Method and system for hevc video selective encryption based on multi-objective tribe competition and member cooperation optimization

Abstract

The application discloses a HEVC video selective encryption method based on multi-target tribe competition and member cooperation optimization, constructs a three-dimensional parameterized chaotic system and takes a Lyapunov index, a correlation dimension and a 0-1 test value as optimization targets to establish a multi-target optimization chaotic model; a multi-target tribe competition and member cooperation algorithm is used to iteratively optimize the model, optimal chaotic parameter combinations are obtained, and an optimized three-dimensional parameterized chaotic system is constructed; the optimized chaotic system is used to generate a multi-dimensional chaotic sequence according to a key, target syntax elements are extracted in an HEVC encoding process, then the chaotic sequence is converted into a pseudo-random key stream, selective encryption is implemented on the target syntax elements, CABAC entropy encoding and code stream packaging are completed, and an encrypted HEVC video code stream is generated. The application automatically optimizes three-dimensional chaotic parameters with optimal chaotic dynamic characteristics through a multi-target algorithm, and combines HEVC syntax elements, so that video encryption with high security strength and low calculation complexity is realized.

Description

Technical Field

[0001] This application relates to the field of information security technology, specifically to a selective encryption method and system for HEVC video based on multi-target tribal competition and member cooperation optimization. Background Technology

[0002] With the rapid development of the internet and multimedia technologies, high-definition and even ultra-high-definition video has become the mainstream carrier of information dissemination. The High Efficiency Video Coding (HEVC) standard, with its superior compression performance, is widely used in scenarios such as video conferencing, telemedicine, online education, and digital media distribution. However, video data faces security risks of illegal theft, tampering, and dissemination during transmission over open networks. Therefore, efficient and reliable encryption protection for HEVC video content, especially high-value or privacy-related data, is crucial.

[0003] Chaotic systems, due to their initial value sensitivity, long-term unpredictability, and quasi-randomness, have a natural connection with cryptography and have been widely used in video encryption. However, existing chaos-based video encryption schemes have many drawbacks. On the one hand, classical low-dimensional chaotic systems (such as the Logistic map) have simple structures, but their chaotic intervals are discontinuous and have periodic windows, resulting in a limited key space and weak security, making them vulnerable to attacks such as phase space reconstruction. On the other hand, although high-dimensional chaotic systems have more complex dynamic behaviors, their chaotic performance (such as Lyapunov exponent, correlation dimension, randomness, etc.) is highly dependent on the selection of system parameters. Traditional methods often rely on manual trial and error or empirically preset parameters, making it difficult to find the optimal trade-off among multiple conflicting chaos evaluation metrics (such as maximizing the Lyapunov exponent while simultaneously satisfying high correlation dimensions). This results in chaotic systems failing to achieve their maximum theoretical performance, thereby affecting the security of encryption algorithms. Furthermore, existing swarm intelligence algorithms for optimizing chaotic parameters are mostly single-objective optimization algorithms, or suffer from low search efficiency and a tendency to get trapped in local optima. They cannot effectively solve complex parameter optimization problems involving multiple objectives and high dimensions. Moreover, existing video encryption methods mostly use full-frame encryption, which incurs huge computational overhead and is difficult to meet the requirements of real-time transmission of high-definition video.

[0004] Therefore, there is an urgent need for an HEVC video selective encryption method that can automatically find the best comprehensive chaotic performance parameters and deeply integrate with HEVC video coding characteristics, while taking into account high security, high efficiency and video coding compatibility. Summary of the Invention

[0005] Purpose of the Invention: In order to overcome the above shortcomings, the purpose of this application is to provide a selective encryption method and system for HEVC video based on multi-objective tribal competition and member cooperation optimization. By automatically optimizing the three-dimensional chaotic parameters with optimal chaotic dynamic characteristics through multi-objective algorithms and combining the importance differences of HEVC syntax elements, video encryption that balances extremely high security strength and low computational complexity can be achieved.

[0006] To address the aforementioned technical problems, this application provides a selective encryption method for HEVC video based on multi-objective tribal competition and member cooperation optimization, comprising: S1: Construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, with the preset chaotic performance index as the optimization objective and the preset parameters to be optimized as the decision variables, construct a multi-objective optimized chaotic model. S2: The multi-objective tribal competition and member cooperation algorithm is used to iteratively optimize and solve the multi-objective optimized chaotic model to obtain a non-dominated solution set that satisfies the Pareto optimality condition. The optimal compromise solution is selected from the non-dominated solution set and substituted into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system. S3: An optimized three-dimensional parametric chaotic system is adopted. A multi-dimensional chaotic sequence is generated iteratively based on the initial value of the key and the input video frame is divided into coding units during the HEVC encoding process. Then, the syntax elements including at least the sign bits of the transform coefficients, the sign bits of the motion vector difference, and the intra-frame prediction mode index are parsed. S4: According to the preset selective encryption rules, the multidimensional chaotic sequence is converted into a pseudo-random key stream with the same length as the syntax element, and the syntax element is encrypted using a preset encryption method to generate an encrypted syntax element. S5: Rewrite the encrypted syntax elements into the HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

[0007] As a preferred embodiment of this application, in S1, the method for constructing a three-dimensional parameterized chaotic system includes the following steps: S11: Establish a three-dimensional nonlinear dynamic model, treat the state variables as three independent dimensions in the system state space, and construct a three-dimensional chaotic equation system containing nonlinear coupling terms, state feedback terms, and parameter modulation terms. S12: Set preset parameters to be optimized as control parameters and establish the coupling relationship between each control parameter and the state variable respectively. Then, establish parameter boundary constraints on the preset parameters to be optimized and map the value range of each parameter to be optimized to the search space of the multi-objective optimization algorithm. S13: Perform numerical integration calculations based on the preset initial state to obtain the chaotic trajectory sequence formed by the change of state variables over time, and use the chaotic trajectory sequence to construct a time series dataset. Then, analyze the time series dataset from the dimensions of phase space distribution characteristics, trajectory traversal characteristics, attractor complexity characteristics, and randomness characteristics to establish a mapping relationship between parameter space and chaotic performance indicators. S14: Set the preset parameters to be optimized as a decision vector and use the decision vector as the optimized individual in the multi-objective tribal competition and member cooperation algorithm. Then, calculate the corresponding target evaluation index containing Lyapunov index, correlation dimension and 0-1 test value according to each optimized individual. S15: Construct a three-dimensional parameterized chaotic system based on the state variables, control parameters, parameter constraints, and target evaluation indicators.

[0008] As a preferred embodiment of this application, in S1, the method for constructing a multi-objective optimization chaotic model includes the following steps: S16: The decision variables in the three-dimensional parameterized chaotic system are used as the search objects in the multi-objective optimization algorithm, and numerical solutions are performed on the three-dimensional parameterized chaotic system for each set of decision variables to obtain the dynamic evolution trajectory of the state variables within a preset time interval. S17: Calculate the maximum Lyapunov exponent and correlation dimension of the chaotic system based on the dynamic evolution trajectory, and perform a 0-1 chaos test based on the dynamic evolution trajectory to generate a 0-1 test value; S18: Using the maximum Lyapunov exponent, correlation dimension, and 0-1 test value as the optimization objective function, construct a multi-objective optimization chaotic model consisting of a decision variable space, an optimization objective function space, and parameter constraints.

[0009] As a preferred embodiment of this application, in S2, the method for iteratively optimizing and solving the multi-objective chaotic optimization model using a multi-objective tribal competition and member cooperation algorithm includes the following steps: S21: Randomly generate multiple candidate parameter vectors according to the value range of the decision variables and use the candidate parameter vectors as initial individuals to construct an initial population. Then, divide the initial population into multiple tribes containing multiple optimized individuals and assign corresponding decision variables to each optimized individual. S22: Calculate the corresponding Lyapunov index, association dimension, and 0-1 test value for each optimized individual, generate multi-objective fitness evaluation results, and perform fast non-dominated sorting based on the multi-objective fitness evaluation results to divide the optimized individuals in the population into multiple non-dominated levels. S23: Calculate the crowding distance for optimized individuals in the same non-dominated level and filter optimized individuals according to the non-dominated level and crowding distance, and store the non-dominated solutions that meet the preset conditions to the external file; S24: Use the external archive to save and update the Pareto optimal solution obtained in the current iteration, wherein when the capacity of the external archive reaches a preset upper limit, uniformly distributed non-dominated solutions are preferentially retained according to the crowding distance.

[0010] As a preferred embodiment of this application, in S3, the method for iteratively generating a multidimensional chaotic sequence using an optimized three-dimensional parameterized chaotic system based on an initial value generated by a key includes the following steps: S31: Generate initial state parameters of the three-dimensional parameterized chaotic system based on the user-input key, initialization vector, and associated information of the video to be encrypted, and start the three-dimensional parameterized chaotic system using the initial state parameters to perform a preset number of pre-iteration operations; S32: After completing the pre-iteration, continue to perform iterative calculations to obtain the chaotic trajectory sequence corresponding to the state variable and perform normalization, quantization and bit mapping on the chaotic trajectory sequence to generate a pseudo-random sequence that meets the encryption requirements. S33: Based on the data length of the syntax elements in the HEVC encoding process, the pseudo-random sequence is truncated, recombined, or expanded to generate a key stream of corresponding length. Based on the key stream, a multidimensional chaotic sequence is generated using an optimized three-dimensional parameterized chaotic system.

[0011] As a preferred embodiment of this application, in S3, the method for dividing the input video frame into coding units during HEVC encoding and then parsing syntax elements including at least transform coefficient sign bits, motion vector difference sign bits, and intra-frame prediction mode index includes the following steps: S34: Perform HEVC encoding on the input video frame and acquire the corresponding video bitstream data during the encoding process; S35: Perform syntax parsing on the video stream data, extract syntax elements from the HEVC stream, and determine the target syntax elements for implementing selective encryption.

[0012] In a preferred embodiment of this application, the target syntax element includes a non-zero transform coefficient sign bit, a motion vector difference sign bit, and an intra-frame prediction mode index; wherein, sign bit encryption is performed on the non-zero transform coefficient sign bit, direction information encryption is performed on the motion vector difference sign bit, and mode index encryption is performed on the intra-frame prediction mode index.

[0013] As a preferred embodiment of this application, in step S4, the method of encrypting the syntax element using a preset encryption method to generate the encrypted syntax element includes the following steps: S41: Calculate the bit length information corresponding to the target syntax element and determine the corresponding key stream length based on the bit length information; S42: Extract a data sequence corresponding to the length of the target syntax element from the multidimensional chaotic sequence and convert the data sequence into a binary pseudo-random key stream; S43: Establish a corresponding keystream mapping relationship based on the type of the target syntax element in the HEVC codestream, so that different types of target syntax elements correspond to keystream data at different positions; S44: The pseudo-random key stream and the target syntax element are operated bit by bit using the XOR operation method, while keeping the content of the unselected syntax elements unchanged, and only the target syntax element is encrypted to generate an encrypted syntax element.

[0014] As a preferred embodiment of this application, in S5, the method of rewriting the encrypted syntax element into HEVC encoding and performing CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax element includes the following steps: S51: Replace the original syntax element at the corresponding position during HEVC encoding with the encrypted syntax element; S52: The replaced syntax elements and the unencrypted syntax elements are subjected to CABAC entropy encoding to complete context modeling and binary arithmetic encoding. S53: Perform NAL unit encapsulation and bitstream reassembly on the entropy-encoded syntax elements to generate an encrypted video bitstream that conforms to the HEVC standard syntax structure.

[0015] This application also provides an HEVC video selective encryption system based on multi-objective tribal competition and member cooperation optimization to implement the above method, comprising: The three-dimensional chaos modeling module is used to construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, with preset chaotic performance indicators as optimization objectives and the preset parameters to be optimized as decision variables, a multi-objective optimization chaotic model is constructed. The parameter optimization module is used to iteratively optimize and solve the multi-objective optimized chaotic model using a multi-objective tribal competition and member cooperation algorithm, obtain a non-dominated solution set that satisfies the Pareto optimality condition, select the optimal compromise solution from the non-dominated solution set, and substitute it into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system. The syntax parsing module is used to generate a multidimensional chaotic sequence iteratively based on the initial value of the key using an optimized three-dimensional parametric chaotic system. During the HEVC encoding process, the input video frame is divided into coding units, and then the syntax elements, including at least the sign bits of the transform coefficients, the sign bits of the motion vector difference, and the intra-frame prediction mode index, are parsed. A selective encryption module is used to convert the multidimensional chaotic sequence into a pseudo-random key stream with the same length as the syntax element according to a preset selective encryption rule, and to encrypt the syntax element using a preset encryption method to generate an encrypted syntax element. The bitstream reconstruction module is used to rewrite the encrypted syntax elements into HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

[0016] The technical solution described in this application has the following advantages over the prior art: 1. This application utilizes the powerful multi-objective optimization capability of multi-objective tribal competition and member cooperation algorithms to find the optimal balance point among three mutually exclusive objectives: Lyapunov exponent, correlation dimension, and 0-1 test value. This results in the optimized three-dimensional parameterized chaotic system being significantly superior to existing systems based on empirical parameter settings in terms of ergodicity, randomness, and phase space complexity, thus providing a guarantee for high-security encryption from the source.

[0017] 2. This application, through in-depth analysis of HEVC encoding characteristics, selects only transform coefficient symbols and motion vector symbols, which are crucial for decoding and recovering the image, for encryption. This selective strategy ensures that the amount of encrypted computation data accounts for only a very small portion of the total bitstream (usually less than 5%), yet the entire video image becomes completely unrecognizable due to the loss of core information. This achieves high security while significantly reducing encryption latency, meeting the needs of real-time high-definition video applications.

[0018] 3. The encryption operation of this application is applied directly to the syntax elements and does not change the HEVC bitstream structure and format information. The encrypted video bitstream can still be received and processed by the standard HEVC decoder (although it will be displayed as a garbled image when not decrypted), and has good network transmission and terminal adaptation capabilities.

[0019] 4. The multi-objective tribal competition and member cooperation algorithm of this application effectively balances global search and local exploitation by introducing a two-layer mechanism of tribal competition and member cooperation, adaptive parameter adjustment and Gaussian mutation strategies. Its search efficiency and convergence accuracy are better than existing mainstream multi-objective optimization algorithms. It can stably find high-quality chaotic parameter solutions and ensure the robust performance of the encryption system. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of this application. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0021] Figure 1 This is a flowchart illustrating an exemplary HEVC video selective encryption method provided in an embodiment of this application.

[0022] Figure 2 This is a flowchart illustrating an exemplary method for constructing a three-dimensional parameterized chaotic system provided in an embodiment of this application.

[0023] Figure 3 This is a flowchart illustrating an exemplary method for constructing a multi-objective optimization chaotic model provided in an embodiment of this application.

[0024] Figure 4 This is a schematic diagram of the overall process of an exemplary 3D-MOC optimization process provided in an embodiment of this application.

[0025] Figure 5 This is a schematic diagram of the module connections of an exemplary HEVC video selective encryption system provided in an embodiment of this application.

[0026] Figure 6 This is an exemplary overall process diagram of MOCTCM provided in an embodiment of this application.

[0027] Figure 7 This is a schematic diagram of an exemplary multi-objective optimization solution management strategy provided in an embodiment of this application.

[0028] Figure 8 This is a schematic diagram of an exemplary parameter dynamic adjustment strategy provided in an embodiment of this application.

[0029] Figure 9 This is an exemplary schematic diagram of Gaussian perturbation and competitive learning provided in an embodiment of this application.

[0030] Figure 10 This is a flowchart illustrating an exemplary optimization solution method provided in an embodiment of this application.

[0031] Figure 11 This is a flowchart illustrating an exemplary multidimensional chaotic sequence generation method provided in an embodiment of this application.

[0032] Figure 12 This is a flowchart illustrating an exemplary syntax element generation method provided in an embodiment of this application.

[0033] Figure 13 This is a flowchart illustrating an exemplary method for generating encrypted syntax elements provided in this application embodiment.

[0034] Figure 14 This is a schematic flowchart of an exemplary entropy encoding and bitstream encapsulation processing method provided in an embodiment of this application.

[0035] Figure 15 This is an exemplary convergence diagram of 3D-POS optimization provided in an embodiment of this application. Detailed Implementation

[0037] The embodiments of this application are described in detail below. Examples of these embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0038] To facilitate understanding of the technical solutions provided in this application, the relevant technologies and terms are explained below.

[0039] Chaotic systems, as a special class of nonlinear dynamic systems, possess characteristics such as initial condition sensitivity, unpredictable long-term behavior, and complex topological structures, making them widely applicable in fields such as information encryption, communication security, and data protection. Based on system form, chaotic systems can be divided into continuous chaotic systems and discrete chaotic systems; based on dimensionality, they can be divided into low-dimensional chaotic systems and high-dimensional chaotic systems. Among these, high-dimensional chaotic systems, due to their greater number of state variables and coupling relationships, can generate more complex dynamic behaviors and more random chaotic sequences, thus possessing greater application potential in cryptography and information security.

[0040] To evaluate the performance of chaotic systems, indices such as the Lyapunov exponent (LE), correlation dimension (CD), 0-1 test value, sample entropy, permutation entropy, and Kolmogorov entropy are commonly used to analyze the dynamic characteristics of the system. Among them, the Lyapunov exponent is used to measure the sensitivity of the system to initial conditions, the correlation dimension is used to characterize the complexity of the attractor, and the 0-1 test value is used to determine the degree of chaos of the system.

[0041] As research into chaos theory continues to deepen, traditional chaotic systems have gradually revealed problems in practical applications, such as insufficient chaotic performance, limited parameter sensitivity range, and easy degradation of dynamic behavior. For example, some classical chaotic systems only exhibit stable chaotic characteristics within a narrow parameter range. When the parameters change slightly, they are prone to entering a periodic state, which leads to a decrease in chaotic performance and limits their effectiveness in high-security application scenarios.

[0042] In recent years, research focus has gradually shifted from low-dimensional chaotic systems to three-dimensional and high-dimensional chaotic systems with higher complexity. Compared to low-dimensional chaotic systems, three-dimensional parameterized chaotic systems exhibit richer dynamic behaviors and can generate more complex chaotic sequences. However, the performance of three-dimensional parameterized chaotic systems usually requires evaluation of multiple chaotic indices, and there are complex nonlinear coupling relationships between the parameters. Different optimization objectives often constrain each other, making it difficult for traditional trial-and-error methods or single-objective optimization methods to obtain the parameter combination with the best overall performance. Therefore, how to achieve coordinated optimization among multiple chaotic performance indices to obtain a three-dimensional parameterized chaotic system with better overall performance has become an important research problem.

[0043] Multi-objective optimization problems aim to simultaneously optimize multiple conflicting objective functions. Their optimal solutions are usually represented by a set of Pareto optimal solutions rather than a unique solution. Swarm intelligence optimization algorithms have been widely used in solving multi-objective optimization problems due to their strong global search capabilities and good solution efficiency. However, existing methods such as differential evolution algorithms, artificial bee colony algorithms, and gray wolf optimization algorithms still have problems such as being prone to getting trapped in local optima, slow convergence speed, and insufficient multi-objective coordination ability in the process of optimizing parameters of complex chaotic systems.

[0044] Tribal Competition and Member Cooperation (CTCM) is a novel swarm intelligence optimization algorithm that achieves global search through tribal competition and local development through member cooperation, exhibiting good convergence performance and search capability. However, existing research mainly focuses on single-objective optimization, with relatively little research on its application in multi-objective optimization problems. Therefore, this application extends the CTCM algorithm to the field of multi-objective optimization, using the Lyapunov exponent, correlation dimension, and 0-1 test value as optimization objectives to perform collaborative optimization on a three-dimensional parameterized chaotic system. This yields a parameter combination with excellent chaotic performance, and the optimized three-dimensional parameterized chaotic system is applied to the HEVC video selective encryption process to improve the security and anti-attack capability of the video encryption system.

[0045] Therefore, for reference Figure 1 As shown in some embodiments of this application, this application provides a selective encryption method for HEVC video based on multi-objective tribal competition and member cooperation optimization. The method includes the following steps: S1: Construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, with preset chaotic performance indicators as optimization objectives and the preset parameters to be optimized as decision variables, construct a multi-objective optimized chaotic model.

[0046] The dynamic performance of a three-dimensional parametric chaotic system is influenced by multiple system parameters. Determining the optimal parameters for chaotic system performance is a typical multi-objective optimization problem. This paper designs a three-dimensional parametric chaotic system (3D-POS) and optimizes its parameters using the MOCTCM algorithm. By adjusting parameters related to the effective objective function, chaotic performance is optimized globally. The chosen 3D-POS includes nine unknown coefficients, and a representative model is constructed as follows: ; in, For state variables, For control parameters, vector These constitute 9-dimensional decision variables; a suitable objective function is needed during the optimization process to evaluate the chaotic characteristics of the mapping.

[0047] There are many precise metrics for measuring the chaotic performance of a mapping, such as Ley-Leadership (LE), 0-1 test, Sequence of Action (SE), Correlation Diagnosis (CD), Principle of Action (PE), Key-Earnings Analysis (KE), bifurcation, and trajectory analysis. In the study of multi-objective functions, this application aims to identify the three most influential objective functions in optimization and apply them to improve chaotic performance. Therefore, this application selects and optimizes the three most effective metrics: LE, CD, and 0-1 test. Then, the decision parameters of 3D-POS are determined using a multi-objective optimization algorithm to improve the following three objective values: 1. Lyapunov Index (LE) In the study of chaotic systems, the Lyapunov exponent (LE) is one of the core indicators for quantifying the dynamic characteristics of a system. It reveals whether a system exhibits chaotic behavior by describing the separation or convergence rate of neighboring trajectories in phase space. The Lyapunov exponent can measure the complexity of a dynamic system and assess its unpredictability and sensitivity to control parameters and initial values. A larger exponent indicates greater sensitivity to initial conditions, shorter predictability, more disordered dynamic behavior, and higher information generation rate. Therefore, the Lyapunov exponent can be chosen as the first objective exponent in a multi-objective optimization algorithm. Its calculation formula is: ; ; Correlation dimension (CD) is a fractal dimension used to measure the distributional complexity of data points in an attractor of a chaotic system. It describes how the number of points required to cover a given volume of phase space varies with the size of that volume. As the level of chaos or randomness in a system increases, its attractor tends to become more complex and exhibit greater self-similarity over a wide range of scales. If a nonlinear system has a positive CD, then it has a singular attractor, and a large CD implies a large spatial dimension, meaning the singular attractor possesses high randomness and singularity. For a time series of dimension m... CD is defined as: ; in The integral related to the relationship can be calculated as follows: ; Where θ(ω) is the step function, θ(ω)=0 when ω≤0; θ(ω)=1 when ω>0; ζ is the time delay, usually set to 1; sequence For the new sequence: The 0-1 test was developed to measure the growth rate of the trajectory of a chaotic system. It can serve as an efficient method for determining whether a time series exhibits chaotic behavior. It simply detects the trajectory characteristics of any sequence of a system in a two-dimensional plane; the 0-1 test value is obtained by calculating its mean square displacement growth rate; given the observed time series of a chaotic system... Perform a random projection transformation on the time series to generate two new series ( , ), expressed by the formula: ; ; Where θ is a randomly selected angle, typically a random value within (0,π); here, a 1D time series is mapped to a 2D plane, and the dynamic characteristics of the system are captured through the random angle θ. Its mean square displacement growth rate can be defined as: ; When K≈0, ( , The trajectory of K is confined to a finite region, representing a regular system; when K≈1, the trajectory will exhibit unbounded but ordered expansion, representing a completely chaotic system.

[0048] The goal of Pareto multi-objective optimization is to optimize the decision variables of 3D-POS to improve the following three objectives: ; in, , and These are the average values ​​of LE, CD, and K, respectively. The objective function for multi-objective optimization is typically a minimization function, so the optimal objective criterion for selecting 3D-POS parameters consists of the following three objective functions: .

[0049] refer to Figure 2 As shown. Specifically, in S1, the method for constructing a three-dimensional parameterized chaotic system includes the following steps: S11: Establish a three-dimensional nonlinear dynamic model, treat the state variables as three independent dimensions in the system state space, and construct a three-dimensional chaotic equation system containing nonlinear coupling terms, state feedback terms, and parameter modulation terms.

[0050] S12: Set preset parameters to be optimized as control parameters and establish the coupling relationship between each control parameter and the state variable. Then, establish parameter boundary constraints for the preset parameters to be optimized and map the value range of each parameter to be optimized to the search space of the multi-objective optimization algorithm.

[0051] S13: Perform numerical integration calculations based on the preset initial state to obtain the chaotic trajectory sequence formed by the change of state variables over time, and use the chaotic trajectory sequence to construct a time series dataset. Then, analyze the time series dataset from the dimensions of phase space distribution characteristics, trajectory traversal characteristics, attractor complexity characteristics, and randomness characteristics to establish a mapping relationship between parameter space and chaotic performance indicators.

[0052] S14: Set the preset parameters to be optimized as decision vectors and use the decision vectors as optimization individuals in the multi-objective tribal competition and member cooperation algorithm. Then, calculate the corresponding target evaluation index containing Lyapunov index, correlation dimension and 0-1 test value according to each optimization individual.

[0053] S15: Construct a three-dimensional parameterized chaotic system based on the state variables, control parameters, parameter constraints, and target evaluation indicators.

[0054] refer to Figure 3 As shown. Specifically, in S1, the method for constructing a multi-objective optimization chaotic model includes the following steps: S16: The decision variables in the three-dimensional parameterized chaotic system are used as the search objects in the multi-objective optimization algorithm. For each set of decision variables, the three-dimensional parameterized chaotic system is numerically solved to obtain the dynamic evolution trajectory of the state variables within a preset time interval.

[0055] S17: Calculate the maximum Lyapunov exponent and correlation dimension of the chaotic system based on the dynamic evolution trajectory, and perform a 0-1 chaos test based on the dynamic evolution trajectory to generate a 0-1 test value.

[0056] S18: Using the maximum Lyapunov exponent, correlation dimension, and 0-1 test value as the optimization objective function, construct a multi-objective optimization chaotic model consisting of a decision variable space, an optimization objective function space, and parameter constraints.

[0057] Multi-objective optimization problems (MOPs) involve optimizing two or more conflicting objectives. Therefore, the solution to a MOP is not a specific set of solutions, but rather a set of Pareto solutions. A Pareto solution set is the set of solutions where optimizing one objective function inevitably worsens the others. Thus, the performance of a MOP algorithm is evaluated by how well the generated Pareto curve fits the actual Pareto curve. The optimization problem involves finding all solutions in a population; finding the optimal solution for all objectives requires considering the outcome of the entire objective function. Generally, minimizing a multi-objective optimization problem (MOP) is defined as follows: ; ; in For the first objective function, For the first One objective function, For decision-making space, Let x be the lower and upper bounds of the i-th dimension, respectively. Let be the mapping function from the decision space to the target space. For n-dimensional real space, For transpose operation, Let n be an m-dimensional objective space; when n≥4, the multi-objective optimization problem is usually referred to as a high-dimensional multi-objective optimization problem (MOPs).

[0058] For multi-objective optimization problems, there is no single solution that can simultaneously optimize all objectives; rather, there exists a set of optimal solutions that balance the objectives. The following definition is essential: Definition 1 (Pareto Dominance): Let a, b∈Ω be two distinct solutions; solution a is said to dominate solution b (denoted as a≺b) if and only if the following condition is satisfied: (1) , (2) .

[0059] Definition 2 (Pareto optimality): If a solution a does not have a dominant solution b, then solution a is Pareto optimal, expressed as: .

[0060] Definition 3 (Pareto set): The Pareto set (PS) consists of all Pareto optimal solutions and can be represented as: PS = {x | x is the Pareto optimal solution in Ω}.

[0061] Definition 4 (Pareto Front): The Pareto front (PF) is the mapping of the PS in the target space, and can be represented as: .

[0062] Therefore, when solving multi-objective optimization problems, the objective optimization algorithm should consider both convergence and diversity to make the obtained solution converge toward the true Pareto front. Convergence aims to find the front of the non-dominated solution that is closest to the true Pareto front, while diversity aims to obtain a solution that is well distributed along the Pareto front.

[0063] Therefore, for reference Figure 4 As shown, Figure 4 Before optimization, the 3D-MOC system parameters were all set to 1 by default. After MOCTCM optimization, the state variables of 3D-MOC are more evenly distributed, have stronger randomness, and have a higher LE.

[0064] S2: The multi-objective tribal competition and member cooperation algorithm is used to iteratively optimize and solve the multi-objective optimized chaotic model to obtain a non-dominated solution set that satisfies the Pareto optimality condition. The optimal compromise solution is selected from the non-dominated solution set and substituted into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system.

[0065] The Tribal Competition and Member Cooperation (CTCM) algorithm is inspired by the survival behaviors of ancient human tribes, simulating two key mechanisms: intra-tribal member cooperation and inter-tribal competition. Each tribe has a leader who controls the optimal resource location. Members explore for better resources based on their own experience and the leader's instructions, with loyalty exhibiting chaotic changes over time. Loyalty is described by a sinusoidal chaotic mapping, reflecting the balance between individual autonomy and group collaboration. Tribes randomly clash over resources; weaker tribes flee in the opposite direction to stronger tribes, while stronger tribes remain unaffected. This expands the search range and avoids local optima.

[0066] CTCM is built upon the cooperative and competitive behaviors of ancient tribes, encompassing core elements such as initialization, member cooperation, tribal competition, location updating, and boundary handling. Let the total number of individuals in the primitive society be... p The number of tribes is nEach tribe contains m There are members, and the solution space has a dimension of . d Then the initial position matrix of all individuals is: ; ; in For the first n Location matrix of each tribe For the first n The first tribe m The member in the d The position of the dimension.

[0067] If the quality of an individual is assessed using its fitness value, then the fitness value of all individuals is... F It can be represented as: ; in, For the first n Fitness matrix of each tribe For the tribe m The fitness of each member.

[0068] Each individual has a velocity vector, which determines its next direction; the overall velocity matrix of all individuals. V Represented as: ; ; in, For the first n The first tribe m The member in the d The speed of the dimension.

[0069] Tribe members explore resources through cooperation, with loyalty calculation and speed updates being the core elements. Members' loyalty to the tribe leader exhibits chaotic fluctuations, described using a sinusoidal chaotic mapping: ; in U (0,1) is a uniformly random number in the interval (0,1), and loyalty is periodically reset to enhance randomness; It reflects the balance between members' independent decision-making and obedience to the leader.

[0070] Each individual has a different mindset, and will adjust their speed by combining their own experience, the leader's instructions, and inertia. This can be represented as: ; in, For the first n The first tribe m Each membert+1 velocity at time constant The inertia factor controls the influence of historical velocity. For the first t The speed of time This represents the optimal fitness position found for the member throughout the entire cycle. for t Location at any given moment The best fit location found for the tribe throughout the entire period; c 1 and c 2 These respectively represent the individual's experience factor and the leader's obedience factor; and Chaotic loyalty for individual members.

[0071] Tribes will compete with each other, leading to widespread exploration; random conflicts can occur between tribes, in which weaker tribes will flee, while stronger tribes will remain unaffected; competing tribes are randomly selected. Calculate the escape term and adjust the speed as follows: ; ; in, The escape term is related to the optimal position difference between competing tribes; Indicates the tribe escape coefficient. A chaotic random factor reflecting the escape speed. If the current tribe adapts... Weaker than competing tribes Then, the speed is reduced by the escape item to distance oneself from competing tribes. Meanwhile, the positions of tribe members are updated as follows: ; If a tribe member's location updates beyond the feasible domain, a mirror bounce occurs to ensure the individual searches within the effective range. This is represented as: .

[0072] CTCM faces challenges such as slow convergence and premature convergence when dealing with multi-objective optimization problems. Therefore, a multi-objective CTCM algorithm (MOCTCM) is proposed to improve upon these shortcomings. Based on CTCM, this algorithm divides the population into multiple tribes, balances global search and local exploitation through a two-layer search mechanism and dynamic parameter adjustment, maintains the diversity of the Pareto optimal solution set by combining external archiving and crowding distance, and enhances the algorithm's search capability by incorporating strategies such as Gaussian mutation and competitive learning. Multiple objective functions are used to optimize the problem. (See the MOCTCM algorithm flowchart for reference.) Figure 6 As shown.

[0073] MOCTCM's solution management strategy is one of its core mechanisms. Its core objective is to efficiently maintain the non-dominated solution set, balance convergence and diversity during iteration, and guide the search towards the Pareto front through strategy design. MOCTCM's solution management strategy employs a hierarchical collaborative mechanism, mainly comprising three modules: non-dominated solution hierarchies, crowding distance evaluation, and external archive management. Through hierarchical management, dynamic pruning, and cross-tribe guidance, this strategy maintains solution diversity while ensuring convergence, exhibiting better solution set distribution and convergence speed compared to traditional multi-objective algorithms for complex problems. The strategy is described in detail below: First, MOCTCM uses a fast non-dominated sorting algorithm (similar to NSGA-II) to divide the solutions into different levels. It uses the dominance relationship defined in Definition 1 above to determine the Pareto fronts. The first level consists of completely non-dominated solutions, i.e., Pareto optimal solutions, such as... Figure 7 The red curve represents the second layer, which consists of solutions dominated only by the solutions in the first layer. Subsequent layers follow the same pattern. By using non-dominated sorting, the superior and inferior levels of the solutions are clearly distinguished, guiding subsequent selection.

[0074] Then, to avoid the solution set concentrating in certain regions of the Pareto front, MOCTCM uses crowding distance to assess the uniformity of solution distribution; to obtain an estimate of the crowding around a specific solution in the population, the average distance between the two points on either side of that point is calculated according to each objective function; this value serves as an estimate of the perimeter of a cuboid with its nearest neighbor as the vertex (called the crowding coefficient); in Figure 7 In this algorithm, the crowding coefficient of the i-th solution on its front edge is the length of the cuboids surrounding it (as shown in the dashed box). The calculation of crowding coefficient ensures population diversity. Crowding distance is a key indicator in multi-objective optimization for measuring the density of solutions in the objective space, used to further distinguish the quality of solutions within the same front edge layer after non-dominated sorting. The core idea is that the more open the area around a solution (the greater the distance between adjacent solutions), the higher the crowding coefficient, and these solutions should be prioritized to maintain population diversity. Low crowding coefficients indicate solution clustering and may be discarded.

[0075] Finally, MOCTCM's external archive management mechanism is a core component of its multi-objective optimization capabilities. By maintaining and updating a solution set independent of the current population, it achieves efficient approximation and maintenance of the Pareto front. Figure 7 As shown by the green dashed line, the external archive is an independent set storing historically discovered non-dominated solutions, complementing the current iteration's population. It records all non-dominated solutions encountered by the algorithm throughout the search process, preserving historical best solutions and preventing the loss of high-quality solutions during iterations. Furthermore, it guides subsequent searches, providing reference points for generating new solutions and accelerating convergence to the true Pareto front. By periodically merging the archive with the current population, it ensures that the solution set possesses both high quality and uniform distribution.

[0076] In current conventional CTCM algorithms, the empirical factor c1, the compliance factor c2, the escape factor c3, and the tribe size m are usually set as fixed constants, failing to fully utilize their impact on algorithm performance. A brief analysis of each factor follows: c1 represents the degree to which an individual relies on its own historical experience, similar to the cognitive parameter in the Particle Swarm Optimization (PSO) algorithm. It controls the intensity with which an individual pursues its own optimal solution during the search process. When c1 is large, the individual is more inclined to use its own experience, accelerating the convergence to its own optimal solution, but this may cause the algorithm to get stuck in a local optimum too early. When c1 is small, the individual's reliance on its own experience decreases, making it more receptive to external information and enhancing its global exploration ability, but the convergence speed may be slower.

[0077] c2 represents the degree of obedience an individual has to other members of the tribe or group, similar to the social parameter in PSO, controlling the strength of an individual's following of the group's optimal solution. When c2 is large, individuals are more inclined to follow the group's optimal solution, promoting information sharing and cooperation within the tribe, accelerating overall convergence, and strengthening local development capabilities, but it may lead to premature homogenization of the group and loss of the ability to explore new solutions. When c2 is small, individual independence is enhanced, reducing dependence on the group, which helps maintain diversity and enhances global exploration capabilities, but it may lead to a decrease in the efficiency of cooperation within the tribe.

[0078] c3 represents a parameter specific to the tribal competition algorithm, used to control the tendency of individuals to leave their current tribe and explore new areas. It may manifest as the probability of an individual leaving its original tribe, the intensity of migration, or sensitivity to the external environment. A larger c3 makes individuals more likely to leave their current tribe, essentially jumping to a new area in the solution space, effectively preventing the algorithm from getting trapped in local optima, optimizing resource allocation, and improving global search efficiency. A smaller c3 makes individuals tend to stay in their original tribe, relying on cooperative search within the tribe, which may cause the algorithm to iterate repeatedly in local areas, thus lacking the ability to escape local optima. c3 also needs to be matched with c1 and c2 to maintain parameter balance. If c3 is too high and c1 / c2 is too low, individuals may migrate blindly; if c3 is too low, the algorithm may lack the ability to escape local optima.

[0079] m represents the number of individuals in the population tribes in the algorithm, determining the group's organizational structure and competition / cooperation patterns. When m is small, the population is divided into more small tribes, each exploring different solution space regions independently, resulting in parallel search effects and significantly improving global exploration capabilities and diversity. However, if m is too large, the number of individuals in each tribe will be too small, potentially reducing search efficiency due to insufficient information. When m is large, the number of individuals within a tribe is large, information exchange is sufficient, and cooperation efficiency is high, but it is easy to get trapped in local optima due to insufficient competition. A moderate m can balance competition and cooperation, with inter-tribe competition driving the survival of the fittest, eliminating inefficient tribes and retaining the experience of high-quality tribes, while intra-tribe cooperation promotes information sharing.

[0080] By appropriately adjusting the above parameters, the algorithm can achieve optimal performance between exploration and development, convergence speed, and solution quality. These parameters are typically set as fixed constants, but adaptive mechanisms can be introduced to adjust them. Assume the maximum number of iterations is G, and the current iteration number is t; when t... <G 0.3, with parameters [c1,c2,c3,m] set to [1.0,0.5,0.2,10]. In the early stages of iteration, reducing c1, c2, and m enhances inter-group competition and promotes the algorithm's global exploration capability; when G... 0.3 <t<G 0.7, parameters [c1,c2,c3,m] are set to [1.5,1.0,0.1,20]. During the middle of the iteration, the parameters are set to moderate values ​​to ensure smooth exploration and development; when t≥G 0.7, with parameters [c1,c2,c3,m] set to [2.0,1.5,0.05,30]. In the later stages of iteration, c1, c2, and m are increased, while c3 is decreased to strengthen intra-group collaboration and improve the algorithm's local exploitation capability. This effectively balances the algorithm's global search capability and local convergence capability, significantly improving the solution efficiency of complex multi-objective optimization problems. See details... Figure 8 As shown.

[0081] To further improve the robustness and optimization performance of the algorithm, MOCTCM proposes an exploration capability enhancement strategy that combines Gaussian perturbation and competitive learning; in the Iterth iteration, the Gaussian perturbation strategy is applied to the current population POP. I To generate another n individuals, we call this new set of individuals . ;for For each individual in the search, a dimension d is randomly selected, and a Gaussian random perturbation is added to that dimension. To ensure that the magnitude of the perturbation can adaptively adjust according to the range of each dimension, the magnitude of the perturbation is proportional to the search range. The calculation formula is as follows: ; In the formula, This represents the d-th dimension component of the i-th individual in the population. To increase the population after Gaussian perturbation, and Let these represent the upper and lower bounds of the j-th dimension variable, respectively. N (0,1) represents a Gaussian random variable with a mean of 0 and a standard deviation of 1; for example, Figure 9 As shown, Gaussian perturbation introduces random noise following a Gaussian distribution to mutate individuals, which performs well in local searches. Since the Gaussian distribution tends to concentrate around the mean, the perturbation is usually small-scale, which helps to perform a fine-grained search near the current location. At the same time, boundary handling ensures the legitimacy of individuals in the population. In addition, Gaussian perturbation applies the perturbation to one random dimension of each individual, rather than all dimensions. This allows for exploration of a single dimension while keeping other dimensions of the individual constant, reducing the amount of change to the individual and helping to maintain population diversity while avoiding overly drastic changes.

[0082] After obtaining the Gaussian perturbation population Subsequently, MOCTCM utilizes a competitive learning mechanism to enhance its global search capability. Competitive learning enables individual selection and elite reorganization, guiding offspring generation by comparing the fitness of individuals in the two current populations and the Gaussian perturbation population, thus helping to maintain population diversity and convergence. Specifically, in the Iter-th iteration, MOCTCM selects individuals from the current population... and Gaussian perturbation population An individual is randomly selected from the dataset, and its fitness value is calculated based on displacement density, using the following formula: ; Where X is an individual in population P. Y Let represent the nearest displacement point to X in population P, and m represent the number of targets. Therefore, the fitness value obtained by this method is the distance between an individual and its nearest displacement point. The density estimation strategy based on offset can evaluate the solution quality in terms of convergence and diversity. If the fitness of an individual in the Gaussian perturbation population is less than or equal to the fitness of the current individual (i.e., the Gaussian perturbation individual wins the competition), then the Gaussian perturbation individual is selected as the basis vector for the individual's migration direction; otherwise, the current individual is selected as the basis vector for the migration direction. The entire population, through competitive learning, can guide the group towards the Pareto front through the directional transmission of excellent solutions. Competitive pressure prompts individuals to continuously optimize, improving the convergence and diversity of solutions.

[0083] Gaussian perturbation and competitive learning mechanisms in MOCTCM, as exploration-enhancing strategies, organically combine global search and local refinement in the multi-objective solution space, effectively improving the algorithm's optimization performance. These, along with tribal competition and member cooperation in the MOCTCM algorithm, are four different but collaborative mechanisms, each playing a different role in the algorithm. When the algorithm gets stuck in local optima or the solution distribution is too dense, Gaussian perturbation can enhance population diversity, escape local optima, and expand the search range. Competitive learning can strengthen the characteristics of high-quality solutions, accelerate convergence, and maintain solution quality; it is mainly applied in scenarios with slow convergence speed and discontinuous Pareto fronts. Tribal competition can prevent premature homogenization of the algorithm, achieve parallel search, and balance exploration and exploitation; it mainly solves multimodal optimization problems and is applied when multiple regions need to be explored in parallel. Member cooperation enables information sharing, accelerates local convergence, and improves solution accuracy; it is mainly applied in local fine-grained search and situations requiring rapid utilization of existing information.

[0084] In summary, Gaussian perturbation and competitive learning are the micro-operations in MOCTCM that enhance exploration capabilities, mainly improving the quality of solutions through randomness and directed learning; while tribal competition and member cooperation are the macro-architecture of the algorithm, optimizing the search strategy through hierarchical organization and information flow; the four complement each other and work together to achieve a dynamic balance between exploration and development in multi-objective optimization.

[0085] Computational complexity is a convenient and effective tool for estimating the computational resource consumption of an algorithm, and it can help guide the application of the algorithm. First, we discuss the time complexity of MOCTCM, which generally consists of five steps: initialization, fitness evaluation, state update, exploration enhancement, and multi-objective solution optimization. Assume that n is the number of the entire MOCTCM population, m is the number of objectives to be optimized, and d is the dimension of the objectives to be optimized. The complexities of initialization, fitness evaluation, state update, exploration enhancement, and multi-objective solution optimization are O(n×d), O(n), O(n), O(n×d×m), and O(n×d), respectively. If the maximum number of iterations is considered, the overall time complexity of MOCTCM is O(n×d×m). Space complexity refers to estimating the computer storage space occupied by the algorithm during computation. According to the optimization process of the algorithm, the space computational complexity of MOCTCM includes the following parts: the initialization space complexity of the population is O(n), the Gaussian perturbation complexity is O(n), and the complexity of the non-dominated sorting based on the reference point is... The complexity of solution selection based on crowding is O(2nlog(2n)), the complexity of the learning strategy based on the competition mechanism is O(n), and the complexity of polynomial mutation is O(n); therefore, the overall space computation complexity of MOCTCM is... .

[0086] MOCTCM has the same computational complexity as algorithms such as NSGAII, MOPSO, and SPEA-II, indicating its comparability with other leading algorithms and its superiority over some popular methods, such as NSGA3 and SPEA, which have higher computational complexity. This indicates that the complexity of MOCTCM is within an acceptable range, and the use of dynamic parameter optimization and exploration enhancement strategies in this application has not increased the complexity.

[0087] refer to Figure 10 As shown, specifically, in S2, the method for iteratively optimizing and solving the multi-objective chaotic optimization model using a multi-objective tribal competition and member cooperation algorithm includes the following steps: S21: Randomly generate multiple candidate parameter vectors according to the value range of the decision variables, and use the candidate parameter vectors as initial individuals to construct an initial population. Then, divide the initial population into multiple tribes containing multiple optimized individuals and assign corresponding decision variables to each optimized individual.

[0088] S22: Calculate the corresponding Lyapunov index, association dimension, and 0-1 test value for each optimized individual, generate multi-objective fitness evaluation results, and perform fast non-dominated sorting based on the multi-objective fitness evaluation results to divide the optimized individuals in the population into multiple non-dominated levels.

[0089] S23: Calculate the congestion distance for optimized individuals in the same non-dominated level and filter optimized individuals according to the non-dominated level and congestion distance, and store the non-dominated solutions that meet the preset conditions to the external file.

[0090] S24: Use the external archive to save and update the Pareto optimal solution obtained in the current iteration, wherein when the capacity of the external archive reaches a preset upper limit, uniformly distributed non-dominated solutions are preferentially retained according to the crowding distance.

[0091] S3: An optimized three-dimensional parametric chaotic system is adopted. Based on the initial value of the key, a multi-dimensional chaotic sequence is iteratively generated and the input video frame is divided into coding units during the HEVC encoding process. Then, the syntax elements including at least the sign bit of the transform coefficient, the sign bit of the motion vector difference, and the intra-frame prediction mode index are parsed.

[0092] refer to Figure 11 As shown. Specifically, in S3, the method for iteratively generating a multidimensional chaotic sequence using an optimized three-dimensional parameterized chaotic system based on an initial value generated by a key includes the following steps: S31: Generate initial state parameters of the three-dimensional parameterized chaotic system based on the user-input key, initialization vector, and associated information of the video to be encrypted, and start the three-dimensional parameterized chaotic system using the initial state parameters to perform a preset number of pre-iteration operations.

[0093] S32: After completing the pre-iteration, continue to perform iterative calculations to obtain the chaotic trajectory sequence corresponding to the state variable, and perform normalization, quantization and bit mapping processing on the chaotic trajectory sequence to generate a pseudo-random sequence that meets the encryption requirements.

[0094] S33: Based on the data length of the syntax elements in the HEVC encoding process, the pseudo-random sequence is truncated, recombined, or expanded to generate a key stream of corresponding length. Based on the key stream, a multidimensional chaotic sequence is generated using an optimized three-dimensional parameterized chaotic system.

[0095] refer to Figure 12 As shown. Specifically, in S3, the method of dividing the input video frame into coding units during HEVC encoding and then parsing syntax elements including at least transform coefficient sign bits, motion vector difference sign bits, and intra-frame prediction mode index includes the following steps: S34: Perform HEVC encoding on the input video frame and obtain the corresponding video bitstream data during the encoding process.

[0096] S35: Perform syntax parsing on the video bitstream data, extract syntax elements from the HEVC bitstream and determine the target syntax elements for implementing selective encryption, wherein the target syntax elements include non-zero transform coefficient sign bits, motion vector difference sign bits and intra-frame prediction mode index.

[0097] Specifically, sign bit encryption is performed on the sign bits of non-zero transform coefficients, direction information encryption is performed on the sign bits of motion vector difference, and mode index encryption is performed on the intra-frame prediction mode index.

[0098] S4: According to the preset selective encryption rules, the multidimensional chaotic sequence is converted into a pseudo-random key stream with the same length as the syntax element, and the syntax element is encrypted using a preset encryption method to generate an encrypted syntax element.

[0099] refer to Figure 13 As shown. Specifically, in S4, the method of encrypting the syntax element using a preset encryption method to generate the encrypted syntax element includes the following steps: S41: Calculate the bit length information corresponding to the target syntax element and determine the corresponding key stream length based on the bit length information.

[0100] S42: Extract the data sequence corresponding to the length of the target syntax element from the multidimensional chaotic sequence and convert the data sequence into a binary pseudo-random key stream.

[0101] S43: Establish a corresponding keystream mapping relationship based on the type of the target syntax element in the HEVC codestream, so that different types of target syntax elements correspond to keystream data at different positions.

[0102] S44: The pseudo-random key stream and the target syntax element are operated bit by bit using the XOR operation method, while keeping the content of the unselected syntax elements unchanged, and only the target syntax element is encrypted to generate an encrypted syntax element.

[0103] S5: Rewrite the encrypted syntax elements into the HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

[0104] refer to Figure 14 As shown. Specifically, in S5, the method of rewriting the encrypted syntax element into HEVC encoding and performing CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax element includes the following steps: S51: Replace the original syntax element at the corresponding position during HEVC encoding with an encrypted syntax element.

[0105] S52: The replaced syntax elements and the unencrypted syntax elements are subjected to CABAC entropy encoding to complete context modeling and binary arithmetic encoding.

[0106] S53: Perform NAL unit encapsulation and bitstream reassembly on the entropy-encoded syntax elements to generate an encrypted video bitstream that conforms to the HEVC standard syntax structure.

[0107] Therefore, to verify the MOCTCM algorithm proposed in this application, experimental comparisons were conducted with several other excellent algorithms on a multi-objective function test set. Taking 3D-POS as an example, the chaotic parameter selection scheme considering LE, CD, and 0-1 test proposed in this application was optimized. The experimental simulations in this part were conducted on a laptop running Windows 10 with 16GB of RAM and an Intel(R) Core(TM) i5-10210U CPU. To ensure the unbiasedness of the comparisons, all experiments were conducted using PlatEMO v4.0 and MATLAB R2023a programs and followed the parameters recommended by various comparison methods.

[0108] MOCTCM will be compared with five other multi-objective algorithms proposed in the last three years: Multi-objective Manta Ray Foraging Optimizer (MOMRFO, 2023), Multi-objective Chimpanzee Optimizer (MOChOA, 2023), Multi-objective Boxing Match Algorithm (MOBMA, 2024), Multi-objective Siege and Conquest Algorithm (MOBCA, 2024), and Multi-objective Ant Colony Algorithm (MOANA, 2025). These algorithms have all been shown to have highly competitive performance. To ensure fairness in the experiment, all algorithms will be run 30 times on the benchmark function, with population size, archive size, and number of iterations set to 100, 100, and 1000, respectively. The parameter settings for all algorithms are consistent with those in the original paper, as detailed in Table 1 below. Table 1 Algorithm Parameter Settings In the experiments, three widely used performance metrics were used to evaluate the multi-objective algorithm: inverse generation distance (IGD), hypervolume (HV), and Δp. These metrics comprehensively evaluate the convergence and distribution of the obtained solutions.

[0109] IGD is used to evaluate an approximate Pareto solution set P and a reference set P. The average distance between them measures the reference set. How close is the average distance from a point in the approximate solution set P to the nearest point? A good algorithm finds a solution set... P It should be as close as possible to the true Pareto front; IGD calculates... P The proximity is quantified by the average difficulty of finding the nearest neighbor for each point. To calculate IGD, approximately 10,000 reference points are uniformly sampled from the true power factors (PF) of each test problem to form a reference point set, which provides higher efficiency than traditional sampling. The formula for calculating IGD is as follows: ; Where d represents each reference point The Euclidean distance to the nearest solution in P; IGD is a powerful metric for evaluating the convergence of multi-objective optimization algorithms; a small IGD value indicates that the approximate solution set P is close to the true Pareto front, and the smaller the IGD value, the better.

[0110] Hypervolume (HV) is one of the most commonly used and important metrics for evaluating the quality of the non-dominated solution set (also known as the Pareto front approximation set) found by a multi-objective optimization algorithm. It comprehensively measures the breadth and uniformity of the solution set's coverage across multiple objectives; given an approximate Pareto solution set found by the algorithm... P and a carefully selected reference point in the target space r Hypervolume HV is the volume of material in the target space composed of… P andr The volume of the commonly defined dominated region; HV is formally expressed as: ; That Let Lebesgue measure, representing P The middle is composed of a reference point r and any point Dominated all points The merged volume; normalize all solutions in the target space, with the reference point set to (1.0, 1.0, ..., 1.0); the larger the HV value, the larger the solution set. P The better the overall quality. Careful selection of reference points is required during calculation. r and ensure that it is P Total domination.

[0111] ∆p is specifically used to quantize an approximate Pareto solution set. P ∆p evaluates the uniformity and breadth of the solution distribution along the True Pareto Front in the target space; unlike HV or IGD which consider both convergence and distribution, ∆p focuses solely on assessing the uniformity and coverage of the solution distribution along the True Pareto Front. For a given p If the value is greater than 0, the Δp metric is defined as follows: ; Δ p The metric consists of two indicators: and Their definitions are as follows: ; ; In the experiment, the value of p was set to 2; Δp achieves a more accurate evaluation by explicitly utilizing the boundary points of the true frontier and the gap distribution of the solution calculated in the parameter space, which captures the true coverage better than the classic Spread index; the smaller the value of Δp, the more uniformly the solution points are distributed along the true frontier, and the better the coverage (especially at the boundaries).

[0112] In addition, all algorithms were tested independently 30 times, and the mean and standard deviation of IGD, HV, and Δp values ​​were collected. To show the significant differences between different algorithms, two nonparametric statistical tests were used: the Wilcoxon rank-sum test, which uses the symbols "+ / - / =" to confirm that the compared algorithms are significantly better / worse / equal to MOCTCM in performance; and the Friedman test, which sets the significance level to 0.05 to obtain the average performance ranking score of multiple algorithms on the same dataset.

[0113] Therefore, in the experimental study, three MOP benchmark sets were selected to examine the performance of MOCTCM, including: 1. The DTLZ (Deterministic Transformation of the Target) test set is a classic test problem set in the field of multi-objective optimization. Proposed in 2002, it consists of seven test problems, namely DTLZ1-DTLZ7. The design goal of DTLZ is to provide test scenarios with different characteristics for multi-objective optimization algorithms. It can flexibly adjust the number of objectives and the dimension of decision variables, and is especially suitable for evaluating the performance of algorithms in high-dimensional multi-objective optimization problems (the number of objectives ≥ 3). Through ingenious mathematical construction, the DTLZ test set provides test scenarios covering various characteristics such as linear, nonlinear, continuous, discontinuous, convex / non-convex for multi-objective algorithms. Its flexibility (adjustable number of objectives, dimension of decision variables, and complexity parameters) makes it a standard tool for high-dimensional multi-objective optimization research.

[0114] 2. The UF test set is derived from 10 test functions, UF1-UF10, from the CEC2009 competition. These test functions simulate the characteristics of real-world engineering problems and address various challenging Pareto fronts, such as linear, convex, concave, and discontinuous fronts. Furthermore, they contain many Pareto local optima, making convergence to the true Pareto front more difficult, thus allowing for testing the algorithm's ability to avoid local optima. The UF test set provides a testing platform for multi-objective optimization algorithms targeting complex Pareto fronts through systematically designed unconstrained problems.

[0115] 3. The IMOP (Irregular Multi-objective Optimization Problems) series was proposed at IEEE TEVC in 2020. It aims to solve the problem that traditional test sets cannot effectively evaluate the performance of algorithms on irregular Pareto fronts. The series consists of 8 test problems, which simulate complex non-ideal geometric structures in engineering optimization through ingenious problem construction. Multi-objective optimization algorithms can be verified to be robust in extremely irregular scenarios through the rigorous testing of the IMOP series. Its high geometric complexity makes it the gold standard for evaluating cutting-edge multi-objective optimization algorithms.

[0116] These three classic multi-objective optimization test problem sets have different characteristics, as shown in Table 2 below: Table 2. Set of test problems for different multi-objective optimization By combining these three series of test sets, the performance limits of the algorithm can be comprehensively evaluated in terms of convergence, diversity preservation, high-dimensional optimization, nonlinear dependence, and adaptation to complex frontiers.

[0117] Optimize 3D-POS by determining decision variables through MOCTCM to make the three objectives in the 3D-POS calculation formula more efficient. Minimize; in MOCTCM, the population size and number of iterations are set to 100 and 50 respectively, and Lb and Ub are set to -10 and 10 respectively. The convergence path of 3D-POS optimization is as follows: Figure 15 As shown, after approximately 30 iterations, the average values ​​of the three objectives of 3D-POS gradually converged, reaching final values ​​of 0.0864, ​​0.3299, and 0.0035, respectively. Finally, all non-dominated solutions for optimizing 3D-POS were obtained, and the solution with the largest HV value was selected. At this point, the optimal variable values ​​are... c j , j∈ [1,2,…,9] are 7.4351, 6.1469, 9.8213, 8.9315, 6.7086, 9.6542, 2.5977, 3.2139, and 4.2468 respectively; the final 3D-POS after optimization variable substitution is given below: .

[0118] Therefore, the Multi-Objective Tribal Competition and Member Cooperation Algorithm (MOCTCM) proposed in this application effectively solves the problems of low efficiency, susceptibility to local optima, and single-objective limitations in parameter optimization of chaotic systems. MOCTCM achieves multi-objective solution management through non-dominated sorting and crowding calculation, enhances exploration capabilities by utilizing Gaussian perturbation and competitive learning, and balances the exploration and utilization of the algorithm by leveraging adaptive parameter adjustment strategies. On benchmark sets such as DTLZ, UF, and IMOP, MOCTCM demonstrates significant advantages over comparative algorithms such as MOMRFO and MOChOA in terms of convergence accuracy, solution diversity, and computational efficiency.

[0119] When MOCTCM is applied to 3D-POS (3D Parametric Optimization) chaotic systems, with Lyapunov exponent, correlation dimension, and 0-1 value as optimization objectives, the optimized system significantly outperforms existing chaotic systems in terms of ergodicity, randomness, and other chaotic performance indicators. The maximum Lyapunov exponent is increased by 270%, the correlation dimension reaches 3.246, and the 0-1 test value reaches 0.999. The experimental results fully demonstrate that MOCTCM has outstanding multi-objective processing capabilities in chaotic parameter optimization, providing a powerful tool for the in-depth application of chaotic systems in the engineering field, and also providing new ideas and methods for parameter optimization of other complex systems.

[0120] Therefore, for reference Figure 5 As shown. In some embodiments of this application, this application also provides a HEVC video selective encryption system based on multi-objective tribal competition and member cooperation optimization, which implements the method described in any of the above embodiments, comprising: The three-dimensional chaos modeling module 201 is used to construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, it constructs a multi-objective optimized chaotic model with preset chaotic performance indicators as optimization objectives and the preset parameters to be optimized as decision variables. The parameter optimization module 202 is used to iteratively optimize and solve the multi-objective optimized chaotic model using a multi-objective tribal competition and member cooperation algorithm, obtain a non-dominated solution set that satisfies the Pareto optimality condition, select the optimal compromise solution from the non-dominated solution set, and substitute it into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system. The syntax parsing module 203 is used to use an optimized three-dimensional parametric chaotic system to iteratively generate a multi-dimensional chaotic sequence based on the initial value of the key and divide the input video frame into coding units during HEVC encoding, thereby parsing syntax elements that include at least the sign bits of the transform coefficients, the sign bits of the motion vector difference, and the intra-frame prediction mode index. The selective encryption module 204 is used to convert the multidimensional chaotic sequence into a pseudo-random key stream with the same length as the syntax element according to a preset selective encryption rule, and to encrypt the syntax element using a preset encryption method to generate an encrypted syntax element. The bitstream reconstruction module 205 is used to rewrite the encrypted syntax elements into HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

[0121] In some application embodiments of this application, the application will be further described in detail with reference to the accompanying drawings and examples.

[0122] This application provides an exemplary embodiment illustrating the specific implementation process of the HEVC video selective encryption method optimized based on multi-objective tribal competition and member cooperation.

[0123] First, a three-dimensional parametric chaotic system (3D-POS) with rich dynamic behavior is designed, whose mathematical expression is a set of nonlinear differential equations with nine adjustable parameters: ; in, For state variables, vector The decision variables consist of nine dimensions; the optimization objective is set as three chaotic performance metrics that need to be maximized simultaneously: the maximum Lyapunov exponent (LE), the correlation dimension (CD), and the 0-1 test value.

[0124] Subsequently, the multi-objective tribal competition and member cooperation algorithm of this application was used to solve the model, with the population size set to 100, file size to 50, and the number of iterations to 100. The algorithm first generates initial tribes that satisfy the boundary constraints in the parameter space. In each iteration: Calculate the LE, CD, and 0-1 test values ​​for each parameter vector.

[0125] The fast non-dominated sort is used to stratify individuals, calculate the crowding distance of individuals within the same stratum, and filter the non-dominated solutions to store them in an external archive.

[0126] In tribal competition, tribes led by non-dominant tribes have a higher probability of winning. Members of the losing tribes are driven by decoupling and migrate to the area where the elite tribes are located, thus achieving global exploration.

[0127] In member collaboration, individuals within a tribe learn from two better members within their own tribe or neighboring areas, thereby enhancing their local development capabilities.

[0128] After 100 iterations, the algorithm outputs a solution set uniformly distributed on the Pareto front. From this front, a fuzzy decision-making method is used to select a compromise optimal solution that considers all three objectives. Substituting this solution into the 3D-POS equation yields the optimized three-dimensional parametric chaotic system. Figure 6 It is evident that the phase space attractor of the optimized three-dimensional parameterized chaotic system exhibits more complex folding and stretching states, indicating a significant enhancement in its chaotic properties.

[0129] Therefore, when it is necessary to securely transmit a 4K video clip acquired in real time, the following operations should be performed: First, the encryption and decryption ends share a 256-bit master key (K). The encryption system uses the secure hash algorithm SHA-256 to apply (K) and the timestamp (T) of the current video frame to generate three 256-bit hash values ​​(K, T, T). ), respectively mapped to the three initial states of the optimized three-dimensional parameterized chaotic system ( ).

[0130] Secondly, ( The optimal parameters are loaded into the chaotic sequence generator, pre-iterate 1000 times to eliminate transient effects, and then iterate N times to generate a three-dimensional floating-point chaotic sequence of length N. Threshold quantization is then used to convert the floating-point sequence into a binary pseudo-random keystream. .

[0131] Next, perform syntax element parsing and encryption: 1. The HEVC encoder divides the input video frames into 64×64 coding tree units (CTUs) and performs prediction and transformation.

[0132] 2. After the transformation encoding is completed, the selective encryption engine extracts the sign bits of all non-zero transform coefficients. For example, if a 4×4 transform block has 5 non-zero coefficients, then 5 sign bits are extracted. .

[0133] 3. For motion compensation, extract the sign bits of the horizontal and vertical components of the MVD of the inter-frame prediction block.

[0134] 4. For intra-frame prediction blocks, if the best prediction mode is one of the 33 angle modes other than Planar and DC, then the codeword prefix representing its mode index is extracted.

[0135] 5. Encryption operations: the engine obtains the key stream... The key, of the same length as the bits to be encrypted, is extracted sequentially and a bitwise XOR operation is performed. For example, the transform coefficient symbol sequence [0,1,1,0,1] is encrypted with the key [1,0,1,1,0] to become [1,1,0,1,1]. The encrypted syntax element bits are put back into the HEVC encoding process and enter the CABAC entropy encoder along with other syntax elements that have been bypassed (such as absolute values ​​of motion vectors, coefficient magnitudes, etc.) to finally form a standard format encrypted video stream.

[0136] Finally, the video decryption process is performed: The decryption end has the same chaotic system and key management mechanism as the encryption end. After receiving and parsing the encryption stream, the decryption system uses the synchronously generated same key stream to perform an XOR operation on the encrypted syntax element bits again, so as to recover the original bits without loss. Then, the decrypted syntax elements are input into the standard HEVC decoder to reconstruct the original video without loss.

[0137] As can be seen from the above application examples, this application utilizes the MOCTCM algorithm to discover the ultimate dynamic performance of a three-dimensional parameterized chaotic system and creatively applies it to the millisecond-level precise masking of HEVC video content-level core bits, achieving a perfect balance between security strength and computational overhead.

[0138] In some embodiments of this application, this application also provides a computer medium storing a computer program, which is executed by a processor to implement the HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization.

[0139] In some embodiments of this application, this application also provides a computer, including the aforementioned computer medium.

[0140] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0141] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A selective encryption method for HEVC video based on multi-objective tribal competition and member cooperation optimization, characterized in that, Includes the following steps: S1: Construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, with the preset chaotic performance index as the optimization objective and the preset parameters to be optimized as the decision variables, construct a multi-objective optimized chaotic model. S2: The multi-objective tribal competition and member cooperation algorithm is used to iteratively optimize and solve the multi-objective optimized chaotic model to obtain a non-dominated solution set that satisfies the Pareto optimality condition. The optimal compromise solution is selected from the non-dominated solution set and substituted into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system. S3: An optimized three-dimensional parametric chaotic system is adopted. A multi-dimensional chaotic sequence is generated iteratively based on the initial value of the key and the input video frame is divided into coding units during the HEVC encoding process. Then, the syntax elements including at least the sign bits of the transform coefficients, the sign bits of the motion vector difference, and the intra-frame prediction mode index are parsed. S4: According to the preset selective encryption rules, the multidimensional chaotic sequence is converted into a pseudo-random key stream with the same length as the syntax element, and the syntax element is encrypted using a preset encryption method to generate an encrypted syntax element. S5: Rewrite the encrypted syntax elements into the HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

2. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 1, characterized in that, In S1, the method for constructing a three-dimensional parameterized chaotic system includes the following steps: S11: Establish a three-dimensional nonlinear dynamic model, treat the state variables as three independent dimensions in the system state space, and construct a three-dimensional chaotic equation system containing nonlinear coupling terms, state feedback terms, and parameter modulation terms. S12: Set preset parameters to be optimized as control parameters and establish the coupling relationship between each control parameter and the state variable respectively. Then, establish parameter boundary constraints on the preset parameters to be optimized and map the value range of each parameter to be optimized to the search space of the multi-objective optimization algorithm. S13: Perform numerical integration calculations based on the preset initial state to obtain the chaotic trajectory sequence formed by the change of state variables over time, and use the chaotic trajectory sequence to construct a time series dataset. Then, analyze the time series dataset from the dimensions of phase space distribution characteristics, trajectory traversal characteristics, attractor complexity characteristics, and randomness characteristics to establish a mapping relationship between parameter space and chaotic performance indicators. S14: Set the preset parameters to be optimized as a decision vector and use the decision vector as the optimized individual in the multi-objective tribal competition and member cooperation algorithm. Then, calculate the corresponding target evaluation index containing Lyapunov index, correlation dimension and 0-1 test value according to each optimized individual. S15: Construct a three-dimensional parameterized chaotic system based on the state variables, control parameters, parameter constraints, and target evaluation indicators.

3. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 2, characterized in that, In S1, the method for constructing a multi-objective optimization chaotic model includes the following steps: S16: The decision variables in the three-dimensional parameterized chaotic system are used as the search objects in the multi-objective optimization algorithm, and numerical solutions are performed on the three-dimensional parameterized chaotic system for each set of decision variables to obtain the dynamic evolution trajectory of the state variables within a preset time interval. S17: Calculate the maximum Lyapunov exponent and correlation dimension of the chaotic system based on the dynamic evolution trajectory, and perform a 0-1 chaos test based on the dynamic evolution trajectory to generate a 0-1 test value; S18: Using the maximum Lyapunov exponent, correlation dimension, and 0-1 test value as the optimization objective function, construct a multi-objective optimization chaotic model consisting of a decision variable space, an optimization objective function space, and parameter constraints.

4. A selective encryption method for HEVC video based on multi-objective tribal competition and member cooperation optimization as described in claim 1 or 3, characterized in that, In S2, the method for iteratively optimizing and solving the multi-objective chaotic optimization model using a multi-objective tribal competition and member cooperation algorithm includes the following steps: S21: Randomly generate multiple candidate parameter vectors according to the value range of the decision variables and use the candidate parameter vectors as initial individuals to construct an initial population. Then, divide the initial population into multiple tribes containing multiple optimized individuals and assign corresponding decision variables to each optimized individual. S22: Calculate the corresponding Lyapunov index, association dimension, and 0-1 test value for each optimized individual, generate multi-objective fitness evaluation results, and perform fast non-dominated sorting based on the multi-objective fitness evaluation results to divide the optimized individuals in the population into multiple non-dominated levels. S23: Calculate the crowding distance for optimized individuals in the same non-dominated level and filter optimized individuals according to the non-dominated level and crowding distance, and store the non-dominated solutions that meet the preset conditions to the external file; S24: Use the external archive to save and update the Pareto optimal solution obtained in the current iteration, wherein when the capacity of the external archive reaches a preset upper limit, uniformly distributed non-dominated solutions are preferentially retained according to the crowding distance.

5. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 4, characterized in that, In S3, the method for iteratively generating a multidimensional chaotic sequence using an optimized three-dimensional parameterized chaotic system based on an initial value generated by a key includes the following steps: S31: Generate initial state parameters of the three-dimensional parameterized chaotic system based on the user-input key, initialization vector, and associated information of the video to be encrypted, and start the three-dimensional parameterized chaotic system using the initial state parameters to perform a preset number of pre-iteration operations; S32: After completing the pre-iteration, continue to perform iterative calculations to obtain the chaotic trajectory sequence corresponding to the state variable and perform normalization, quantization and bit mapping on the chaotic trajectory sequence to generate a pseudo-random sequence that meets the encryption requirements. S33: Based on the data length of the syntax elements in the HEVC encoding process, the pseudo-random sequence is truncated, recombined, or expanded to generate a key stream of corresponding length. Based on the key stream, a multidimensional chaotic sequence is generated using an optimized three-dimensional parameterized chaotic system.

6. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 5, characterized in that, In S3, the method for dividing the input video frame into coding units during HEVC encoding and then parsing syntax elements including at least transform coefficient sign bits, motion vector difference sign bits, and intra-frame prediction mode index includes the following steps: S34: Perform HEVC encoding on the input video frame and acquire the corresponding video bitstream data during the encoding process; S35: Perform syntax parsing on the video stream data, extract syntax elements from the HEVC stream, and determine the target syntax elements for implementing selective encryption.

7. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 6, characterized in that, in, The target syntax element includes a non-zero transform coefficient sign bit, a motion vector difference sign bit, and an intra-frame prediction mode index; wherein, the non-zero transform coefficient sign bit is subjected to sign bit encryption, the motion vector difference sign bit is subjected to direction information encryption, and the intra-frame prediction mode index is subjected to mode index encryption.

8. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 6, characterized in that, In S4, the method of encrypting the syntax element using a preset encryption method to generate the encrypted syntax element includes the following steps: S41: Calculate the bit length information corresponding to the target syntax element and determine the corresponding key stream length based on the bit length information; S42: Extract a data sequence corresponding to the length of the target syntax element from the multidimensional chaotic sequence and convert the data sequence into a binary pseudo-random key stream; S43: Establish a corresponding keystream mapping relationship based on the type of the target syntax element in the HEVC codestream, so that different types of target syntax elements correspond to keystream data at different positions; S44: The pseudo-random key stream and the target syntax element are operated bit by bit using the XOR operation method, while keeping the content of the unselected syntax elements unchanged, and only the target syntax element is encrypted to generate an encrypted syntax element.

9. The HEVC video selective encryption method based on multi-objective tribal competition and member cooperation optimization according to claim 8, characterized in that, In S5, the method of rewriting the encrypted syntax element into HEVC encoding and performing CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax element includes the following steps: S51: Replace the original syntax element at the corresponding position during HEVC encoding with the encrypted syntax element; S52: The replaced syntax elements and the unencrypted syntax elements are subjected to CABAC entropy encoding to complete context modeling and binary arithmetic encoding. S53: Perform NAL unit encapsulation and bitstream reassembly on the entropy-encoded syntax elements to generate an encrypted video bitstream that conforms to the HEVC standard syntax structure.

10. A HEVC video selective encryption system based on multi-objective tribal competition and member cooperation optimization, implementing the method of any one of claims 1-9, characterized in that, include: The three-dimensional chaos modeling module is used to construct a three-dimensional parameterized chaotic system and define a chaotic mapping equation containing preset parameters to be optimized. Then, with preset chaotic performance indicators as optimization objectives and the preset parameters to be optimized as decision variables, a multi-objective optimization chaotic model is constructed. The parameter optimization module is used to iteratively optimize and solve the multi-objective optimized chaotic model using a multi-objective tribal competition and member cooperation algorithm, obtain a non-dominated solution set that satisfies the Pareto optimality condition, select the optimal compromise solution from the non-dominated solution set, and substitute it into the three-dimensional parameterized chaotic system to obtain the optimized three-dimensional parameterized chaotic system. The syntax parsing module is used to generate a multidimensional chaotic sequence iteratively based on the initial value of the key using an optimized three-dimensional parametric chaotic system. During the HEVC encoding process, the input video frame is divided into coding units, and then the syntax elements, including at least the sign bits of the transform coefficients, the sign bits of the motion vector difference, and the intra-frame prediction mode index, are parsed. A selective encryption module is used to convert the multidimensional chaotic sequence into a pseudo-random key stream with the same length as the syntax element according to a preset selective encryption rule, and to encrypt the syntax element using a preset encryption method to generate an encrypted syntax element. The bitstream reconstruction module is used to rewrite the encrypted syntax elements into HEVC encoding and perform CABAC entropy encoding and bitstream encapsulation processing together with the unencrypted syntax elements to generate an encrypted HEVC video bitstream.

Images