Heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G network

Abstract

The invention discloses a heterogeneous encrypted multi-mode multimedia message real-time fragmentation transmission method for a 5G network, and relates to the technical field of multimedia transmission optimization. According to the heterogeneous encryption multi-mode multimedia message real-time fragmentation transmission method for the 5G network, the fragmentation granularity and priority are optimized through a reinforcement learning algorithm by obtaining the sensitive label and the value density of multimedia data and combining the 5G slice resource state. And dynamically selecting an encryption algorithm cluster according to a fragmentation rule, and realizing cross-slice key distribution and updating by using an alliance chain. On the basis of space-time relevance of multi-modal data, a fountain code driven redundant coding strategy is constructed, and a fragment transmission path is optimized in combination with a game theory model. And finally, analyzing a transmission flow characteristic and a key distribution log, and dynamically adjusting an encryption strength threshold value and a fragment retransmission strategy. According to the method, the security, the reliability and the transmission efficiency of the multimedia data in the 5G network are improved, and the method is suitable for a multimedia communication scene with low time delay and high reliability.

Description

Technical Field

[0001] The present invention relates to the technical field of multimedia transmission optimization, and specifically to a real-time sharding transmission method for heterogeneous encrypted multimodal multimedia messages for 5G networks. Background Technique

[0002] With the commercialization of 5G networks, the demand for high-bandwidth and low-latency transmission of multimodal multimedia applications has increased sharply. At the same time, the threat of data security has become increasingly complex, and the differentiated requirements for encryption intensity and transmission reliability in different business scenarios have been significantly improved. Traditional single encryption and static sharding mechanisms are difficult to adapt to the heterogeneous resource characteristics and dynamic service requirements of 5G network slices, and there is an urgent need to build an intelligent sharding transmission system for multi-slice collaboration.

[0003] Existing methods mostly adopt fixed sharding granularity and static encryption strategies, and cannot dynamically adjust the sharding priority and encryption intensity according to the network slice resource status, resulting in high-value data shards may cause transmission congestion or security vulnerabilities due to resource competition; traditional centralized key distribution mechanisms are difficult to support multi-slice collaboration scenarios, and there are problems such as high key synchronization delay and low credibility of cross-domain distribution paths; existing redundancy strategies lack intelligent perception of the spatio-temporal correlation of shards, and the fixed redundancy factor leads to bandwidth waste or insufficient error correction ability, and cannot adapt to the dynamic network packet loss rate; traditional routing algorithms only consider delay or bandwidth metrics, ignoring the multi-objective game relationship of security level, energy consumption cost and slice resource competition, and it is difficult to achieve global optimal transmission. Summary of the Invention

[0004] Aiming at the deficiencies of the prior art, the present invention provides a real-time sharding transmission method for heterogeneous encrypted multimodal multimedia messages for 5G networks, which solves the problems in the above background technique.

[0005] To achieve the above objectives, the present invention is realized through the following technical solutions: A heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for a 5G network, comprising the following steps: S1. Obtain the sensitivity label and value density of multimedia data, and real-time monitor the resource status of 5G network slices. Dynamically generate a fragmentation granularity and priority mapping rule through a reinforcement learning algorithm, extract spatio-temporal key features by combining a convolutional network, and output a fragmentation sequence adapted to the 5G slice resources and the corresponding encryption intensity label; S2. Receive the fragmentation sequence and the encryption intensity label, dynamically select an encryption algorithm cluster according to the fragmentation granularity and priority mapping rule, realize cross-slice key distribution and update through a consortium blockchain, and generate encrypted fragmentation data blocks and key distribution logs; S3. According to the encrypted fragmentation data blocks and the spatio-temporal correlation of multi-modal data, construct a fountain code-driven redundant coding strategy, optimize the fragmentation transmission path by combining a game theory model, and output a mapping table of associated fragmentation groups and transmission paths; S4. According to the mapping table of associated fragmentation groups and transmission paths, analyze the transmission traffic characteristics and key distribution logs, and dynamically adjust the encryption intensity threshold and fragmentation retransmission strategy.

[0006] Further, the specific process of obtaining the sensitivity label and value density of multimedia data and real-time monitoring the resource status of 5G network slices is as follows: According to the distribution density of sensitive words in the text data annotated by natural language processing, extract the key frame motion trajectory features of video data, analyze the frequency domain burst frequency of audio data, and combine the mutation amplitude of the sensor waveform to dynamically calculate the value density level of multimedia data; Real-time collect the remaining bandwidth of the uRLLC slice, the delay jitter of the eMBB slice, and the packet loss rate of the mMTC slice through the 5G core network interface, and generate a 5G network resource status vector.

[0007] Further, the specific process of dynamically generating a fragmentation granularity and priority mapping rule through a reinforcement learning algorithm is as follows: Construct a state space with the value density feature vector of multimedia data and the 5G network slice resource status vector as inputs, define the fragmentation granularity level and transmission priority weight as the action space, design a multi-objective reward function to comprehensively consider the encryption intensity matching degree, delay deviation, and terminal energy consumption, and generate a dynamic mapping matrix of fragmentation granularity and priority through policy iteration.

[0008] Further, the specific process of extracting spatio-temporal key features by combining a convolutional network and outputting a fragmentation sequence adapted to the 5G slice resources and the corresponding encryption intensity label is as follows: Extract spatio-temporal features of the video stream through a lightweight convolutional network, identify the motion correlation between key frames and the burst silent segments in the audio stream, fuse the frequency domain mutation points of the sensor waveform, and generate a spatio-temporal key feature vector; Dynamically divide the fragmentation boundary and mark the encryption intensity label according to the value density feature vector of multimedia data and the 5G slice resources, and output a fragmentation sequence adapted to the characteristics of the uRLLC, eMBB, and mMTC slices.

[0009] Furthermore, the specific process of dynamically selecting the encryption algorithm cluster according to the sharding granularity and priority mapping rules is as follows: Construct a heterogeneous encryption algorithm pool, generate a multi-dimensional matching index based on the sharding priority weight and encryption strength label, and statistically analyze the network jitter coefficient and computational load fluctuation characteristics of historical shards through a sliding time window; According to the remaining bandwidth of the current slice resource and the terminal hardware acceleration ability, dynamically combine symmetric encryption algorithms and lightweight asymmetric encryption algorithms to form an algorithm cluster with linked sharding granularity and encryption strength, including: Fine-grained sharding: Bind the attribute-based homomorphic encryption algorithm, generate dynamic keys with the spatio-temporal characteristics of multi-modal data as attribute parameters, and the spatio-temporal characteristics of multi-modal data include video key frame hash values and sensor waveform peaks; Medium-grained sharding: Through the hierarchical hybrid encryption algorithm, cross-encrypt data blocks, and the keys are dynamically distributed through the edge node consensus protocol; Coarse-grained sharding: Through dynamic mask perturbation encryption, the keys are sent on demand through the 5G control plane signaling to adapt to the low-power consumption transmission requirements; According to the spatio-temporal key feature vector of the sharding sequence, trigger the key length adaptive adjustment mechanism and encryption mode switching event to realize the real-time coupling of the algorithm cluster parameters and the network slice QoS indicators.

[0010] Furthermore, the specific process of realizing cross-slice key distribution and update through the consortium chain and generating encrypted sharding data blocks and key distribution logs is as follows: Split the sharding key into multiple sub-key components according to the slice domain, and define the threshold signature rule and consensus verification condition for cross-domain key distribution; Through the dynamic authorization channel between consortium chain nodes, push the sub-key components to the edge computing nodes corresponding to the uRLLC, eMBB, and mMTC slices, and record the key version number and distribution path topology in the immutable block log; Combine the encryption timestamp and life cycle label of the sharding data block to trigger periodic key update events, and use the cross-chain protocol to synchronously update the key copies of the heterogeneous blockchain network.

[0011] Furthermore, the specific process of constructing a fountain code-driven redundant coding strategy is as follows: Dynamically generate a redundancy factor according to the encryption strength label and spatio-temporal correlation of the sharding data block, and adjust the coding degree distribution function of the fountain code according to the spatio-temporal feature distribution density of multi-modal data; Arrange the encrypted sharding data blocks and redundant check blocks in a two-dimensional matrix in the frequency domain and time domain, dynamically detect the association strength of the sharding group through a sliding window and inject differential redundancy; Combine the real-time packet loss rate and transmission reliability index of the slice to establish a dynamic adaptation relationship between the redundant coding parameters and the QoS requirements, and generate a multi-slice collaborative redundancy strategy.

[0012] Furthermore, the specific process of optimizing the shard transmission path in combination with the game theory model and outputting the mapping table of associated shard groups and transmission paths is as follows: Define the available bandwidth of transmission nodes, the slice resource competition coefficient, and the shard security level as the strategy space, and design a composite utility function that includes delay cost, security risk loss, and energy consumption cost; Solve the Nash equilibrium point through the evolutionary game algorithm to generate an initial matching scheme for shard groups and transmission paths, and dynamically correct the path selection weight according to the spatio-temporal key feature vector of the shard sequence; Combine the blockchain key distribution topology to construct a credibility evaluation mechanism and output the dynamic mapping table of shard groups and slice resources.

[0013] Furthermore, according to the mapping table of associated shard groups and transmission paths, the specific process of analyzing the transmission traffic characteristics and key distribution logs and dynamically adjusting the encryption intensity threshold and shard retransmission strategy is as follows: Real-time monitor the packet loss rate, key distribution delay, and blockchain consensus verification success rate of the shard transmission path, and construct a joint evaluation matrix of network status and security risk; Establish a dynamic adjustment model for the encryption intensity threshold according to the reinforcement learning algorithm, trigger the shard encryption mode switching event according to the risk level, and generate the shard retransmission priority queue at the same time.

[0014] The specific process of the further step is as follows: Construct a multi-objective optimization space that includes the encryption intensity gradient, end-to-end delay constraint, and energy consumption threshold, define the dynamic approximation condition of the Pareto front, solve the optimal solution set of security and efficiency, and dynamically adjust the weight coefficient of the objective function in combination with the fluctuation characteristics of network slice resources; Generate a hierarchical response instruction set according to the shard value density level, including encryption algorithm cluster reorganization instructions, shard path switching instructions, and redundant coding reconstruction instructions.

[0015] The present invention has the following beneficial effects:

[0016] (1) The heterogeneous encryption multi-modal multimedia message real-time shard transmission method for 5G networks obtains the sensitivity label and value density of multimedia data, and real-time monitors the resource status of 5G network slices. Combining with the reinforcement learning algorithm, it dynamically generates the shard granularity and priority mapping rules, and can flexibly adjust the shard granularity and encryption intensity of data according to the changes in network resources. This adjustment based on the dynamic algorithm can effectively ensure that high-sensitivity and high-value data are preferentially transmitted when the network load is high, reducing the data packet loss rate and improving the transmission efficiency. At the same time, combining the spatio-temporal key features extracted by the convolutional network makes the generation of shard sequences and encryption intensity labels more accurate, thus improving the data security and transmission reliability.

[0017] (2)The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks can ensure the reliability and stability of multi-modal data transmission in complex network environments through the redundant coding strategy based on fountain codes and the fragmentation transmission path design optimized by game theory. By combining a multi-objective optimization algorithm to dynamically adjust the encryption intensity threshold and the fragmentation retransmission strategy, the balance between security and transmission efficiency during the transmission process is optimized. When network fluctuations or congestion occur during the transmission process, the system can also dynamically adjust the strategy to ensure the rapid recovery and secure transmission of data, ultimately ensuring an efficient and stable data transmission experience.

[0018] Of course, it is not necessary for any product implementing the present invention to simultaneously achieve all the above-mentioned advantages. Brief Description of the Drawings

[0019] Figure 1 It is a flowchart of the heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to the present invention.

[0020] Figure 2 It is a flowchart of the multimedia message transmission steps according to the present invention.

[0021] Figure 3 It is a flowchart of step S1 according to the present invention Detailed Embodiments

[0022] The embodiment of the present application solves the problems of low resource utilization efficiency, insufficient credibility of cross-slice key distribution, mismatch between redundant coding and network status, and imbalance between security and efficiency caused by the static coupling of fragmentation strategies and encryption mechanisms in the prior art through the heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks. Specifically, it includes: breaking through the traditional fixed fragmentation mode, dynamically generating fragmentation granularity and priority mapping rules through reinforcement learning, and realizing the real-time adaptation of network slice resource status and multi-modal data value density; cracking the cross-domain key management problem, constructing a key distribution mechanism combining threshold signature and dynamic authorization channel based on the consortium chain to ensure the credibility and timeliness of key distribution in multi-slice scenarios; overcoming the rigid defect of redundant coding, proposing a dynamic redundant strategy of fountain codes driven by spatio-temporal correlation, and realizing the intelligent matching of the coding degree distribution function with the network packet loss rate and fragmentation priority; eliminating the limitation of single-dimensional path optimization, constructing a multi-objective game model that integrates security level, delay cost, and slice resource competition coefficient, and generating an optimal path mapping that takes into account both security and transmission efficiency.

[0023] The overall idea of the solution in the embodiment of the present application is as follows:

[0024] Obtain the sensitivity label and value density of multimedia data, and real-time monitor the resource status of 5G network slices. Dynamically generate the mapping rules of slice granularity and priority through the reinforcement learning algorithm, extract spatio-temporal key features by combining convolutional networks, and output the slice sequence adapted to 5G slice resources and the corresponding encryption strength label.

[0025] Receive the slice sequence and encryption strength label, dynamically select the encryption algorithm cluster according to the mapping rules of slice granularity and priority, realize cross-slice key distribution and update through the consortium blockchain, and generate the encrypted slice data block and key distribution log.

[0026] According to the encrypted slice data block and the spatio-temporal correlation of multimodal data, construct a fountain code-driven redundant coding strategy, optimize the slice transmission path by combining the game theory model, and output the mapping table of associated slice groups and transmission paths.

[0027] According to the mapping table of associated slice groups and transmission paths, analyze the transmission traffic characteristics and key distribution log, and dynamically adjust the encryption strength threshold and slice retransmission strategy.

[0028] Please refer to Figure 1 and Figure 2 For this, an embodiment of the present invention provides a technical solution: a real-time slice transmission method for heterogeneous encrypted multimodal multimedia messages for 5G networks, including the following steps: S1. Obtain the sensitivity label and value density of multimedia data, and real-time monitor the resource status of 5G network slices. Dynamically generate the mapping rules of slice granularity and priority through the reinforcement learning algorithm, extract spatio-temporal key features by combining convolutional networks, and output the slice sequence adapted to 5G slice resources and the corresponding encryption strength label; S2. Receive the slice sequence and encryption strength label, dynamically select the encryption algorithm cluster according to the mapping rules of slice granularity and priority, realize cross-slice key distribution and update through the consortium blockchain, and generate the encrypted slice data block and key distribution log; S3. According to the encrypted slice data block and the spatio-temporal correlation of multimodal data, construct a fountain code-driven redundant coding strategy, optimize the slice transmission path by combining the game theory model, and output the mapping table of associated slice groups and transmission paths; S4. According to the mapping table of associated slice groups and transmission paths, analyze the transmission traffic characteristics and key distribution log, and dynamically adjust the encryption strength threshold and slice retransmission strategy.

[0029] In this implementation scheme, step S1: The system first obtains the sensitivity label and value density of the multimedia data. This is to judge the security requirements of the data and its transmission priority. Specifically, the sensitivity label is used to characterize the importance of the data (such as privacy data, financial data, etc.), and the value density represents the high or low value of the data, so that priority sorting can be carried out during the transmission process. The system will monitor the resource status of the 5G network slice in real time to evaluate the load condition of the current network resources. This information will be input into the reinforcement learning algorithm to dynamically generate the mapping rule between the slice granularity and the priority, and combined with the spatio-temporal key features of the data extracted by the convolutional neural network, further generate the slice sequence adapted to the current 5G network resource status and the corresponding encryption strength label. In this way, the system can make an optimized data slicing and encryption strategy according to different data characteristics and network resource conditions. Step S2: Select the encryption algorithm and perform cross-slice key distribution. In this step, after the system receives the slice sequence and its corresponding encryption strength label, according to the mapping rule between the slice granularity and the priority generated in the first step, it dynamically selects a suitable encryption algorithm cluster. Here, the "encryption algorithm cluster" refers to a set of multiple encryption algorithms, and a suitable algorithm is selected according to different data requirements. After selecting the encryption algorithm, the system realizes the secure distribution and update of the key between different 5G network slices through the consortium chain technology (i.e., a form of distributed ledger technology). In this way, it can be ensured that the cross-slice key management and update are synchronized and secure, ensuring that the data transmission within each slice can be properly encrypted and protected. Finally, the system generates the encrypted slice data block and records the key distribution log for subsequent traceability and management. Step S3: Construct a redundant coding strategy and optimize the transmission path. This step constructs a fountain code-driven redundant coding strategy based on the encrypted slice data block and the spatio-temporal correlation of the multimodal data. The fountain code is an efficient coding method that can ensure that even if some data is lost during the transmission process, the original information can still be recovered from the remaining data. During the transmission process, the slice transmission path is optimized according to the game theory model, that is, by simulating the game between multiple transmission nodes, the optimal transmission path is determined to maximize the success rate of data transmission and minimize the delay. Finally, the system outputs the mapping table of the associated slice group and the transmission path, indicating the paths of each slice data transmitted in the network and the relationship between them. Step S4: The system analyzes the transmission traffic characteristics of the network and the key distribution log according to the mapping table of the associated slice group and the transmission path obtained in step S3. By monitoring the transmission situation in real time, the encryption strength threshold and the retransmission strategy of the slice are dynamically adjusted. For example, when a certain network path is congested or the packet loss rate is high, the system will automatically reduce the encryption strength or select a more reliable transmission path to ensure that the data can be transmitted as smoothly as possible. During the adjustment process, combined with the multi-objective optimization algorithm, the system weighs between security and transmission efficiency and generates risk response instructions to cope with different network conditions.In this way, the system can effectively cope with the unstable factors in the 5G network and ensure the security and efficiency of the data transmission process.

[0030] Please refer to Figure 3 , specifically, the specific process of obtaining the sensitivity label and value density of multimedia data and real-time monitoring the resource status of 5G network slices is as follows: According to the distribution density of sensitive words in the text data annotated by natural language processing, extract the key frame motion trajectory features of video data, analyze the frequency domain burst frequency of audio data, and combine the mutation amplitude of the sensor waveform to dynamically calculate the value density level of multimedia data; Real-time collect the remaining bandwidth of the uRLLC slice, the delay jitter of the eMBB slice, and the packet loss rate of the mMTC slice through the 5G core network interface to generate a 5G network resource status vector.

[0031] In this implementation scheme, in the transmission of multimedia data, the distribution density of sensitive words in the text data annotated by natural language processing: For text data (such as messages, documents, etc.), identify sensitive words (such as ID card numbers) in it through natural language processing (NLP) technology. The distribution density of sensitive words reflects the privacy level of the text data. If the text contains a large number of sensitive words, the sensitivity label of the text will be rated as high, otherwise it will be low. Extraction of key frame motion trajectory features of video data: Process the video data, extract the key frames in the video, and analyze the motion trajectories of these key frames. Areas in the video with fast movement or change (such as rapidly changing scenes or actions) may contain higher sensitive information, so their sensitivity labels are higher. The motion trajectory features reflect the motion patterns of objects in the video and can help identify areas that need special protection. Analysis of the frequency domain burst frequency of audio data: Perform frequency domain analysis on the audio data to identify the burst frequency in it (such as prominent sound peaks in the audio). If the audio contains sensitive content (such as specific voice signals or noises), its burst frequency will change, thus affecting the generation of the sensitivity label. Analysis of the mutation amplitude of the sensor waveform: Perform waveform analysis on the data collected by sensors (such as temperature sensors, motion sensors, etc.) to identify the mutation amplitude. The mutation of sensor data often means the occurrence of important events (such as security-related alarms or abnormal situations), so the sensitivity label of this data is higher. Dynamically calculate the value density level of multimedia data: By comprehensively analyzing various data features (such as text sensitive words, video motion features, audio frequency domain features, sensor data, etc.), dynamically calculate the "value density" of each data. The value density represents the value of the data, and the higher the value density, the more important the data (such as key business data, important decision support data, etc.). Dynamically calculate the value density level: ; Parameter description: : Distribution density of sensitive words, : Motion trajectory intensity of adjacent key frames, : Burst frequency in the frequency domain, : Waveform mutation amplitude, , , , : Weight coefficient (needs to satisfy ); V: Value density level. This dynamic calculation can update the value level of data in real time according to the changes in multimedia data. Real-time monitoring of the resource status of 5G network slices, which is a key mechanism in 5G technology and can slice network resources according to different requirements (such as low latency, high bandwidth, etc.). When real-time monitoring the network resource status, the system collects and analyzes the status information of the 5G network in the following ways: Remaining bandwidth of the uRLLC slice: The uRLLC (Ultra-Reliable Low-Latency Communication) slice provides ultra-high reliability and low-latency communication services. The system real-time monitors the remaining bandwidth of the uRLLC slice, that is, the available bandwidth capacity in this slice. Bandwidth is a key factor affecting transmission efficiency. The larger the remaining bandwidth, the stronger the transmission ability of the network, and vice versa, it means a higher network load. Delay jitter of the eMBB slice: The eMBB (Enhanced Mobile Broadband) slice supports high-speed and large-capacity broadband communication and is suitable for scenarios such as high-definition video and large data transmission. Delay jitter refers to the change in delay during data transmission. Real-time monitoring of delay jitter can help judge the stability of the network. Excessive delay jitter will lead to a decline in transmission quality and affect the experience of applications such as video streams and online games. Packet loss rate of the mMTC slice: The mMTC (Massive Machine-Type Communications) slice is suitable for the connection of a large number of Internet of Things devices and has the characteristics of high device density and low energy consumption. The packet loss rate refers to the proportion of lost data packets during data transmission. Monitoring the packet loss rate helps evaluate the reliability of the network. A higher packet loss rate means a poorer network quality and may affect the normal communication of IoT devices. Based on the above monitoring data, the system generates a 5G network resource status vector, which includes key network performance indicators such as the remaining bandwidth of the uRLLC slice, the delay jitter of the eMBB slice, and the packet loss rate of the mMTC slice. Through these data, the system can evaluate the network status in real time and provide a basis for subsequent sharding transmission decisions.

[0032] Specifically, the specific process of dynamically generating the mapping rules of slice granularity and priority through the reinforcement learning algorithm is as follows: Construct a state space with the feature vector of multimedia data value density and the state vector of 5G network slice resources as inputs, define the slice granularity level and transmission priority weight as the action space, design a multi-objective reward function to comprehensively consider the encryption strength matching degree, delay deviation, and terminal energy consumption, and generate a dynamic mapping matrix of slice granularity and priority through policy iteration.

[0033] In this implementation plan, the detailed explanation of the dynamic generation of slice rules by the reinforcement learning algorithm, the input vectors for constructing the state space: the feature vector of multimedia data value density: ; : The distribution density of text sensitive words; M: The motion trajectory intensity of video key frames; B: The burst frequency in the audio-visual domain; A: The mutation amplitude of the sensor waveform (physical quantity unit). The state vector of 5G network slice resources: : The remaining bandwidth (Mbps) of the uRLCC slice; : The delay jitter of the eMBB slice; : The packet loss rate (percentage) of the mMTC slice. State space: · Operator ⊕: Represents vector concatenation to form a 7-dimensional state vector. Definition of the action space, slice granularity level (discrete action): Fine granularity: slice duration ; Medium granularity: ; Coarse granularity: . Transmission priority weight (continuous action): High priority: weight ; Medium priority: weight ; Low priority: weight . Action space: Z = { {(T, w)}}, ; . Design of the multi-objective reward function, sub-objective quantization: Encryption strength matching degree: ; : The actual encryption strength (such as 1.0 for AES-256 and 0.8 for ECC); : The encryption strength required according to the sensitivity label (high sensitivity = 1.0, medium sensitivity = 0.8, low sensitivity = 0.5). Delay deviation: ; : The actual transmission delay (ms); : The target delay (10 ms for the uRLLC slice, 50 ms for the eMBB, and 100 ms for the mMTC). Terminal energy consumption: ; : The encryption and transmission energy consumption of the terminal; : The maximum allowable energy consumption of the terminal. Comprehensive reward function (weighted harmonic mean): ; Weight coefficient: : Represent the priorities of security, real-time performance, and energy efficiency respectively (default ). Policy iteration generates a dynamic mapping matrix, and the policy function (Deep Deterministic Policy Gradient, DDPG): Actor network: Input state S, output action Z; Critic network: Input S and Z, output Q-value estimation. Policy update rule: ; θ: Actor network parameters; μ(S): Policy function; Q(S,Z): Q-value estimation of the Critic network. Dynamic mapping matrix: ; Matrix elements: Each row represents an action combination (fragmentation granularity + priority) and its corresponding expected reward R.

[0034] Specifically, the specific process of combining the convolutional network to extract spatio-temporal key features and output the fragmentation sequence adapted to 5G slice resources and the corresponding encryption intensity label is as follows: The spatio-temporal features of the video stream are extracted through a lightweight convolutional network, the motion correlation between key frames is identified, and the sudden silent segments in the audio stream are fused. The frequency-domain mutation points of the sensor waveform are generated to form a spatio-temporal key feature vector; according to the multimedia data value density feature vector and the 5G slice resource status, the fragmentation boundary is dynamically divided and the encryption intensity label is marked, and a fragmentation sequence adapted to the uRLLC, eMBB, and mMTC slice characteristics is output.

[0035] In this implementation scheme, the process of spatio-temporal feature extraction of the video stream: A lightweight convolutional network (such as MobileNetV3) is used to extract spatio-temporal features of the video stream to capture the motion correlation between key frames. ; Parameter description: : Input video frame sequence (dimension: time × space × channel); : Convolution kernel weight (initialized through a pre-trained model); : Output spatio-temporal feature vector. The process of detecting sudden silent segments in the audio stream: Detect sudden silent segments in the audio stream through short-time energy analysis and extract sudden frequency features. Formula: ; ; Parameter description: : The th sampling point of the audio signal; : Silent segment energy threshold (dynamically adjusted according to environmental noise); : Proportion of the silent segment (unitless ratio). The process of analyzing the frequency-domain mutation points of the sensor waveform: Perform a fast Fourier transform (FFT) on the sensor waveform (such as vibration, temperature) to detect the mutation of the frequency-domain energy. Formula: ; ; Parameter description: : Sensor waveform time series; : Frequency-domain energy distribution vector; : Frequency domain mutation amplitude (physical quantity unit). Multimodal spatio-temporal key feature fusion process: Weightedly fuse video, audio, and sensor features to generate a unified spatio-temporal key feature vector. Formula: ; Parameter description: : Weight coefficient, dynamically adjusted according to data value density, satisfying ; Fused spatio-temporal key feature vector. Dynamic shard boundary division and encryption label marking process: Combine the value density feature vector with the 5G slice resource status to dynamically divide the shard boundary and assign encryption strength labels. Shard duration calculation: ; Parameter description: : Value density level (0 - 1); : 5G slice resource status vector; : Normalization coefficient (ensuring within a reasonable range); : Shard duration.

[0036] Specifically, the specific process of dynamically selecting the encryption algorithm cluster according to the shard granularity and priority mapping rules is as follows: Construct a heterogeneous encryption algorithm pool, generate a multi-dimensional matching index according to the shard priority weight and encryption strength label, and statistically analyze the network jitter coefficient and computing load fluctuation characteristics of historical shards through a sliding time window; According to the remaining bandwidth of the current slice resource and the terminal hardware acceleration ability, dynamically combine symmetric encryption algorithms and lightweight asymmetric encryption algorithms to form an algorithm cluster with shard granularity and encryption strength linkage, including: Fine-grained sharding: Bind the attribute-based homomorphic encryption algorithm, generate dynamic keys with the spatio-temporal features of multimodal data as attribute parameters, and the spatio-temporal features of multimodal data include video key frame hash values and sensor waveform peaks; Medium-grained sharding: Cross-encrypt data blocks through a hierarchical hybrid encryption algorithm, and the key is dynamically distributed through the edge node consensus protocol; Coarse-grained sharding: Through dynamic mask perturbation encryption, the key is sent on demand through the 5G control plane signaling to adapt to the low-power consumption transmission requirements; According to the spatio-temporal key feature vector of the shard sequence, trigger the key length adaptive adjustment mechanism and encryption mode switching event to realize the real-time coupling of the algorithm cluster parameters and the network slice QoS index.

[0037] In this implementation scheme, a heterogeneous encryption algorithm pool is constructed: a variety of encryption algorithms, including symmetric encryption and lightweight asymmetric encryption, are aggregated to form an algorithm pool. Each algorithm is classified according to different performance requirements and resource constraints to adapt to different granularity and encryption strength requirements. Generate multi-dimensional matching indexes: Generate multi-dimensional matching indexes through shard priority weights and encryption strength labels. These indexes determine the appropriate encryption algorithm based on the spatiotemporal characteristics of the shards, encryption strength labels, and network status characteristics (such as bandwidth, load, etc.). Historical network jitter and computational load statistics: The network jitter coefficient and computational load fluctuation characteristics of historical shards are counted through sliding time windows. This process helps optimize the current encryption algorithm selection to adapt to changes in network conditions and load fluctuations. Dynamically select encryption algorithm clusters: Dynamically combine and select appropriate encryption algorithms based on the remaining bandwidth of the current slice and the terminal hardware acceleration capability. For shards of different granularities (fine-grained, medium-grained, and coarse-grained), select the corresponding encryption algorithm: Fine-grained shards: Bind attribute-based homomorphic encryption algorithms to generate dynamic keys through multimodal data spatiotemporal characteristics. Features such as video keyframe hash values ​​and sensor waveform peaks Medium-granular sharding: Using a layered hybrid encryption algorithm, data blocks are cross-encrypted, and keys are dynamically distributed through edge node consensus protocols. Coarse-grained sharding: Encrypted through dynamic mask perturbation, keys are issued on demand through 5G control plane signaling to meet low energy consumption requirements. Key length adaptive adjustment mechanism and encryption mode switching: Trigger key length adaptive adjustment and encryption mode switching based on the spatiotemporal key feature vectors of the sharding sequence. This mechanism adjusts the strength and efficiency of encryption according to real-time network conditions and data requirements to ensure that the QoS (quality of service) indicators of the network slice are met.

[0038] Specifically, the specific process of cross-slice key distribution and update and generating encrypted shard data blocks and key distribution logs through the alliance chain is as follows: split the shard key into multiple sub-key components according to the slice domain, and define the threshold signature rules and consensus verification conditions for cross-domain key distribution; through the dynamic authorization channel between alliance chain nodes, push the sub-key components to the edge computing nodes corresponding to the uRLLC, eMBB, and mMTC slices, and record the key version number and distribution path topology to the tamper-proof block log; combine the encryption timestamp and life cycle label of the shard data block to trigger periodic key update events, and use the cross-chain protocol to synchronously update the key copies of the heterogeneous blockchain network.

[0039] In this implementation plan, the sharded key is split into multiple sub-key components according to the slice domain: the key is split into multiple sub-key components according to different slice domains (such as uRLLC, eMBB, mMTC). This method ensures that different slices have independent and dedicated key parts according to their specific quality of service requirements (such as latency, bandwidth, number of connections, etc.), thereby improving security and flexibility. Define the threshold signature rule and consensus verification condition for cross-domain key distribution: A threshold signature rule is designed, in which the verification of multiple nodes is required to complete the key distribution and update. This strategy ensures that the key distribution process reaches an agreement among multiple consortium chain nodes, thereby improving the security of the key update process. The consensus verification condition ensures that in the cross-domain key distribution process, key update can only be carried out when specific conditions (such as the number of nodes, signature requirements, etc.) are met, thus avoiding the risks of single point of failure or malicious operations. Push the sub-key components to the slice edge computing nodes through the dynamic authorization channel between consortium chain nodes: Using the dynamic authorization channel between consortium chain nodes, each sub-key component is directed to the edge computing nodes of the uRLLC, eMBB, and mMTC slices. This step ensures the security of the key during transmission through encrypted transmission and access control, and at the same time ensures that the key is only distributed to the appropriate nodes. Record the key version number and the distribution path topology in an immutable block log: Each key distribution records the key version number and the distribution path topology in the block log of the consortium chain. This makes the key distribution process completely transparent, traceable, and cannot be tampered with or retrospectively modified, increasing the credibility and auditability of the system. Combine the encrypted timestamp and the life cycle label of the sharded data block to trigger periodic key update events: Combining the encrypted timestamp and the life cycle label of the sharded data block, the system can automatically trigger periodic key update events. This mechanism ensures that the key is updated regularly during its life cycle, thereby ensuring the long-term security of the data and avoiding the security risks caused by key leakage or failure. Use the cross-chain protocol to synchronously update the key copies of heterogeneous blockchain networks: Through the cross-chain protocol, the key update is not only carried out within a single consortium chain, but also synchronized to the relevant nodes in the heterogeneous blockchain network. This step ensures the consistency and synchronization of the key copies in different blockchain networks, contributing to the security and consistency of the cross-chain system.

[0040] Specifically, the specific process of constructing the fountain code-driven redundant coding strategy is as follows: Dynamically generate the redundancy factor according to the encryption intensity label and spatio-temporal correlation of the sharded data block, and adjust the coding degree distribution function of the fountain code according to the spatio-temporal feature distribution density of the multi-modal data; Arrange the encrypted sharded data block and the redundant check block in a two-dimensional matrix in the frequency domain and the time domain, and dynamically detect the correlation intensity of the sharded group through a sliding window and inject differential redundancy; Combine the real-time packet loss rate of the slice and the transmission reliability index, establish a dynamic adaptation relationship between the redundant coding parameters and the QoS requirements, and generate a multi-slice collaborative redundancy strategy.

[0041] In this implementation scheme, the redundancy factor is dynamically generated based on the encryption intensity label and spatio-temporal correlation of the sharded data blocks: According to the encryption intensity label of each data block (such as high, medium, and low-intensity encryption) and the spatio-temporal correlation of the data (such as the temporal relationship between video frames, the time characteristics of sensor data, etc.), the generation of the redundancy factor is dynamically adjusted. This means that sharded data with a higher encryption intensity may require more redundancy to cope with the computational and transmission overheads brought by high-intensity encryption. For shards with strong spatio-temporal correlation, by generating the redundancy factor, the integrity and redundancy of the data are ensured to be maintained during transmission, preventing packet loss or data corruption. Adjust the encoding degree distribution function of the fountain code according to the spatio-temporal feature distribution density of multimodal data: Analyze the spatio-temporal feature distribution density of multimodal data (video, audio, sensor data, etc.) to adjust the encoding degree distribution function of the fountain code. The spatio-temporal feature distribution density refers to the distribution of data in the spatio-temporal domain, and the encoding degree distribution function determines the allocation method of redundant codes. Through this adjustment, an appropriate amount of redundant data can be generated between different data blocks, optimizing the storage and transmission efficiency of redundancy. Arrange the encrypted sharded data blocks and redundant check blocks in a two-dimensional matrix in the frequency domain and time domain: Combine the encrypted sharded data blocks and redundant check blocks into a two-dimensional matrix in the frequency domain and time domain for arrangement. This process utilizes the characteristics of the frequency domain and time domain, taking into account the distribution characteristics of time and frequency during the encoding process, and can more effectively adapt to the transmission requirements of different network environments. For example, the arrangement in the frequency domain and time domain helps to reduce the packet loss rate of data under different transmission conditions and improve the transmission stability. Dynamically detect the association strength of shard groups through a sliding window and inject differential redundancy: Use the sliding window technique to dynamically detect the association strength between different shard groups. This detection method can evaluate the correlation and redundancy between data. If the association between certain shards is strong, differential redundancy can be injected as needed. For example, shard groups with strong association may require less redundancy, while shard groups with weak association require more redundant data. This strategy can intelligently adjust the quantity and distribution of redundancy according to data characteristics. Combine the real-time packet loss rate and transmission reliability index of slices to establish a dynamic adaptation relationship between redundancy coding parameters and QoS requirements: Dynamically adjust the parameters of redundancy coding according to the real-time packet loss rate and transmission reliability index. A network environment with a high packet loss rate may require more redundancy to ensure data integrity and reliability, while a network with higher transmission reliability can reduce the generation of redundant data. By establishing a dynamic adaptation relationship, it can be ensured that the redundancy coding strategy can meet the required quality of service (QoS) requirements under various network conditions, optimizing the transmission efficiency and data security. Generate a multi-slice collaborative redundancy strategy: In a multi-slice environment, combine information such as the network status, packet loss rate, and transmission reliability of each slice to generate a collaborative redundancy strategy.This means that different slices (such as uRLLC, eMBB, mMTC) will jointly optimize the configuration of redundant coding according to their respective QoS requirements, network conditions, and redundancy requirements, ensuring the coordinated operation between multiple slices and avoiding network overload or waste of redundant resources.

[0042] Specifically, the specific process of optimizing the sharding transmission path in combination with the game theory model and outputting the mapping table of associated shard groups and transmission paths is as follows: Define the available bandwidth of transmission nodes, the slice resource competition coefficient, and the shard security level as the strategy space, and design a composite utility function that includes delay cost, security risk loss, and energy consumption cost; Solve the Nash equilibrium point through the evolutionary game algorithm to generate an initial matching scheme for shard groups and transmission paths, and dynamically correct the path selection weights according to the spatio-temporal key feature vectors of the shard sequence; Combine the blockchain key distribution topology to construct a credibility evaluation mechanism and output the dynamic mapping table of shard groups and slice resources.

[0043] In this implementation plan, define the strategy space: Available bandwidth of transmission nodes: It represents the bandwidth capacity of each transmission node (such as a base station, router, etc.), with the unit of bits per second (bps). This is an important factor affecting the data transmission rate. Slice resource competition coefficient: It represents the degree of competition for each slice resource in a multi-slice network environment. This coefficient measures the difficulty of resource sharing between slices, and the larger the value, the more intense the competition. Shard security level: It represents the security requirements of each data shard, usually based on factors such as encryption strength and data sensitivity. This level affects the security strategy for shard path selection. These parameters constitute the strategy space in the game model, that is, the strategy space that game participants (transmission paths, shards, etc.) can choose. Design a composite utility function: Design a composite utility function: The composite utility function combines delay cost, security risk loss, and energy consumption cost to evaluate the advantages and disadvantages of different path selections. The composite utility function can be expressed as: ; where: : Represents the total utility of path P. : Represents the delay cost of path P, usually calculated as the sum of the delays of each node in the path, with the unit of seconds. : Represents the security risk loss of path P, usually related to the security level and encryption strength of the path. The higher the risk loss, the poorer the security of the path. : Represents the energy consumption cost of path P, usually related to factors such as the number of nodes on the path and bandwidth requirements, with the unit of watts. Parameter description: 、 、 : respectively represent the importance coefficients of time delay, risk, and energy consumption, reflecting the influence degree of different factors on path selection. Solve the Nash equilibrium point through the evolutionary game algorithm: The evolutionary game algorithm is used to simulate the game between shards and transmission paths, and find the strategy balance point between each participant (shard group and transmission path), that is, the Nash equilibrium. Each shard group adjusts its strategy according to the utility function, and the transmission path also makes a choice based on its own resource status. The strategy update in the evolutionary game can be expressed as: ; where: : represents the strategy of the i-th shard group in the t-th round (i.e., the selected transmission path). : represents the average utility of all current paths. : represents the step size of strategy update, controlling the update rate. Through continuous iteration, the evolutionary game finally obtains a stable Nash equilibrium point, representing the optimal strategy allocation between each shard group and the transmission path. Dynamically correct the path selection weight: According to the spatio-temporal key feature vector of the shard sequence (such as factors like time delay, packet loss rate, network congestion, etc.), dynamically adjust the path selection weight. This means that as the network state changes, the weights of certain paths may increase or decrease. For example, when the network is congested, it may be more preferable to select a low-latency path. The update formula for weight adjustment can be expressed as: where: represents the path selection weight of the i-th packet and the j-th transmission path at the t-th moment. represents the spatio-temporal characteristics of the packets corresponding to path j, including time delay, network load, and packet loss. represents the average spatio-temporal characteristics of all paths. represents the step size of weight update. Through this dynamic correction mechanism, the path selection weight can reflect the changes in the network state in real time.

[0044] Specifically, according to the mapping table of associated shard groups and transmission paths, analyze the transmission traffic characteristics and key distribution logs, and the specific process of dynamically adjusting the encryption intensity threshold and shard retransmission strategy is as follows: Real-time monitor the packet loss rate, key distribution time delay, and blockchain consensus verification success rate of the shard transmission path, and construct a joint evaluation matrix of network state and security risk; Establish a dynamic adjustment model for the encryption intensity threshold according to the reinforcement learning algorithm, trigger the shard encryption mode switching event according to the risk level, and generate a shard retransmission priority queue at the same time.

[0045] In this implementation, the performance metrics of the transmission path are monitored in real time: Packet loss rate: The packet loss rate is the proportion of data packets that fail to reach the receiving end successfully during the transmission process among the total transmitted packets. Real-time monitoring of the packet loss rate can help evaluate the reliability of the network, especially in high-load or congested situations. Key distribution delay: The key distribution delay represents the delay from key generation to actual distribution to the target node. A higher key distribution delay may lead to delays in encryption operations and affect the timeliness of data transmission. Blockchain consensus verification success rate: The blockchain consensus verification success rate measures whether the consensus mechanism among nodes is effective and whether it can quickly and reliably verify the legitimacy of key distribution and the transmission path. These performance metrics are collected and analyzed in real time through monitoring and recording. Construct a joint evaluation matrix of network status and security risks: The evaluation matrix is constructed by integrating network performance (such as packet loss rate, latency) and security risks (such as the security of key transmission, the efficiency of blockchain verification). The role of this matrix is to dynamically adjust the transmission strategy according to different network statuses and security risks. The joint evaluation matrix usually includes the following dimensions: Network performance dimension: including packet loss rate, latency, bandwidth utilization, etc. Security risk dimension: including the security of key distribution, the strength of encryption algorithms, the security of the transmission path, etc. Dynamically adjust the encryption strength threshold based on the reinforcement learning algorithm: The reinforcement learning algorithm is used to dynamically adjust the encryption strength threshold according to the network status and security risks. Reinforcement learning continuously updates the strategy through interaction with the environment to maximize a certain goal (such as maximizing the security or reliability of the network). The core of the reinforcement learning model is the state-action-reward mechanism. In this process: State: The current network state (such as packet loss rate, latency, etc.) and the security risk level. Action: The strategy for adjusting the encryption strength threshold. For example, the strength of the encryption algorithm can be adjusted (such as switching from symmetric encryption to stronger asymmetric encryption), or the key length of the encryption algorithm can be changed. Reward: Given a reward based on the joint evaluation of network performance and security risks. For example, a lower packet loss rate and lower latency can obtain a higher reward. Update formula of the reinforcement learning model: ; where: : Represents the expected return of taking action a in state s. : Learning rate, controlling the influence degree of new experiences. r: Reward value of the current step. : Discount factor, measuring the importance of future rewards. s’: New state. a’: New action. According to the results of the joint evaluation matrix, if the current network state and security risk level exceed the set threshold, the event of switching the encryption mode will be triggered. For example, if the packet loss rate is very high or the security of key distribution is insufficient, it can be automatically switched to a higher security encryption mode. The switching event can be achieved through a dynamic trigger mechanism. For example, using a time-window-based trigger mechanism.

[0046] In summary, the present application has at least the following effects:

[0047] The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks balances the encryption intensity, transmission delay, and energy consumption through a multi-objective optimization algorithm, achieving the optimal trade-off between data security and transmission efficiency in different network environments. By using a dynamic adjustment mechanism, it ensures the full utilization of network resources while guaranteeing high data transmission security. According to the fluctuations of network slice resources and the actual transmission situation, it automatically adjusts the encryption algorithm, transmission path, and redundant coding strategy, flexibly coping with various network environment changes and achieving real-time optimization. By precisely controlling the encryption intensity, path selection, and redundancy strategy, it optimizes the network load, reduces unnecessary delay and energy consumption, improves the overall transmission efficiency, and adapts to the resource constraint requirements in different scenarios. Through the path optimization and redundant coding strategy based on game theory, it can effectively avoid data loss during transmission, ensure the integrity and stable transmission of key data, and thus improve the transmission reliability of the system. Through risk assessment and dynamic response instruction generation, it can real-time identify security risks and performance bottlenecks in network transmission and automatically adjust the encryption mode, path selection, and retransmission strategy to cope with different security requirements and transmission conditions.

[0048] Those skilled in the art should understand that the embodiments of the present invention can be provided as a method, a system, or a computer program product. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present invention can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0049] The present invention is described with reference to the flowcharts and / or block diagrams of systems, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, as well as the combination of flows and / or blocks in the flowcharts and / or block diagrams, can be realized by computer program instructions. These computer program instructions can be provided to the processors of general-purpose computers, special-purpose computers, embedded processors, or other programmable data processing devices to generate a machine, such that the instructions executed by the processors of the computer or other programmable data processing devices generate means for realizing the functions specified in one flow Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0050] These computer program instructions can also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to work in a particular manner, such that the instructions stored in the computer-readable memory produce a manufacture including an instruction device that implements the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0051] These computer program instructions can also be loaded onto a computer or other programmable data processing apparatus, such that a series of operational steps are performed on the computer or other programmable apparatus to produce a computer-implemented process, whereby the instructions executed on the computer or other programmable apparatus provide steps for implementing the functions specified in one or more of the processes and / or blocks Figure 1 in one or more of the processes and / or blocks Figure 1 specified in one or more of the blocks or blocks.

[0052] Although the preferred embodiments of the present invention have been described, additional changes and modifications can be made by those skilled in the art once they learn of the basic creative concept. Therefore, the appended claims are intended to be construed to include the preferred embodiments as well as all changes and modifications that fall within the scope of the present invention.

[0053] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention is also intended to include these modifications and variations.

Claims

1. A real-time fragmentation transmission method for heterogeneous encrypted multimodal multimedia messages for 5G networks, characterized in that It includes the following steps: S1. Obtain the sensitivity label and value density of multimedia data, and monitor the resource status of 5G network slices in real time. Dynamically generate the mapping rule between slice granularity and priority through a reinforcement learning algorithm, extract spatio-temporal key features by combining a convolutional network, and output a slice sequence adapted to 5G slice resources and the corresponding encryption intensity label; S2. Receive the slice sequence and encryption intensity label, dynamically select an encryption algorithm cluster according to the mapping rule between slice granularity and priority, realize cross-slice key distribution and update through a consortium blockchain, and generate encrypted slice data blocks and key distribution logs; S3. According to the spatio-temporal correlation between the encrypted slice data blocks and multimodal data, construct a fountain code-driven redundant coding strategy, optimize the slice transmission path by combining a game theory model, and output a mapping table of associated slice groups and transmission paths; S4. According to the mapping table of associated slice groups and transmission paths, analyze the transmission traffic characteristics and key distribution logs, and dynamically adjust the encryption intensity threshold and slice retransmission strategy.

2. The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 1, characterized in that: The specific process of obtaining the sensitivity label and value density of multimedia data and monitoring the resource status of 5G network slices in real time is as follows: According to the distribution density of sensitive words in the text data annotated by natural language processing, extract the key frame motion trajectory features of video data, analyze the frequency domain burst frequency of audio data, and combine the mutation amplitude of the sensor waveform to dynamically calculate the value density level of multimedia data; Real-time collect the remaining bandwidth of the uRLLC slice, the delay jitter of the eMBB slice, and the packet loss rate of the mMTC slice through the 5G core network interface to generate a 5G network resource status vector.

3. The heterogeneous encryption multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 2, characterized in that: The specific process of dynamically generating the mapping rule between slice granularity and priority through a reinforcement learning algorithm is as follows: Construct a state space with the multimedia data value density feature vector and the 5G network slice resource status vector as inputs, define the slice granularity level and transmission priority weight as the action space, design a multi-objective reward function to comprehensively consider the encryption intensity matching degree, delay deviation, and terminal energy consumption, and generate a dynamic mapping matrix of slice granularity and priority through policy iteration.

4. The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 3, wherein: The specific process of extracting spatio-temporal key features by combining a convolutional network and outputting a slice sequence adapted to 5G slice resources and the corresponding encryption intensity label is as follows: Extract spatio-temporal features of the video stream through a lightweight convolutional network, identify the motion correlation between key frames and the burst silent segments in the audio stream, and fuse the frequency domain mutation points of the sensor waveform to generate a spatio-temporal key feature vector; According to the multimedia data value density feature vector and the 5G slice resource status, dynamically divide the slice boundary and mark the encryption intensity label, and output a slice sequence adapted to the characteristics of the uRLLC, eMBB, and mMTC slices.

5. The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 4, characterized in that: The specific process of dynamically selecting an encryption algorithm cluster according to the mapping rule between slice granularity and priority is as follows: Construct a heterogeneous encryption algorithm pool, generate a multi-dimensional matching index according to the slice priority weight and encryption intensity label, and statistically analyze the network jitter coefficient and computing load fluctuation characteristics of historical slices through a sliding time window; According to the remaining bandwidth of the current slice resource and the terminal hardware acceleration capability, dynamically combine symmetric encryption algorithms and lightweight asymmetric encryption algorithms to form an algorithm cluster with linked shard granularity and encryption strength, including: Fine-grained sharding: Bind the attribute-based homomorphic encryption algorithm, generate dynamic keys with the spatio-temporal characteristics of multimodal data as attribute parameters. The spatio-temporal characteristics of multimodal data include video key frame hash values and sensor waveform peaks; Medium-grained sharding: Through the hierarchical hybrid encryption algorithm, cross-encrypt data blocks, and the keys are dynamically distributed through the edge node consensus protocol; Coarse-grained sharding: Through dynamic mask perturbation encryption, the keys are issued on demand through 5G control plane signaling to adapt to the low-power consumption transmission requirements; According to the spatio-temporal key feature vector of the shard sequence, trigger the key length adaptive adjustment mechanism and encryption mode switching event to realize the real-time coupling of the algorithm cluster parameters and the network slice QoS indicators.

6. The heterogeneous encrypted multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 5, characterized in that: The specific process of realizing cross-slice key distribution and update through the consortium chain, generating encrypted shard data blocks and key distribution logs is as follows: Split the shard key into multiple sub-key components according to the slice domain, and define the threshold signature rules and consensus verification conditions for cross-domain key distribution; Through the dynamic authorization channel between consortium chain nodes, push the sub-key components to the edge computing nodes corresponding to the uRLLC, eMBB, and mMTC slices, and record the key version number and distribution path topology in the immutable block log; Combined with the encryption timestamp and life cycle label of the shard data block, trigger the periodic key update event, and use the cross-chain protocol to synchronously update the key copies of heterogeneous blockchain networks.

7. The real-time sharding transmission method of heterogeneous encrypted multi-modal multimedia messages for 5G networks according to claim 6, characterized in that: The specific process of constructing a fountain code-driven redundant coding strategy is as follows: Dynamically generate redundant factors according to the encryption strength label and spatio-temporal correlation of the shard data block, and adjust the encoding degree distribution function of the fountain code according to the spatio-temporal feature distribution density of multimodal data; Arrange the encrypted shard data blocks and redundant check blocks in a two-dimensional matrix in the frequency domain and time domain, dynamically detect the correlation strength of the shard group through a sliding window and inject differential redundancy; Combined with the real-time packet loss rate and transmission reliability index of the slice, establish a dynamic adaptation relationship between the redundant coding parameters and QoS requirements, and generate a multi-slice collaborative redundant strategy.

8. The heterogeneous encryption multi-modal multimedia message real-time fragmentation transmission method for 5G networks according to claim 7, wherein: The specific process of optimizing the shard transmission path in combination with the game theory model and outputting the mapping table of associated shard groups and transmission paths is as follows: Define the available bandwidth of the transmission node, the slice resource competition coefficient, and the shard security level as the strategy space, and design a composite utility function including delay cost, security risk loss, and energy consumption cost; Solve the Nash equilibrium point through the evolutionary game algorithm, generate the initial matching scheme of the shard group and the transmission path, and dynamically correct the path selection weight according to the spatio-temporal key feature vector of the shard sequence; Construct a credibility evaluation mechanism in combination with the blockchain key distribution topology, and output the dynamic mapping table of the shard group and slice resources.

9. The heterogeneous encryption multi-modal multimedia message real-time fragmentation transmission method for a 5G network according to claim 8, wherein: According to the mapping table of associated shard groups and transmission paths, analyze the transmission traffic characteristics and key distribution logs, and dynamically adjust the encryption strength threshold and shard retransmission strategy. The specific process is as follows: Real-time monitor the packet loss rate, key distribution delay, and blockchain consensus verification success rate of the shard transmission path, and construct a joint evaluation matrix of network status and security risks; Establish a dynamic adjustment model for the encryption strength threshold based on the reinforcement learning algorithm, trigger the sharding encryption mode switching event according to the risk level, and generate a sharding retransmission priority queue at the same time.

Images