Unmanned aerial vehicle remote control system and method based on radio cracking technology

Abstract

The invention relates to the technical field of unmanned aerial vehicle radio cracking control, in particular to an unmanned aerial vehicle remote control system and method based on the radio cracking technology, and the system comprises a password distribution unit, a cracking progress estimation unit, and a GPU core control unit. A password distribution unit accurately divides a password database combination into subsets according to a specific algorithm and distributes the subsets to a GPU core to carry out cracking operation, a cracking progress estimation unit cooperates by means of multiple modules, constructs a model to deeply analyze password features, constructs a wide set sample of a module, reduces dimensions to disjunction key features, and carries out training optimization. The password cracking module quantifies difference weighting to obtain a coefficient, dynamically corrects and fuses multiple factors to predict remaining time, the difficulty evaluation module clusters subsets and accurately evaluates adaptation according to an encrypted submodel, the GPU core control unit intelligently regulates and controls frequency according to the estimated remaining time, stores and outputs a result after successful cracking, and the cracking efficiency is improved.

Description

Technical Field

[0001] The present invention relates to the technical field of unmanned aerial vehicle (UAV) radio cracking control, and more specifically, to a UAV remote control system and method based on radio cracking technology. Background Art

[0002] UAV radio cracking control is an important technology. In today's digital age, UAV systems transmit commands and data via radio signals, and their encryption mechanisms safeguard flight control rights and sensitive information security. However, with the development of encryption technology, UAV encryption has become increasingly complex, the cracking difficulty has soared, and there is an urgent need for efficient cracking technology.

[0003] Traditional cracking methods have significant defects in estimating the remaining cracking time. Most rely on fixed parameters or simple empirical formulas, without deeply considering password characteristics, the tried situations, and the real-time state of the system, resulting in serious inaccuracies in predicting the remaining cracking time. This leads to blind allocation of GPU core cracking resources and low overall cracking efficiency. When facing large-scale UAV encryption tasks, such as controlling unknown UAV swarms during military exercises and detecting illegally intruding UAVs in key areas, if the remaining cracking time cannot be accurately predicted, it will endanger security. To solve this problem, we provide a UAV remote control system and method based on radio cracking technology. Summary of the Invention

[0004] The purpose of the present invention is to provide a UAV remote control system and method based on radio cracking technology to solve the problems raised in the above background art.

[0005] To achieve the above purpose, a UAV remote control system and method based on radio cracking technology are provided, including a password distribution unit, a cracking progress estimation unit, and a GPU core control unit; The password distribution unit is configured with a password database for storing password combinations of common UAV encryption types. The database divides the password combinations into multiple password subsets based on a password distribution algorithm. When receiving a cracking start instruction, it transmits the divided password subsets to the corresponding GPU cores for password attempt cracking operations; The cracking progress estimation unit establishes a password combination feature analysis model for extracting features of password combinations. During the password attempt cracking process, it receives the tried password information and attempt result information fed back from the GPU cores, compares and analyzes the features of the tried passwords with the feature data in the password combination feature analysis model, combines the current number of tried passwords and the computing speed of the GPU cores, uses a pre-constructed progress estimation algorithm to calculate the cracking difficulty coefficient of the remaining passwords, and predicts the remaining cracking time based on the cracking difficulty coefficient and the current cracking speed, and transmits the prediction result to the GPU core control unit; The GPU core control unit receives the subset of passwords transmitted by the password distribution unit and distributes it to the corresponding GPU cores for processing. During the cracking process, it dynamically adjusts the working frequency of the GPU cores according to the remaining cracking time transmitted by the cracking progress estimation unit. When receiving the cracking success signal, it controls the GPU cores to stop the password attempt operation and stores and outputs the cracking result.

[0006] As a further improvement of this technical solution, the cracking progress estimation unit includes a model construction module. The construction method of the password combination feature analysis model in the model construction module is as follows: Collect a large number of different types of UAV encryption password combinations as sample data, perform multi-dimensional feature extraction on each password combination to obtain high-dimensional feature data, and use the principal component analysis algorithm to perform dimensionality reduction processing on the high-dimensional feature data to obtain key features; Based on the key features, use machine learning algorithms for model training until the model reaches a predetermined accuracy rate on the test data set, and construct an analysis model for evaluating the cracking difficulty characteristics of password combinations.

[0007] As a further improvement of this technical solution, the cracking progress estimation unit includes a password cracking module. The calculation method of the cracking difficulty coefficient in the password cracking module is as follows: For the received attempted password information, extract its password feature vector and the standard feature vector output by the password combination feature analysis model, and calculate their difference value; Use the analytic hierarchy process to determine the weight of each feature dimension in the cracking difficulty evaluation, and substitute the weight and the difference value into the formula to calculate the cracking difficulty coefficient. The cracking difficulty coefficient is used to reflect the deviation degree of the password from the easily cracked mode. At the same time, dynamically correct the difficulty coefficient according to the number of attempted passwords and the GPU core calculation speed.

[0008] As a further improvement of this technical solution, the remaining cracking time prediction method in the password cracking module is as follows: Let the total number of passwords be , and the number of attempted passwords be . According to the current cracking difficulty coefficient and the GPU core calculation speed , calculate the expected number of passwords cracked per unit time ; Use the exponential smoothing prediction method to predict and adjust the future cracking speed. Based on the historical cracking speed data , calculate the smoothing coefficient , and predict the future cracking speed ; where is the cracking speed predicted at the previous moment, and the predicted remaining cracking time is calculated accordingly 。

[0009] As a further improvement of this technical solution, the cracking progress estimation unit includes a difficulty evaluation module. During the process of attempting to crack the password, the difficulty evaluation module dynamically evaluates the overall cracking difficulty of the password subset. The specific method is as follows: Regularly perform clustering analysis on the attempted password subset, and use the K-Means algorithm to cluster the passwords into clusters with different difficulty levels; Calculate the proportion of the number of passwords in each cluster and the average cracking difficulty coefficient, and calculate the overall cracking difficulty index of the password subset by weighted calculation according to the proportion and the difficulty coefficient. If the overall cracking difficulty index exceeds the preset threshold, it is determined that the overall cracking difficulty of the password subset is high.

[0010] As a further improvement of this technical solution, when the difficulty evaluation module processes password combinations of different encryption types, the encryption type recognition and difficulty adaptation method is as follows: Use a convolutional neural network to identify the encryption type of the password combination, and train the network with password samples of different encryption types to identify the encryption algorithm feature pattern of the password combination; For different encryption types, construct corresponding cracking difficulty evaluation sub-models. The sub-models work together with the symmetric encryption key length and the password combination feature analysis model. After identifying the encryption type of the new password combination, switch to the corresponding sub-model to calculate the cracking difficulty coefficient and the predicted remaining cracking time.

[0011] As a further improvement of this technical solution, when the password cracking module predicts the remaining cracking time, the uncertainty factor compensation mechanism is as follows: Establish an uncertainty factor set, which includes the update frequency of the UAV encryption algorithm and the external electromagnetic interference intensity; Use the fuzzy comprehensive evaluation method to quantitatively score the influence degree of each factor, determine the weight, then calculate the comprehensive influence coefficient according to the score and the weight, and finally, based on the predicted remaining cracking time, construct a compensation time formula.

[0012] The second object of the present invention is to provide a method for implementing a UAV remote control system based on the radio cracking technology including any one of the above, including the following steps: S1. The system starts, and the password distribution unit loads common encrypted password combinations from the password database, divides them into subsets according to the password distribution algorithm, and then distributes them to the corresponding cores according to the GPU core performance and load balancing principle; S2. The cracking progress estimation unit first extracts multi-dimensional features from each password combination to obtain high-dimensional feature data. Then, it uses the principal component analysis algorithm to process the high-dimensional feature data, forms a transformation matrix, projects it into the high-dimensional space to obtain key features. Finally, based on the key features, it selects the support vector machine algorithm to construct a password combination cracking difficulty feature analysis model, and starts the cracking progress estimation process, receiving the tried password information and results fed back by the GPU core. S3. The GPU core conducts password cracking work according to the assigned password subset and feeds back the tried password information and trial results in real time. During the password cracking process, the cracking progress estimation unit uses a convolutional neural network to identify the encryption type of the new password combination, and switches the cracking difficulty evaluation sub-model according to the results. The sub-model calculates the cracking difficulty coefficient and predicts the remaining cracking time in combination with the password features. S4. When the GPU core cracks successfully, the GPU core control unit stores and outputs the cracking result. If the cracking fails, it retraces the password subset division strategy of the password distribution unit, the model parameters and evaluation strategy of the cracking progress estimation unit, and the regulation strategy of the GPU core control unit.

[0013] Compared with the prior art, the beneficial effects of the present invention are as follows: The cracking progress estimation unit constructs a feature analysis model by collecting a large number of encrypted password samples of drones, extracts features in multiple dimensions and performs dimensionality reduction optimization, deeply mines the internal characteristics and laws of passwords, and builds a precise foundation for cracking. During cracking, it receives the feedback from the GPU to calculate the difficulty coefficient, accurately weighs the weights according to the differences in feature vectors and dynamically corrects them, closely follows the number of tried passwords and the speed of the GPU core, finely quantifies the difficulty, reasonably allocates computing resources, improves the cracking efficiency, and finally dynamically adjusts the predicted remaining cracking time according to the total password amount, the tried password amount and the difficulty coefficient to ensure the efficient progress of the cracking plan. Brief Description of the Drawings

[0014] Figure 1 is the overall block diagram of the present invention; Figure 2 is the overall flowchart of the present invention.

[0015] The meanings of the reference numerals in the figure are as follows: 1. Password distribution unit; 2. Cracking progress estimation unit; 21. Model construction module; 22. Password cracking module; 23. Difficulty evaluation module; 3. GPU core control unit. Detailed Embodiments

[0016] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts shall fall within the protection scope of the present invention. Embodiment 1

[0017] The present invention provides a drone remote control system and method based on radio cracking technology. Please refer to Figure 1 as shown, which includes a password distribution unit 1, a cracking progress estimation unit 2, and a GPU core control unit 3; The password distribution unit 1 is configured with a password database for storing password combinations of common encryption types of drones. The database divides the password combinations into multiple password subsets based on a password distribution algorithm. When a cracking start instruction is received, the divided password subsets are respectively transmitted to the corresponding GPU cores for password attempt cracking operations.

[0018] The cracking progress estimation unit 2 establishes a password combination feature analysis model for extracting features of password combinations. During the password attempt cracking process, it receives the tried password information and attempt result information fed back from the GPU cores, compares and analyzes the features of the tried passwords with the feature data in the password combination feature analysis model, combines the current number of tried passwords and the computing speed of the GPU cores, uses a pre-constructed progress estimation algorithm to calculate the cracking difficulty coefficient of the remaining passwords, and predicts the remaining cracking time based on the cracking difficulty coefficient and the current cracking speed, and transmits the prediction result to the GPU core control unit 3.

[0019] The cracking progress estimation unit 2 includes a model construction module 21. The construction method of the password combination feature analysis model in the model construction module 21 is as follows: A large number of different types of drone encrypted password combinations are widely collected as sample data from publicly available drone security research data sets, actually captured drone encrypted communication cases, and simulated generated password combinations. For each password combination, comprehensive multi-dimensional feature extraction is performed, and the password length is calculated, which is one of the key factors directly affecting the cracking difficulty. A long password theoretically has more combination possibilities.

[0020] The principal component analysis algorithm is used to process high-dimensional feature data. First, the covariance matrix of the feature data is calculated, which reflects the correlation between the features. Then the eigenvalues ​​and eigenvectors of the covariance matrix are solved, and the eigenvalues ​​are sorted. The eigenvectors corresponding to the eigenvalues ​​with a cumulative contribution rate of more than 85% are selected to form a transformation matrix. The high-dimensional feature data is projected into the newly constructed low-dimensional space to achieve dimensionality reduction while retaining the main information, obtaining key features, and greatly reducing the subsequent model training calculation amount and overfitting risk. The focus is on the feature dimensions that play a key role in the difficulty of cracking, thereby improving the model training efficiency and generalization ability.

[0021] Based on the key features after dimensionality reduction, support vector machine is selected as the machine learning algorithm to carry out model training. The sample data is randomly divided into training set and test set in a ratio of 7:3. For the support vector machine model, radial basis function is used as the kernel function, which can effectively handle nonlinear relationships and adapt to the complex distribution of password characteristics. The model parameters are optimized through grid search method, and the combination is exhaustively enumerated within the set parameter range. The accuracy is used as the evaluation indicator to find the optimal parameter combination that enables the model to achieve the predetermined accuracy on the test data set. An analysis model is constructed to accurately evaluate the difficulty characteristics of password combination cracking, laying a solid foundation for cracking progress estimation, effectively distinguishing between easy-to-crack and difficult-to-crack password modes, and facilitating the rational allocation of resources and the efficient formulation of cracking strategies.

[0022] The cracking progress estimation unit 2 includes a password cracking module 22. The calculation method of the cracking difficulty coefficient in the password cracking module 22 is as follows: For the received attempted password information, extract its password feature vector , and the standard feature vector output by the password combination feature analysis model Compare and calculate the difference By quantifying the differences in password features, the uniqueness of the password can be accurately measured, and the difference between the password and common easy-to-crack patterns can be quickly located, providing intuitive basic data for cracking difficulty assessment and clarifying the subsequent cracking direction.

[0023] Then the hierarchical analysis method is used to determine the weight of each feature dimension in the cracking difficulty assessment ,The analytic hierarchy method can integrate expert ,experience and data statistics to scientifically assign weights, reasonably weigh the ,impact of each feature on the difficulty of cracking, avoid excessive ,interference of a single feature in the evaluation, improve the accuracy of the ,cracking difficulty coefficient, and adapt to a diverse combination of password ,features.

[0024] Calculate the cracking difficulty coefficient ; Accurately reflect the degree of password deviation from crackability, concisely and effectively quantify the core elements of cracking difficulty, provide reliable initial evaluation indicators, and help plan cracking resource investment and time arrangements.

[0025] Based on the number of passwords tried and the computing speed of the GPU core Dynamically correct the difficulty coefficient, and the correction formula is ; Since the remaining distribution of the password changes with the attempts during the cracking process, considering these factors can adapt to the dynamic changes, flexibly adjust the difficulty coefficient according to the actual situation, ensure the accuracy of the prediction, continuously optimize the cracking strategy, avoid resource misallocation, and improve the overall cracking efficiency and success rate.

[0026] Among them, the method for predicting the remaining cracking time in the password cracking module 22 is as follows: Let the total number of passwords be , and the number of passwords that have been attempted be . According to the current cracking difficulty coefficient and the computing speed of the GPU core , calculate the expected number of passwords cracked per unit time ; The cracking difficulty coefficient reflects the complexity of password cracking, and the GPU computing speed determines the computing power. The combination of the two can reasonably estimate the amount of passwords cracked per unit time, provide the core parameters for the subsequent prediction of the remaining time, initially frame the cracking progress rhythm from the theoretical level, quantify the cracking ability based on the existing state, and make the prediction of the remaining time have a clear calculation starting point and basis, and build a basic framework for predicting the cracking time.

[0027] Taking every 10 minutes as a time period, collect the GPU core cracking speed data of the past 10 time periods , and use the least squares method to calculate the slope and intercept of the trend line of the data sequence, and analyze the fluctuation characteristics and trend of the data. Select the smoothing coefficient according to the degree of fluctuation, accurately determine that it can optimize the utilization of historical data, improve the stability and accuracy of the cracking speed prediction, filter out noise interference, calibrate the model parameters for accurately predicting the future cracking speed, ensure that the prediction fits the actual change trend of the cracking speed, and improve the reliability of the prediction of the remaining cracking time.

[0028] Based on the determined smoothing coefficient , predict the future cracking speed ; is the cracking speed predicted at the previous moment, and then calculate the predicted remaining cracking time ; Since the cracking speed is dynamically affected by various factors, the exponential smoothing prediction combined with real-time data correction can capture the change trend of the speed, dynamically adjust the predicted cracking speed, timely reflect the system performance fluctuation and the change of password characteristics, make the prediction of the remaining time real-time accurate, provide a dynamic and accurate time plan for the whole process of the cracking task, help to reasonably allocate computing resources, adjust the cracking strategy, improve the overall cracking efficiency and success rate, ensure that the cracking task is completed within the optimal time frame, and enhance the controllability and timeliness of the system cracking process.

[0029] The password cracking module 22 uses an uncertainty factor compensation mechanism when predicting the remaining cracking time, as follows: A set of uncertainty factors is established, covering the update frequency of the drone encryption algorithm and the external electromagnetic interference intensity. The encryption algorithm update frequency is divided into three levels: low, medium, and high, based on industry dynamics, manufacturer release rules, and historical data statistical analysis and evaluation, corresponding to the quantitative score intervals [0,30], [31,60], and [61,100], respectively. The external electromagnetic interference intensity is converted into normalized by real-time data collection of electromagnetic monitoring equipment deployed around the cracking environment. , 0 means no interference, 1 means strong interference causing serious signal distortion, accurately captures key sources of uncertainty, comprehensively covers environmental variables that affect cracking, lays the foundation for subsequent precise compensation, clarifies the full picture of potential interference factors, and enhances the robustness of the cracking plan.

[0030] Use fuzzy comprehensive evaluation method to determine weights, evaluate the impact of various factors on cracking based on historical experience, and set encryption algorithm update weights and EMI weight ,and , quantitative scoring based on factors With weight Calculate the comprehensive impact coefficient The advantage of this method is that it integrates multiple knowledge to accurately weigh the importance of factors, avoids subjective bias, scientifically quantifies the comprehensive interference of uncertain factors, provides key parameters for accurate compensation, and optimizes the accuracy of cracking time estimation.

[0031] Reconstruct the compensation time formula ; To predict the remaining cracking time, the initial prediction time Based on the comprehensive impact coefficient Dynamic adjustment, such as hour, , that is, extend the cracking time by 20%, and the system monitors the changes of factor sets in real time and , dynamically compensate for uncertain interference, ensure that the cracking task is completed on time, enhance the flexibility of cracking planning, resist complex environmental fluctuations, and ensure that the system operates stably and efficiently under the impact of uncertain factors.

[0032] The cracking progress estimation unit 2 includes a difficulty assessment module 23. The difficulty assessment module 23 dynamically assesses the overall cracking difficulty of the password subset during the password cracking attempt, as follows: Set a clustering analysis of the attempted password subset to be performed every 500 passwords cracked. When the evaluation period is reached, collect the attempted password information and the corresponding cracking result data fed back by the GPU core during this period, and construct a password subset data set. Each data point contains a password feature vector and a cracking result label, providing a complete data basis for subsequent clustering analysis. Timed evaluation can capture the dynamic changes in the characteristics of the password subset, balance the computational cost and real-time requirements, ensure that the system timely grasps the trend of the cracking difficulty of the password subset, and accurately determine the timing of resource allocation.

[0033] Perform clustering operations on the password subset data set using the K-Means algorithm. First, determine the number of clusters based on the previous data exploration. , randomly initialize cluster centers, calculate the Euclidean distance from each password feature vector in the data set to the cluster centers, and assign the password points to the clusters where the nearest cluster centers are located. Repeat the steps of updating the cluster centers and assigning the password points until the cluster centers converge and stabilize, forming password clusters of different difficulty levels. The K-Means algorithm is used because it is efficient in processing large-scale data and can mine the internal distribution law of the data, automatically distinguish the password difficulty levels, simplify the subsequent analysis, intuitively present the cracking difficulty structure of the password subset, and help analyze the overall difficulty characteristics.

[0034] Calculate the proportion of the number of passwords in each cluster , is the cluster number, count the cracking result data of the passwords in the cluster and calculate the average cracking difficulty coefficient , with and as weights and coefficients, weighted calculate the overall cracking difficulty index of the password subset ; Continuously monitor , if its change exceeds the set threshold, it indicates that the overall cracking difficulty of the password subset has changed significantly. Considering the cluster scale and difficulty comprehensively, comprehensively measure the overall characteristics, accurately quantify the degree of difficulty change, keenly capture the key difficulty turning points, and send accurate signals for strategy adjustment.

[0035] When it is determined that the overall cracking difficulty of the password subset has changed significantly, immediately adjust the GPU core allocation strategy. If the difficulty increases, allocate additional cores to the cracking tasks associated with the high-difficulty clusters; otherwise, reduce the allocation. At the same time, optimize the password allocation algorithm parameters, subdivide the password subset for the high-difficulty clusters, adjust the allocation priority, or adopt an allocation rule more suitable for the characteristics of high-difficulty passwords; otherwise, simplify the process for the low-difficulty clusters, adapt to the dynamic evolution of the cracking difficulty, dynamically optimize the resource configuration and cracking path, improve the efficiency, ensure that the system always cracks with an efficient strategy, enhance the adaptability and success rate of the cracking tasks, and ensure the efficient utilization of resources and the smooth progress of the cracking process.

[0036] Among them, when the difficulty assessment module 23 processes password combinations of different encryption types, the encryption type recognition and difficulty adaptation method is as follows: Build a convolutional neural network architecture, including convolutional layers, pooling layers and fully connected layers. Collect a large number of password samples of different encryption types from public password libraries, simulated encrypted data and actual UAV encryption cases, covering a variety of symmetric encryption and asymmetric encryption password combinations. Conduct data preprocessing, normalize the password character encoding into the network input format, and input the password combination in the form of an image into the convolutional neural network. The convolutional layer extracts local features, the pooling layer reduces the dimension and abstracts, the fully connected layer integrates the features for classification, and the backpropagation algorithm is used to optimize the network weights according to the encryption type. After thousands of iterations, the network can accurately identify the encryption type feature pattern. The convolutional neural network is good at capturing the spatial structure features of data, accurately identifying the encryption type, laying the foundation for accurate assessment, quickly and accurately determining the encryption root of the password, and guiding the adaptation of cracking strategies.

[0037] For symmetric encryption types, set sub-models according to the key length, such as AES-128 and AES-256. The longer the key, the exponentially increasing cracking space. The sub-model takes the key length as a key variable and calculates the difficulty coefficient by combining general features such as password length and character distribution. For asymmetric encryption, consider the complexity of the algorithm's mathematical principle. For example, the difficulty of RSA factoring large integers increases superlinearly with the key bit length, and ECC is based on the elliptic curve discrete logarithm problem. The sub-model constructs a functional relationship with the core mathematical parameters of the algorithm to evaluate the difficulty, integrates the general features of the password feature analysis model, and the encryption characteristics determine the essential differences in cracking, accurately quantifying the exclusive difficulty, providing customized and accurate difficulty assessment, and improving the rationality and efficiency of cracking planning.

[0038] When a new password combination is input, after the convolutional neural network identifies the encryption type, it immediately switches to the corresponding sub-model. The sub-model receives the character frequency and continuous character rules extracted by the password feature analysis model, and calculates the cracking difficulty coefficient in combination with its own encryption characteristic parameters. When predicting the remaining cracking time, the sub-model uses the dynamic time estimation algorithm to calculate according to the difficulty coefficient, the current GPU core calculation speed, and the progress of the cracked password. Each model collaborates to achieve full-process adaptive optimization, dynamically adapt to encryption diversity, seamlessly connect to improve cracking coherence and accuracy, ensure efficient cracking in complex encryption environments, reduce time costs and resource consumption, and enhance the cracking robustness and universality of the system.

[0039] In the password cracking task, the cracking difficulties of different password combinations vary greatly. If the working frequency of the unified GPU core is used, it will cause energy waste and accelerated hardware loss when facing simple passwords, and may lead to a long cracking time due to insufficient computing power when dealing with complex passwords. Dynamically adjusting the frequency according to the remaining cracking time predicted by the cracking progress estimation unit 2 can accurately match the computing power of the GPU core with the password cracking requirements, ensuring efficient use of resources and efficient progress of tasks.

[0040] The GPU core control unit 3 receives the password subset transmitted by the password allocation unit 1, and distributes it to the corresponding GPU core for processing. During the cracking process, the operating frequency of the GPU core is dynamically adjusted according to the remaining cracking time transmitted by the cracking progress estimation unit 2. The precise frequency control significantly improves energy efficiency. When there is plenty of remaining time and the password difficulty is low, the frequency is reduced to reduce unnecessary energy consumption and extend the service life of the hardware. When time is tight or the password difficulty is high, the frequency is quickly increased to release powerful computing power in a short time, speed up the cracking process, avoid cracking failure due to time exhaustion, enhance the system's adaptability and flexibility to complex tasks, and achieve performance and energy consumption balance optimization.

[0041] When a successful cracking signal is received, the GPU core is controlled to stop the password attempt operation and store and output the cracking results. The stored results provide a data foundation for subsequent analysis, which is convenient for analyzing password patterns, evaluating system performance, and summarizing cracking experience to optimize strategy algorithms. The output results can intuitively present the cracking results, meet the user's demand for instant information acquisition, and help them take quick follow-up actions.

[0042] In the present invention, the password allocation unit 1 accurately divides the password database combination into subsets according to a specific algorithm, and assigns them to the GPU core to carry out cracking operations. The cracking progress estimation unit 2 builds a model to deeply analyze the password characteristics through multi-module collaboration. The model construction module 21 collects samples, reduces the dimension to extract key features and trains and optimizes. The password cracking module 22 quantifies the difference weighted coefficients, dynamically corrects and integrates multiple factors to predict the remaining time. The difficulty assessment module 23 clusters subsets and accurately evaluates and adapts based on the encrypted sub-model. The GPU core control unit 3 intelligently adjusts the frequency according to the estimated remaining time. After the cracking is successful, the results are stored and output to improve the cracking efficiency. Example 2

[0043] The second object of the present invention is to provide a method for implementing any one of the above-mentioned UAV remote control systems based on radio cracking technology, comprising the following steps: S1, system startup, password allocation unit 1 loads common encryption password combinations from the password database, divides them into subsets according to the password allocation algorithm, and then allocates them to the corresponding cores according to the GPU core performance and load balancing principle.

[0044] S2, the cracking progress estimation unit 2 first extracts multi-dimensional features for each password combination to obtain high-dimensional feature data, then uses the principal component analysis algorithm to process the high-dimensional feature data, composes a transformation matrix, and projects it to the high-dimensional space to obtain key features. Finally, based on the key features, the support vector machine algorithm is selected to construct a password combination cracking difficulty feature analysis model, and the cracking progress estimation process is started to receive the attempted password information and results fed back by the GPU core.

[0045] S3. The GPU core conducts password cracking work based on the allocated password subset, and feeds back the tried password information and the trial results in real time. During the password cracking process, the cracking progress estimation unit 2 uses a convolutional neural network to identify the encryption type of the new password combination, and switches the cracking difficulty evaluation sub-model according to the result. The sub-model calculates the cracking difficulty coefficient and predicts the remaining cracking time in combination with the password features.

[0046] S4. When the GPU core cracks successfully, the GPU core control unit 3 stores and outputs the cracking result. If the cracking fails, the password subset division strategy of the password allocation unit 1, the model parameters and evaluation strategy of the cracking progress estimation unit 2, and the regulation strategy of the GPU core control unit 3 are traced back.

[0047] The above shows and describes the basic principles, main features and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited by the above embodiments. The above embodiments and the descriptions in the specification are only preferred examples of the present invention, and are not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed. The scope of the present invention claimed is defined by the appended claims and their equivalents.

Claims

1. A UAV remote control system based on radio cracking technology, characterized in that: It includes a password distribution unit (1), a cracking progress estimation unit (2) and a GPU core control unit (3); The password allocation unit (1) is configured with a password database for storing common encryption type password combinations of drones, the database divides the password combinations into multiple password subsets based on a password allocation algorithm, and when a cracking start instruction is received, the divided password subsets are respectively transmitted to corresponding GPU cores for password cracking attempts; The cracking progress estimation unit (2) establishes a password combination feature analysis model for extracting features from the password combination, receives information about attempted passwords and attempt result information fed back from the GPU core during the password cracking attempt, performs comparative analysis based on the features of the attempted passwords and feature data in the password combination feature analysis model, calculates the cracking difficulty coefficient of the remaining passwords using a pre-built progress estimation algorithm in combination with the number of currently attempted passwords and the computing speed of the GPU core, and predicts the remaining cracking time based on the cracking difficulty coefficient and the current cracking speed, and transmits the prediction result to the GPU core control unit (3); The GPU core control unit (3) receives the password subset transmitted by the password allocation unit (1) and allocates it to the corresponding GPU core for processing. During the cracking process, the operating frequency of the GPU core is dynamically adjusted according to the remaining cracking time transmitted by the cracking progress estimation unit (2). When a cracking success signal is received, the GPU core is controlled to stop the password attempt operation and the cracking result is stored and output.

2. The UAV remote control system based on radio cracking technology according to claim 1 is characterized in that: The cracking progress estimation unit (2) comprises a model construction module (21). The construction method of the password combination feature analysis model in the model construction module (21) is as follows: Collect a large number of different types of drone encryption password combinations as sample data, perform multi-dimensional feature extraction on each password combination to obtain high-dimensional feature data, and use the principal component analysis algorithm to reduce the dimensionality of the high-dimensional feature data to obtain key features; Based on key features, machine learning algorithms are used to train the model until the model reaches a predetermined accuracy rate on the test data set, and an analysis model is constructed to evaluate the difficulty characteristics of password combination cracking.

3. The UAV remote control system based on radio cracking technology according to claim 2 is characterized in that: The cracking progress estimation unit (2) comprises a password cracking module (22). The method for calculating the cracking difficulty coefficient in the password cracking module (22) is as follows: For the received attempted password information, extract its password feature vector and the standard feature vector output by the password combination feature analysis model, and calculate their difference value; The hierarchical analysis method is used to determine the weight of each feature dimension in the cracking difficulty assessment, and the weight and difference value are substituted into the formula to calculate the cracking difficulty coefficient. The cracking difficulty coefficient is used to reflect the degree of deviation of the password from the easy-to-crack mode. At the same time, the difficulty coefficient is dynamically corrected according to the number of passwords tried and the GPU core computing speed.

4. The UAV remote control system based on radio cracking technology according to claim 3 is characterized in that: The remaining cracking time prediction method in the password cracking module (22) is as follows: Assume the total number of passwords is , the number of passwords tried is , according to the current cracking difficulty coefficient and GPU core computing speed , calculate the expected number of cracked passwords per unit time ; The exponential smoothing prediction method is used to predict and adjust the future cracking speed based on historical cracking speed data. , calculate the smoothing coefficient , predicting future cracking speed ;in The cracking speed predicted at the last moment, and the remaining cracking time is calculated based on this .

5. The UAV remote control system based on radio cracking technology according to claim 4 is characterized in that: The cracking progress estimation unit (2) comprises a difficulty assessment module (23). The difficulty assessment module (23) dynamically assesses the overall cracking difficulty of a password subset during the password cracking attempt process. Specifically, the method is as follows: Cluster analysis is performed regularly on subsets of attempted passwords, using the K-Means algorithm to cluster the passwords into clusters of different difficulty levels; Calculate the proportion of passwords in each cluster and the average cracking difficulty coefficient, and calculate the overall cracking difficulty index of the password subset based on the proportion and difficulty coefficient. If the overall cracking difficulty index exceeds the preset threshold, the overall cracking difficulty of the password subset is determined to be high.

6. The UAV remote control system based on radio cracking technology according to claim 5 is characterized in that: When the difficulty assessment module (23) processes different encryption type password combinations, the encryption type identification and difficulty adaptation method are as follows: A convolutional neural network is used to identify the encryption type of the password combination. The network is trained with password samples of different encryption types to identify the encryption algorithm feature patterns of the password combination. For different encryption types, corresponding cracking difficulty assessment sub-models are constructed. The sub-models work together with the password combination feature analysis model based on the symmetric encryption key length. When the new password combination encryption type is identified, it switches to the corresponding sub-model to calculate the cracking difficulty coefficient and predict the remaining cracking time.

7. The UAV remote control system based on radio cracking technology according to claim 4 is characterized in that: The password cracking module (22) has an uncertainty factor compensation mechanism when predicting the remaining cracking time, which is specifically as follows: Establishing a set of uncertain factors, wherein the set of uncertain factors includes the update frequency of the drone encryption algorithm and the intensity of external electromagnetic interference; The fuzzy comprehensive evaluation method is used to quantify the influence of each factor and determine the weight. The comprehensive influence coefficient is calculated based on the score and weight. Finally, the compensation time formula is constructed based on the predicted remaining cracking time.

8. A method for implementing a remote control system for an unmanned aerial vehicle based on radio cracking technology according to any one of claims 1 to 7, characterized in that: The following steps are involved: S1, system startup, password allocation unit (1) loads common encryption password combinations from the password database, divides them into subsets according to the password allocation algorithm, and then allocates them to corresponding cores according to the GPU core performance and load balancing principle; S2, the cracking progress estimation unit (2) first extracts multi-dimensional features from each password combination to obtain high-dimensional feature data, then uses the principal component analysis algorithm to process the high-dimensional feature data to form a transformation matrix, and projects it into a high-dimensional space to obtain key features. Finally, based on the key features, the support vector machine algorithm is used to construct a password combination cracking difficulty feature analysis model, and the cracking progress estimation process is started to receive the attempted password information and results fed back by the GPU core; S3, the GPU core performs password cracking based on the assigned password subset and provides real-time feedback on the password information and the results of the attempts. During the password cracking process, the cracking progress estimation unit (2) uses a convolutional neural network to identify the encryption type of the new password combination and switches the cracking difficulty assessment sub-model based on the result. The sub-model calculates the cracking difficulty coefficient and predicts the remaining cracking time based on the password features. S4. When the GPU core cracking succeeds, the GPU core control unit (3) stores and outputs the cracking result. If the cracking fails, the password subset division strategy of the password allocation unit (1), the model parameters and evaluation strategy of the cracking progress estimation unit (2), and the control strategy of the GPU core control unit (3) are backtracked.

Images