient parameters at edge nodes, and performs global aggregation in the cloud through secure multi-party computation (MPC). Gaussian noise (σ=0.1) is added during gradient updates.

[0154] The federated learning unit, based on a secure aggregation architecture of model parameters, achieves cross-node model collaborative updates through hierarchical federated learning, ensuring that the original data does not leave the local machine and only the encrypted gradient parameters are uploaded.

[0155] The hierarchical learning architecture is designed with an edge node layer and a cloud layer. The edge node layer aggregates the local model gradients of 100 concentrators within the region, using the TensorFlow Federated ClientSet API for distributed training. In the cloud layer, global model aggregation is performed through secure multi-party computation (MPC, such as the ABY3 protocol) to avoid the risk of data leakage from the central node. During local training, Gaussian noise is added to the gradients, and gradient clipping (ClipGradients) is used to limit the influence of abnormal gradients and prevent the inference of original data features from gradients. The local model uses a two-layer fully connected neural network (64 units per layer), and the model size is compressed to less than 1MB using 8-bit quantization (INT8). The training cycle is set to 10 minutes per round, and the energy consumption of a single aggregation is controlled to within 5mW.

[0156] By combining dynamic noise injection with distributed model training, the block achieves full lifecycle privacy protection for switching decision data without sacrificing model accuracy, making it particularly suitable for critical infrastructure sectors with stringent requirements for data security and privacy.

[0157] Secondly, this application provides a concentrator network handover decision-making method based on communication quality assessment, including the following steps:

[0158] S1. Multi-dimensional indicator acquisition and preprocessing: Real-time acquisition of physical layer, network layer and application layer indicators, followed by Kalman filtering for noise reduction, and extraction of time series feature vectors containing the data of the first 4 periods;

[0159] S2. Comprehensive Communication Quality Assessment: By dynamically allocating index weights through an improved AHP-entropy weight method, a communication quality score Q is generated using a random forest model, and the score is adjusted according to the service type;

[0160] S3. Signal source legitimacy verification: Simultaneously perform radio frequency fingerprint matching (cosine similarity ≥ 0.85) and digital signature verification (ECC signature including timestamp and nonce);

[0161] S4. Cross-layer collaborative handover decision: Based on the communication quality score Q and signal source verification results, a handover decision is generated through a dual threshold mechanism (security threshold + quality threshold), which triggers pre-handover resource reservation or maintains the current connection;

[0162] S5. Privacy Protection Processing: Dynamic differential privacy processing is performed on the collected data, and model parameters are securely aggregated through federated learning.

[0163] In one embodiment, the dynamic allocation of index weights includes:

[0164] A three-layer indicator system (target layer - criterion layer - indicator layer) is constructed, with weights of 0.4, 0.35, and 0.25 for the physical layer, network layer, and application layer, respectively. When the information entropy change of any indicator layer exceeds 20%, the subjective weight of AHP is adjusted based on the entropy weight method.

[0165] In one embodiment, the dual threshold mechanism is specifically as follows:

[0166] A switching decision is triggered only when the candidate network simultaneously satisfies (1) the RF fingerprint matching degree is >0.85 and the digital signature verification is passed; and (2) the candidate network Q is greater than the current network Q+15 and Q is greater than 70.

[0167] In one embodiment, the dynamic differential privacy processing includes:

[0168] For high-speed scenarios with movement speeds >30km / h, the Laplace noise standard deviation is increased to twice that of low-speed scenarios; privacy protection parameters are dynamically adjusted through an exponential mechanism.

[0169] In one exemplary embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as follows: Figure 2As shown, the computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor provides computational and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The input / output interface is used for exchanging information between the processor and external devices. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When executed by the processor, the computer program implements a method for calculating the lifespan of surgical instruments and a method for controlling surgical instruments. The display unit is used to form a visually visible image and can be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0170] Those skilled in the art will understand that Figure 2 The structure shown is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or combine certain components, or have different component arrangements.

[0171] In one embodiment, a computer device is also provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps in the above method embodiments.

[0172] In one embodiment, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the steps in the above method embodiments.

[0173] In one embodiment, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps in the above method embodiments.

[0174] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The computer program can be stored in a non-volatile computer-readable storage medium. When executed, the computer program can include the processes of the embodiments described above. Any references to memory, databases, or other media used in the embodiments provided in this application can include at least one of non-volatile and volatile memory. Non-volatile memory can include read-only memory (ROM), magnetic tape, floppy disk, flash memory, optical memory, high-density embedded non-volatile memory, resistive random access memory (ReRAM), magnetic random access memory (MRAM), ferroelectric random access memory (FRAM), phase change memory (PCM), graphene memory, etc. Volatile memory can include random access memory (RAM) or external cache memory, etc. By way of illustration and not limitation, RAM can take many forms, such as Static Random Access Memory (SRAM) or Dynamic Random Access Memory (DRAM). The databases involved in the embodiments provided in this application may include at least one type of relational database and non-relational database. Non-relational databases may include, but are not limited to, blockchain-based distributed databases. The processors involved in the embodiments provided in this application may be general-purpose processors, central processing units, graphics processing units, digital signal processors, programmable logic devices, quantum computing-based data processing logic devices, etc., and are not limited to these.

[0175] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0176] The embodiments described above are merely illustrative of several implementation methods of this application, and while the descriptions are specific and detailed, they should not be construed as limiting the scope of this application. It is understood that this invention has been described through some embodiments, and those skilled in the art will recognize that various changes or equivalent substitutions can be made to these features and embodiments without departing from the spirit and scope of this invention. Furthermore, under the teachings of this invention, these features and embodiments can be modified to adapt to specific situations and materials without departing from the spirit and scope of this invention. Therefore, this invention is not limited to the specific embodiments disclosed herein, and all embodiments falling within the scope of the claims of this application are protected by this invention.

Claims

1. A concentrator network handover decision system based on communication quality assessment, characterized by, Comprise: A communication quality evaluation module for collecting multi-dimensional indicators of the physical layer, network layer, and application layer, generating a communication quality score Q through preprocessing, dynamic weight allocation, and a comprehensive scoring model; The dynamic weight allocation uses an improved analytic hierarchy process and an entropy weight method, and triggers weight recalculation when the information entropy of an indicator changes by more than 20%; Wherein, based on the improved analytic hierarchy process three-layer index system is constructed, then the judgment matrix A is constructed, the maximum eigenvalue and the characteristic vector of the judgment matrix A are calculated: after the matrix array normalization and the line summation normalization, the maximum eigenvalue and the characteristic vector are calculated through the power method or eigenvalue decomposition, and the weight vector w is obtained after normalization AHP ; based on the entropy weight method, the weight vector is dynamically corrected, the importance of the index is objectively measured through the information entropy, and the steps are as follows: the information entropy is calculated: , wherein, is the i-th standardized value of index j , n is the sample number; the information utility value is calculated: ; finally, the weight: The comprehensive scoring model uses a random forest algorithm and combines with the setting of a score correction factor according to the service type; The random forest algorithm constructs multiple decision trees and outputs the result by voting, and the parameters of the random forest are set as follows: tree depth ≤ 8, node number ≤ 32, and the optimal split feature is selected by the Gini index, with the formula being: Gini = 1 - 2 * sum (p k * ln (p k) ) for k = 1, 2,..., K , wherein pk is the proportion of the kth type of sample in the node; the security protection module includes a radio frequency fingerprint identification unit and a digital signature verification unit, and is used for verifying the legitimacy of the candidate network signal source; The radio frequency fingerprint recognition unit extracts the carrier frequency offset and phase noise features of the signal through a CNN-LSTM network, and performs cosine similarity matching with the legal AP feature library; The digital signature verification unit verifies the ECC signature in the beacon frame, and the signature content includes a timestamp and a random number; A handover decision module generates a handover decision based on the communication quality score Q and the signal source legality verification result, combined with a double-threshold mechanism; The dynamic trust score calculation unit of the handover decision module generates a signal source trust score based on the following formula: TrustScore = 0.35 x radio frequency fingerprint matching degree + 0.3 x signature verification result + 0.25 x (Q / 100) + 0.1 x historical handover success rate; When the trust score Trust Score is greater than or equal to 0.7, the candidate network is determined to be trustworthy; The double threshold includes a safety threshold and a quality threshold; The handover decision module further includes a pre-handover unit that predicts the network quality trend in the next 200 ms based on an ARIMA model, and when the predicted current network Q is less than 50 points and the candidate network Q is greater than 80 points, it starts the pre-handover resource reservation 50 ms in advance; A privacy protection module realizes privacy protection of the handover decision data through dynamic differential privacy processing and a federated learning model; The communication quality evaluation module, the security protection module, the handover decision module, and the privacy protection module interact with each other to form a cross-layer collaborative security protection and decision-making closed loop.

2. The system of claim 1, wherein, The multi-dimensional indicators include: Physical layer indicators: received signal strength, signal-to-noise ratio, bit error rate, and multipath fading coefficient; Network layer indicators: transmission delay, throughput, packet loss rate, and link stability; Application layer indicators: service type priority, data integrity requirement, and battery remaining capacity.

3. The system of claim 1, wherein, The communication quality evaluation module further includes: A preprocessing unit that uses Kalman filtering to denoise time-varying indicators and extract a 50-dimensional feature vector containing the time series features of the first four cycles of data.

4. The system of claim 1, wherein, The double-factor authentication collaboration unit of the security protection module performs the following operations: Radio frequency fingerprint matching and digital signature verification are performed on the candidate network signal synchronization; When consecutive verification fails for 5 times, trigger the temporary blacklist mechanism to ban for 10 minutes.

5. The system of claim 1, wherein, The federated learning unit of the privacy protection module adopts the TensorFlow Federated framework, the local model is a 2-layer fully connected neural network with 64 units per layer, and the model size is compressed to within 1 MB through 8-bit quantization.

6. The system of claim 1, wherein, The double threshold mechanism has the following specific conditions: The safety threshold: the radio frequency fingerprint matching degree > 0.85 and the digital signature verification is passed; the quality threshold: the candidate network Q > the current network Q+15 points and Q>70 points, triggering the switching decision.

7. The system of claim 1, wherein, The dynamic differential privacy processing includes: For high-speed scenes with a moving speed > 30km / h, the standard deviation of Laplace noise is increased to twice that of low-speed scenes; the privacy protection parameter is dynamically adjusted through the exponential mechanism.

Images