A communication data transmission method based on data encryption

By constructing a heterogeneous graph for ciphertext transmission and using the ML-KEM algorithm, the problem of inaccurate ciphertext segment scheduling in existing communication data transmission methods is solved, achieving efficient encryption and decryption in multi-session, multi-link, and out-of-order transmission scenarios, thus improving the stability and reliability of data transmission.

CN123055429APending Publication Date: 2026-11-06TIANJIN HANYUAN TECHNOLOGY CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611362112.0
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-09-03
Publication Date
2026-11-06

AI Technical Summary

Technical Problem

Existing communication data transmission methods lack fine-grained correlation modeling of ciphertext fragments, making it difficult to accurately schedule and reassemble in cases of out-of-order transmission, delayed arrival, and duplicate transmission. This results in unclear key correspondences, insufficient basis for fragment reassembly, and delayed feedback updates.

Method used

By employing standardized communication segment processing, ciphertext transmission heterogeneous graph construction, improved MH-Net multi-view heterogeneous graph network, ML-KEM algorithm key encapsulation, segment path key derivation processing, and out-of-order segment key backfilling processing, we can achieve segment-by-segment encryption and verification decryption, and establish a causal chain between segment sending, receiving, acknowledgment feedback, and retransmission feedback.

Benefits of technology

In scenarios involving multiple sessions, multiple links, and out-of-order transmission, the system achieves clear scheduling relationships for encrypted fragments, accurate key matching for out-of-order fragments, minimal interference from abnormal fragment propagation, and timely transmission feedback updates, thereby improving the confidentiality and integrity of communication data transmission.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123055429A_ABST
    Figure CN123055429A_ABST
Patent Text Reader

Abstract

This invention discloses a data encryption-based communication data transmission method, relating to the field of data encryption technology. The method includes: collecting basic communication data and generating a standardized communication fragment dataset; extracting various transmission relationships and constructing a ciphertext transmission heterogeneous graph; constructing an improved MH-Net network and outputting ciphertext fragment transmission status packets; determining the fragment transmission order and generating a ciphertext fragment scheduling packet; performing ML-KEM key encapsulation to generate a key encapsulation encryption packet; encrypting each fragment and binding an identifier to generate a ciphertext transmission packet; and verifying, decrypting, and reassembling the data at the receiving end to generate recovered communication data. This invention, by introducing an improved MH-Net multi-view heterogeneous graph network and the ML-KEM algorithm, achieves secure encrypted transmission, dynamic key derivation, and reliable reassembly at the receiving end of communication data in scenarios involving fragmentation, out-of-order transmission, and multi-link feedback.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of data encryption technology, and in particular to a communication data transmission method based on data encryption. Background Technology

[0002] With the rapid development of network communication, IoT terminals, cloud business systems, and mobile application platforms, communication data is frequently transmitted between terminal devices, servers, edge nodes, and business platforms. The transmitted content includes identity authentication data, business instruction data, payload data, timestamp data, and verification data. Existing communication data transmission methods typically employ a combination of symmetric encryption, public-key negotiation, integrity verification, and transmission protocol protection to ensure secure data transmission. The sending end encrypts and encapsulates the communication data before transmission, and the receiving end decrypts and verifies it upon receipt to guarantee the confidentiality and integrity of the communication data across public or untrusted links. Some methods further introduce session keys, dynamic tokens, digital signatures, or transmission feedback mechanisms to manage the identities of the communicating parties, the transmitted content, and abnormal states, thereby reducing the risk of communication data being intercepted, tampered with, forged, or replayed.

[0003] Existing data encryption-based communication data transmission methods still have shortcomings. Traditional methods often process entire packets of communication data, typically using a unified session key or fixed derivation rules for encryption within a single session. They lack fine-grained correlation modeling for communication data segments, session rounds, link states, and feedback states. When ciphertext segments are sent out of order, arrive late, are repeatedly transmitted, or remain across tasks, existing methods usually rely on receiver verification and retransmission. This makes it difficult to uniformly schedule the transmission order, time window, and index association state of ciphertext segments before transmission, and also makes it difficult to express the causal relationship between segment transmission, segment reception, acknowledgment feedback, and retransmission feedback. Existing technologies also lack handling of the propagation neighborhood of segments to be verified and the state of remnant segments across tasks. This can easily cause abnormal segments to affect the transmission judgment of normal segments, leading to unclear key correspondences, insufficient basis for segment reassembly, and delayed feedback updates at the receiver during decryption, reassembly, and feedback updates.

[0004] Therefore, how to provide a data encryption-based communication data transmission method is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] One objective of this invention is to propose a data encryption-based communication data transmission method. This invention utilizes standardized communication segment processing, ciphertext transmission heterogeneous graph construction, an improved MH-Net multi-view heterogeneous graph network, ML-KEM algorithm key encapsulation, segment path key derivation processing, out-of-order segment key backfilling processing, and receiver-side verification and reassembly feedback technology. It details the implementation process of communication data at the sending end, including field marking, segment division, ciphertext segment scheduling, segment key generation, segment-by-segment encryption, and scheduled transmission, as well as key recovery, verification decryption, and segment reassembly at the receiving end. This achieves secure encrypted transmission of communication data in multi-session, multi-link, and out-of-order transmission scenarios. Compared to traditional fixed-key encryption transmission methods, this invention offers advantages such as refined segment key derivation, clear ciphertext segment scheduling relationships, accurate out-of-order segment key matching, less interference from abnormal segment propagation, and timely transmission feedback updates.

[0006] A data encryption-based communication data transmission method according to an embodiment of the present invention includes: Collect basic communication data, perform preprocessing, field labeling and fragment segmentation on the basic communication data, and generate a standardized communication fragment dataset; Based on the standardized communication segment dataset, segment association, session round relationship, link transmission relationship and feedback status relationship are extracted, and various relationships are mapped into a graph structure to construct a heterogeneous graph for encrypted transmission. An improved MH-Net multi-view heterogeneous graph network is constructed, multi-view relationship encoding is performed on the ciphertext transmission heterogeneous graph, a causal chain between fragment sending, fragment receiving, acknowledgment feedback and retransmission feedback is established based on the session round causal chain reconstruction layer, the propagation neighborhood of the fragment to be verified is shrunk through the ciphertext fragment trusted neighborhood shrinking layer, and the state of the fragments left across tasks is recycled through the task boundary state recycling layer, and the ciphertext fragment transmission state packet is output. The transmission order, transmission time window, and index association status of communication data segments are determined based on the encrypted segment transmission status packet, and an encrypted segment scheduling packet is generated. The key encapsulation is performed based on the ML-KEM algorithm to generate a session shared key and key encapsulation data. Combined with the ciphertext fragment scheduling packet, the key encapsulation encryption packet is generated through fragment path key derivation processing and out-of-order fragment key backfilling processing. The sequence of fragments to be encrypted and the order of fragment transmission are determined based on the key encapsulation encryption packet. Segment-by-segment encryption and integrity identifier binding are performed on the sequence of fragments to be encrypted, and the bound ciphertext fragments are sent in the order of fragment transmission to generate ciphertext transmission packets. The receiving end performs key recovery, verification decryption, and fragment reconstruction based on the encrypted transmission packet, generates recovered communication data and transmission feedback records, updates the encrypted transmission heterogeneous graph, improves the MH-Net multi-view heterogeneous graph network parameters, and the key derivation parameters of the ML-KEM algorithm.

[0007] Optionally, generating the standardized communication fragment dataset includes: Collect basic communication data, which includes communication data to be transmitted, sender identity identifier, receiver identity identifier, session round record, link status record, and historical feedback record; Standardized communication basic data is generated by performing format unification, invalid record removal, duplicate record deletion, and timestamp alignment on the basic communication data. Perform field parsing on the communication data to be transmitted in the standardized communication basic data, and generate field tags according to the field position and field type. The field tags include identity field tags, business field tags, payload field tags, time field tags, and verification field tags. The communication data to be transmitted is segmented according to the field tags, data length, and session round record. Each communication data segment is configured with a communication data packet identifier, segment number, field tags, session round identifier, link identifier, and feedback status identifier to generate a communication data segment sequence. The communication data fragment sequence is associated with the sender's identity identifier, receiver's identity identifier, link status record, and historical feedback record to generate a standardized communication fragment dataset.

[0008] Optionally, constructing the ciphertext transmission heterogeneous graph includes: The standardized communication segment dataset is organized according to communication data packet identifier, segment sequence number, field label, session round identifier, link identifier, and feedback status identifier to generate a segment index record set; Based on the sequence number continuity of fragments in the fragment index record set, the field tag attribution relationship, and the relationship between fragments within the same communication data packet, a fragment association relationship is generated; The session round relationship is generated based on the session round identifier, sender identity identifier, receiver identity identifier, acknowledgment feedback order, and retransmission identifier in the fragment index record set; Extract the link identifier, sending timestamp, receiving timestamp, and segment arrival status from the link status records associated with the segment index record set, and generate the link transmission relationship. Generate feedback status relationships based on historical feedback records, verification status, missing status, duplicate status, and retransmission status in the fragment index record set; The communication data fragments, session rounds, transmission links, and feedback status are mapped as graph nodes, and various relationships are mapped as graph edges. Field tags, link status records, and feedback status identifiers are written to construct a heterogeneous graph for encrypted transmission.

[0009] Optionally, the output encrypted fragment transmission status packet includes: An improved MH-Net multi-view heterogeneous graph network is constructed, which includes a session round causal chain reconstruction layer, a ciphertext fragment trusted neighborhood shrinking layer, and a task boundary state recycling layer. The session round causal chain reconstruction layer receives various relationships in the heterogeneous graph of encrypted transmission, arranges the segment sending node, segment receiving node, acknowledgment feedback node and retransmission feedback node according to the session round identifier, establishes the sending-receiving relationship, receiving-acknowledgment relationship, verification failure-retransmission relationship and acknowledgment feedback-next round sending relationship, and generates the session round causal chain and segment causal state representation; The trusted neighborhood shrinking layer of ciphertext fragments is based on the causal chain of session rounds and the causal state representation of fragments. It divides the neighboring fragment nodes in the heterogeneous graph of ciphertext transmission into neighborhoods. The neighboring fragment nodes with consistent session rounds, consistent link identifiers, and matching feedback states are included in the trusted fragment neighborhood. The neighboring fragment nodes with inconsistent session rounds, missing feedback states, or mismatched link identifiers are included in the neighborhood of fragments to be verified. The propagation range of the neighborhood of fragments to be verified is shrunk to generate a trusted neighborhood fragment graph. The task boundary state recycling layer, based on the trusted neighborhood fragment graph, session round causal chain, and feedback state relationship, marks the start fragment, end fragment, retransmission termination fragment, and cross-task residual fragment of the current communication task. It performs recycling processing on the state of cross-task residual fragments, generates a task boundary recycling graph, and generates ciphertext fragment transmission state packets based on the task boundary recycling graph. The improved MH-Net multi-view heterogeneous graph network was trained using a combination of session round causal chain reconstruction error, trusted neighborhood partitioning error, task boundary state recovery error, and ciphertext fragment transmission state prediction error as the joint optimization objective. The network parameters of each layer were continuously optimized. When the change of the joint loss value in 5 consecutive training rounds was less than 0.001, the improved MH-Net multi-view heterogeneous graph network was considered to have completed convergence training.

[0010] Optionally, the generation of the ciphertext fragment scheduling packet includes: The encrypted segment transmission status representation in the encrypted segment transmission status packet is organized according to the session round identifier, communication data packet identifier and segment sequence number to generate a scheduling candidate segment sequence; Based on the causal state of the segments, trusted neighborhood identifier, task boundary reclamation identifier, and link transmission state in the scheduling candidate segment sequence, the scheduling state of each communication data segment is determined. The scheduling state includes the sending state, the delayed state, the retransmission state, and the stripping state. The communication data segments in the stripped state are removed from the scheduling candidate segment sequence. The remaining communication data segments are sorted according to the session round identifier, segment causal order and link transmission state to generate a segment sequence to be encrypted. Based on the causal order of the fragments in the sequence of fragments to be encrypted, the link transmission status, and the transmission time window, a fragment transmission order table is generated, and the transmission round identifier and transmission time window identifier are configured. Based on the fragment sequence number, transmission round identifier, transmission time window identifier, and receiver reassembly position, generate an out-of-order mapping table and an index association table; The sequence of fragments to be encrypted, the fragment sending order table, the out-of-order mapping table, and the index association table are associated to generate a ciphertext fragment scheduling packet.

[0011] Optionally, the generation of the key to encapsulate the encrypted packet includes: The receiving end generates an ML-KEM encapsulation public key and an ML-KEM decapsulation private key, and sends the ML-KEM encapsulation public key to the sending end; The sending end performs key encapsulation based on the ML-KEM encapsulated public key, generates a session shared key and key encapsulation data, and associates them with the ciphertext fragment scheduling packet to establish a session key derivation extension record of the ML-KEM algorithm; Based on the sequence of fragments to be encrypted, the fragment transmission order table, the out-of-order mapping table, and the index association table in the encrypted fragment scheduling packet, generate a fragment path record including session round identifier, fragment sequence number, field marker, link identifier, transmission time window identifier, and index association identifier; In the session key derivation extension record, the session shared key and the fragment path record are processed to generate the initial fragment key sequence. In the session key derivation extension record, the initial fragment key sequence is backfilled with out-of-order fragment keys according to the fragment transmission order table and the out-of-order mapping table to generate a backfilled fragment key sequence. The key-encapsulated data, the sequence of fragments to be encrypted, the fragment sending order table, the fragment path record, the backfilled fragment key sequence, and the index association table are bound together to generate a key-encapsulated encryption package.

[0012] Optionally, generating the encrypted transmission packet includes: Based on the sequence of fragments to be encrypted, the fragment transmission order table, the backfilled fragment key sequence, and the index association table in the key-encapsulated encryption packet, determine the correspondence between the fragment encryption keys and the fragment transmission order; According to the segment encryption key correspondence, each communication data segment in the segment sequence to be encrypted is matched with the corresponding backfilled segment key to generate a segment encryption task table; Each communication data segment is encrypted segment by segment according to the segment encryption task table, generating a set of ciphertext segments; An integrity identifier is generated based on the ciphertext fragments, fragment sequence numbers, sending time window identifiers, and index association identifiers in the ciphertext fragment set. The integrity identifier is then bound to the corresponding ciphertext fragment, generating a bound ciphertext fragment set. The encrypted fragment set is sent in the order of fragment transmission, and the encrypted transmission packet is generated by associating the key-encapsulated data, fragment path record, index association table and transmission record.

[0013] Optionally, generating recovery communication data and transmission feedback records and performing feedback updates includes: The receiving end parses the encrypted transmission packet to obtain the bound encrypted fragment set, key encapsulation data, fragment path record, index association table and sending record; The receiving end decapsulates the key-encapsulated data based on the ML-KEM decapsulation private key, recovers the session shared key, and generates a fragment decryption key sequence based on the session shared key, fragment path record, index association table, and sending record; Integrity verification is performed based on the integrity identifier in the bound ciphertext fragment set. Ciphertext fragments that pass the verification are written into the fragment set to be decrypted, and ciphertext fragments that fail the verification are written into the transmission feedback record. The decryption process is performed segment by segment according to the decryption key sequence. The decrypted communication data segments are then sorted and reassembled based on the index association table and the sending record to generate the recovered communication data. Based on the verification failure records, missing fragment records, retransmission records, and reassembly status records in the transmission feedback records, update the ciphertext transmission heterogeneous graph, improve the MH-Net multi-view heterogeneous graph network parameters, and improve the ML-KEM algorithm key derivation parameters.

[0014] The beneficial effects of this invention are: This invention proposes a data encryption-based communication data transmission method. By preprocessing, labeling fields, and segmenting basic communication data, a standardized communication segment dataset is formed. Furthermore, a heterogeneous graph of encrypted transmission is constructed, enabling communication data to be transmitted in a structured manner, rather than as a single encrypted packet. This model follows segment relationships, session round relationships, link transmission relationships, and feedback state relationships. Compared to traditional whole-packet encryption methods, this invention can more accurately describe the transmission relationships of communication data segments in multiple sessions, multiple links, and multiple feedback states. This provides a clear data foundation for subsequent encrypted segment scheduling, segment key derivation, and receiver segment reassembly, reducing transmission processing errors caused by out-of-order segments, link anomalies, and inconsistent feedback states.

[0015] This invention constructs an improved MH-Net multi-view heterogeneous graph network, comprising a session round causal chain reconstruction layer, a ciphertext fragment trusted neighborhood shrinking layer, and a task boundary state recovery layer. This network encodes multi-view relationships in the heterogeneous graph of ciphertext transmission, establishing causal chains between fragment transmission, fragment reception, acknowledgment feedback, and retransmission feedback. It also processes the propagation neighborhood of fragments to be verified and the state of residual fragments across tasks. Through this structure, the ciphertext fragment transmission state packet can more accurately reflect the transmission order, transmission time window, and index association state of communication data fragments, reducing interference from abnormal fragments, residual fragments, and error feedback on normal fragment scheduling, and improving the stability and reliability of the ciphertext fragment scheduling process.

[0016] This invention performs key encapsulation using the ML-KEM algorithm, combining fragment path key derivation processing and out-of-order fragment key backfilling processing to generate a key-encapsulated encryption packet. This ensures that the fragment key no longer relies solely on a unified session key or a fixed fragment sequence number, but is integrated with the ciphertext fragment scheduling packet, fragment path record, out-of-order mapping relationship, and index association state. The sending end can perform segment-by-segment encryption and integrity identifier binding on the sequence of fragments to be encrypted, while the receiving end can complete key recovery, verification decryption, and fragment reassembly based on the ciphertext transmission packet. Furthermore, it updates the ciphertext transmission heterogeneous graph through transmission feedback records, improves the MH-Net multi-view heterogeneous graph network parameters, and enhances the key derivation parameters of the ML-KEM algorithm. This improves the confidentiality, integrity, tamper resistance, and recovery accuracy of communication data under conditions of out-of-order transmission, fragment retransmission, link fluctuations, and abnormal feedback. Attached Figure Description

[0017] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of a data encryption-based communication data transmission method proposed in this invention; Figure 2 This is a schematic diagram of the structure of an improved MH-Net multi-view heterogeneous graph network based on a data encryption-based communication data transmission method proposed in this invention. Figure 3 This is a schematic diagram illustrating the process of generating a key-encapsulated encryption packet using the ML-KEM algorithm, a data encryption-based communication data transmission method proposed in this invention. Detailed Implementation

[0018] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are simplified schematic diagrams, illustrating only the basic structure of the invention, and therefore only show the components relevant to the invention.

[0019] refer to Figure 1 , Figure 2 and Figure 3 A data encryption-based communication data transmission method, comprising: Collect basic communication data, perform preprocessing, field labeling and fragment segmentation on the basic communication data, and generate a standardized communication fragment dataset; Based on the standardized communication segment dataset, segment association, session round relationship, link transmission relationship and feedback status relationship are extracted, and various relationships are mapped into a graph structure to construct a heterogeneous graph for encrypted transmission. An improved MH-Net multi-view heterogeneous graph network is constructed, multi-view relationship encoding is performed on the ciphertext transmission heterogeneous graph, a causal chain between fragment sending, fragment receiving, acknowledgment feedback and retransmission feedback is established based on the session round causal chain reconstruction layer, the propagation neighborhood of the fragment to be verified is shrunk through the ciphertext fragment trusted neighborhood shrinking layer, and the state of the fragments left across tasks is recycled through the task boundary state recycling layer, and the ciphertext fragment transmission state packet is output. The transmission order, transmission time window, and index association status of communication data segments are determined based on the encrypted segment transmission status packet, and an encrypted segment scheduling packet is generated. The key encapsulation is performed based on the ML-KEM algorithm to generate a session shared key and key encapsulation data. Combined with the ciphertext fragment scheduling packet, the key encapsulation encryption packet is generated through fragment path key derivation processing and out-of-order fragment key backfilling processing. The sequence of fragments to be encrypted and the order of fragment transmission are determined based on the key encapsulation encryption packet. Segment-by-segment encryption and integrity identifier binding are performed on the sequence of fragments to be encrypted, and the bound ciphertext fragments are sent in the order of fragment transmission to generate ciphertext transmission packets. The receiving end performs key recovery, verification decryption, and fragment reconstruction based on the encrypted transmission packet, generates recovered communication data and transmission feedback records, updates the encrypted transmission heterogeneous graph, improves the MH-Net multi-view heterogeneous graph network parameters, and the key derivation parameters of the ML-KEM algorithm.

[0020] In this embodiment, generating a standardized communication fragment dataset includes: Collect basic communication data, which includes communication data to be transmitted, sender identity identifier, receiver identity identifier, session round record, link status record, and historical feedback record; Standardized communication basic data is generated by performing format unification, invalid record removal, duplicate record deletion, and timestamp alignment on the basic communication data, including: Generate standardized communication base data, specifically: The communication data to be transmitted, the sender's identity identifier, the receiver's identity identifier, the session round record, the link status record, and the historical feedback record are uniformly converted into the basic communication record. The basic communication record includes the data content, data length, sender's identity identifier, receiver's identity identifier, session round identifier, link identifier, feedback status identifier, and collection timestamp. Invalid records are removed from the basic communication records. When the data length is 0, or when any of the sender's identity identifier, receiver's identity identifier, session round identifier, link identifier, or collection timestamp is empty, the corresponding basic communication record is deleted. Duplicate record deletion is performed on the remaining basic communication records. The data content, sender identity identifier, receiver identity identifier, session round identifier and collection timestamp are concatenated in sequence and then digest operation is performed to generate a record digest value. The record with the earliest collection timestamp among the records with the same record digest value is retained. Using the earliest collected timestamp within the same session round as the baseline timestamp, and setting the time alignment window to 5ms, the collected timestamp of each basic communication record is assigned to the corresponding time alignment window and written into the alignment timestamp to generate standardized basic communication data. The system performs field parsing on the communication data to be transmitted in the standardized communication foundation data, generating field tags according to field position and field type. These field tags include identity field tags, service field tags, payload field tags, time field tags, and verification field tags. The process of parsing fields in the communication data to be transmitted is as follows: For each basic communication record in the standardized basic communication data, a byte offset sequence is established. The starting position of the byte offset sequence is 0, and the ending position is the data length minus 1. An identity field marker is generated based on the byte range corresponding to the sender's identity identifier and the receiver's identity identifier. Generate business field tags according to the byte ranges corresponding to the business type code, instruction code, and control parameters; Generate payload field markers according to the byte range corresponding to the text data; Generate a time field marker based on the byte range corresponding to the sending timestamp and the session time window; Generate a verification field marker based on the byte range corresponding to the verification value and the reserved verification field; Associate each field marker with its corresponding byte start position, byte end position, field length, and session round identifier to generate a field marker record; When there is an unmarked byte range between two adjacent field marks, the unmarked byte range is merged into the byte range corresponding to the adjacent payload field mark, and the field mark record is updated; The communication data to be transmitted is segmented according to field tags, data length, and session round record. Each communication data segment is configured with a communication data packet identifier, segment sequence number, field tags, session round identifier, link identifier, and feedback status identifier, generating a communication data segment sequence, where: The communication data to be transmitted is segmented into segments, specifically as follows: Determine the byte start position, byte end position and field length corresponding to each field based on the field marker record, and establish a field queue to be divided according to the arrangement order of the field markers in the communication data to be transmitted. The target segment length of a single communication data segment is set to 4096 bytes. When the field length is greater than 4096 bytes, the byte range corresponding to the field is continuously segmented in units of 4096 bytes. When the remaining byte length of the field is less than 4096 bytes, the remaining bytes are formed into a separate communication data segment. When the field length is less than or equal to 4096 bytes, the byte range corresponding to the field is formed into a communication data segment, and the field tag corresponding to the field is retained; Generate segment numbers according to the order of communication data segments in the communication data to be transmitted, write the communication data packet identifier corresponding to the same communication data to be transmitted into each communication data segment, and write the field marker in the field marker record and the session round identifier in the session round record into the corresponding communication data segment. Configure link identifiers for communication data segments based on link status records, configure feedback status identifiers for communication data segments based on historical feedback records, and generate a sequence of communication data segments; The communication data fragment sequence is associated with the sender's identity identifier, receiver's identity identifier, link status record, and historical feedback record to generate a standardized communication fragment dataset.

[0021] In this embodiment, constructing the ciphertext transmission heterogeneous graph includes: The standardized communication segment dataset is organized according to communication data packet identifier, segment sequence number, field label, session round identifier, link identifier, and feedback status identifier to generate a segment index record set, wherein: Generate a fragment index record set, specifically as follows: Using communication data segments in the standardized communication segment dataset as index objects, the communication data segments are grouped according to the communication data packet identifier to generate segment groups under the same communication data packet; Within each segment group, the communication data segments are arranged in ascending order of their segment numbers; Associate field tags, session round identifiers, link identifiers, feedback status identifiers, sender identity identifiers, receiver identity identifiers, link status records, and historical feedback records with the corresponding segment sequence numbers to form a single segment index record; The individual fragment index records are combined according to the communication data packet identifier and fragment sequence number to generate a fragment index record set; Based on the sequence number continuity of fragments in the fragment index record set, the field tag attribution relationship, and the sequence relationship of fragments within the same communication data packet, a fragment association relationship is generated, where: Generate fragment association relationships, specifically as follows: In the fragment index record set, single fragment index records with the same communication data packet identifier are grouped into the same fragment association group; Within the same fragment association group, single fragment index records are arranged in ascending order of fragment number. When the difference between the fragment number of the subsequent single fragment index record and the fragment number of the previous single fragment index record is 1, a fragment association relationship is established. Single-fragment index records with the same field label are grouped into the same field affiliation group, and field affiliation relationships are established within the same field affiliation group; When there are more than two consecutive segment numbers in the same field's group, the field's continuous association relationship is established according to the order of the consecutive segment numbers. Combine the relationships between fragments before and after, the relationships between fields belonging to, and the relationships between fields consecutively to generate fragment relationships; Based on the session round identifier, sender identity identifier, receiver identity identifier, acknowledgment order, and retransmission identifier in the fragment index record set, a session round relationship is generated, where: The session turn sequence is generated as follows: In the fragment index record set, single fragment index records with the same session round identifier are grouped into the same session round group; Within each session round group, a segment transmission relationship is established between the sender and receiver based on the sender's identity and the receiver's identity. Establish confirmation feedback relationships between the receiving end feedback records and the corresponding communication data segments according to the confirmation feedback order; When a retransmission identifier exists in a single-fragment index record, a retransmission association is established between the communication data fragment corresponding to the retransmission identifier and the original sent fragment. The segment sending relationship, acknowledgment feedback relationship, and retransmission relationship are combined according to the session round identifier to generate the session round relationship; Based on the link status records associated with the fragment index record set, extract the link identifier, sending timestamp, receiving timestamp, and fragment arrival status to generate link transmission relationships, where: The link transmission relationship is generated as follows: In the fragment index record set, single fragment index records with the same link identifier are grouped into the same link transmission group; Within each link transmission group, single-segment index records are arranged in ascending order of transmission timestamps to form a link segment transmission sequence. For each communication data segment, the transmission delay is calculated. The transmission delay is the receive timestamp minus the send timestamp. When the receive timestamp is empty, the segment arrival status is marked as missing. Extract the link availability status, link switching status, and link retransmission status from the link status record, and write the link availability status, link switching status, and link retransmission status into the corresponding link transmission group; The link transmission relationship is generated by associating communication data segments, link identifiers, sending timestamps, receiving timestamps, transmission delays, and segment arrival status. Based on the historical feedback records, checksums, missing statuses, duplicate statuses, and retransmission statuses in the fragment index record set, a feedback status relationship is generated, where: The feedback state relationship is generated as follows: In the fragment index record set, historical feedback records are categorized according to communication data packet identifier, fragment sequence number, and session round identifier to generate fragment feedback record groups; Within each segment feedback record group, extract the verification status, missing status, duplicate status, and retransmission status, and write the verification status, missing status, duplicate status, and retransmission status into the corresponding communication data segment. When the verification status is verification failure, establish a feedback association between the corresponding communication data segment and the verification failure feedback; When a missing state, a duplicate state, or a retransmission state exists, establish a feedback association relationship between the corresponding communication data segment and the missing feedback, duplicate feedback, or retransmission feedback. The feedback relationships corresponding to verification failure feedback, missing feedback, duplicate feedback and retransmission feedback are combined to generate feedback status relationships; The communication data fragments, session rounds, transmission links, and feedback states are mapped as graph nodes, and various relationships are mapped as graph edges. Field tags, link status records, and feedback status identifiers are written to construct a heterogeneous graph for encrypted transmission, where: Constructing a heterogeneous graph for encrypted transmission, specifically: The communication data fragments, session round identifiers, link identifiers, and feedback status are mapped to fragment nodes, session nodes, link nodes, and feedback nodes, respectively. The fragment association relationship, session round relationship, link transmission relationship and feedback state relationship are mapped to fragment association edge, session round edge, link transmission edge and feedback state edge, respectively; Write the field tags, fragment sequence numbers, communication data packet identifiers, link status records, transmission delays, and feedback status identifiers into the corresponding graph nodes or graph edges to generate a heterogeneous graph for encrypted transmission.

[0022] In this embodiment, the output ciphertext fragment transmission status packet includes: An improved MH-Net multi-view heterogeneous graph network is constructed, comprising a session round causal chain reconstruction layer, a ciphertext fragment trusted neighborhood shrinking layer, and a task boundary state recycling layer, wherein: Construct an improved MH-Net multi-view heterogeneous graph network, specifically as follows: In the traditional MH-Net multi-view heterogeneous graph network, the traffic unit generates a traffic representation after the multi-view heterogeneous graph is constructed and the heterogeneous relationship is aggregated. On the basis of the traditional multi-view heterogeneous relationship aggregation structure, a session round causal chain reconstruction layer is added, which connects the fragment sending node, fragment receiving node, acknowledgment feedback node and retransmission feedback node in the encrypted transmission heterogeneous graph into a causal chain according to the session round identifier. Based on the original heterogeneous neighborhood aggregation structure, a trusted neighborhood contraction layer for encrypted fragments is introduced to divide adjacent fragment nodes into trusted fragment neighborhoods and fragment neighborhoods to be verified, and to contract the propagation range of the fragment neighborhoods to be verified. A new task boundary state recycling layer is added after the original graph representation output structure to recycle residual fragment states across tasks and generate a task boundary recycling graph. The session round causal chain reconstruction layer, the ciphertext fragment trusted neighborhood shrinkage layer, and the task boundary state recycling layer are connected in the order of causal chain reconstruction, trusted neighborhood shrinkage, and task boundary recycling to obtain the improved MH-Net multi-view heterogeneous graph network. The session round causal chain reconstruction layer receives various relationships in the heterogeneous graph of encrypted transmission, arranges the segment sending nodes, segment receiving nodes, acknowledgment feedback nodes, and retransmission feedback nodes according to the session round identifier, establishes sending-receiving relationships, receiving-acknowledgment relationships, verification failure-retransmission relationships, and acknowledgment feedback-next round sending relationships, and generates session round causal chains and segment causal state representations, where: The conversation turn causal chain reconstruction layer includes: Relationship input buffer queue: Stores various relationships in the heterogeneous graph of encrypted transmission, and groups them according to communication data packet identifier and session round identifier; Round Index Register: Records the session round identifier, round start segment number, round end segment number, and next round start segment number; Node sorting buffer table: Segment sending nodes, segment receiving nodes, acknowledgment feedback nodes, and retransmission feedback nodes are arranged in session round and time order; Feedback status matching table: records the status of verification passed, verification failed, missing, duplicate, and retransmission, and matches the corresponding fragment nodes; Causal edge generator: generates send-receive relationship, receive-acknowledgement relationship, verification failure-retransmission relationship, and acknowledgment feedback-next round send relationship; Causal chain caching unit: caches and concatenates the causal relationship chains of each session round; Fragment Causal State Output Unit: Generates a fragment causal state representation based on the position of the fragment node in the causal chain; In the session round causal chain reconstruction layer, the relation input buffer queue groups various relations in the encrypted transmission heterogeneous graph according to the communication data packet identifier and the session round identifier and then passes them to the node sorting buffer table. The round index register provides the node sorting buffer table with the round start and end segment number and the next round start segment number, forming the node sequence within the round. The feedback state matching table matches the feedback state with the segment node and sends the matching result to the causal edge generator. The causal edge generator generates various causal relations and writes them into the causal chain buffer unit. The causal chain buffer unit splices the causal relation chain of each session round. The segment causal state output unit generates the segment causal state representation based on the spliced ​​session round causal chain. The fragment sending node, fragment receiving node, acknowledgment feedback node, and retransmission feedback node are arranged according to the session round identifier, specifically as follows: The fragment sending nodes, fragment receiving nodes, acknowledgment feedback nodes, and retransmission feedback nodes in the encrypted transmission heterogeneous graph are grouped according to the session round identifier to generate session round node groups; Within each session round node group, a node sorting field is established based on the fragment sequence number, sending timestamp, receiving timestamp, and confirmation feedback order; The nodes are sorted in ascending order by sorting the sorting field. When two nodes have the same time field or the same confirmation feedback order, they are sorted in the order of segment sending node, segment receiving node, confirmation feedback node, and retransmission feedback node. The sorted fragment sending nodes, fragment receiving nodes, acknowledgment feedback nodes, and retransmission feedback nodes are combined into a node sequence within a round. Establish send-receive relationship, receive-acknowledgement relationship, verification failure-retransmission relationship, and acknowledgement feedback-next round send relationship, specifically as follows: Within the node sequence of a round, connect the segment sending nodes and segment receiving nodes that have the same segment sequence number, the same session round identifier, and whose sender identity identifier and receiver identity identifier correspond to each other to generate a send-receive relationship; Connect the receiving nodes and acknowledgment feedback nodes that have the same segment sequence number, the same session round identifier, and the corresponding acknowledgment feedback order to generate a receive-acknowledgment relationship; When the feedback status in the confirmation feedback node is marked as verification failure or missing, the corresponding segment receiving node, confirmation feedback node and retransmission feedback node are connected and classified into the verification failure-retransmission relationship. Connect the acknowledgment feedback node of the current session round to the starting fragment sending node of the next session round to generate an acknowledgment feedback-next round sending relationship; The send-receive relationship, receive-acknowledgement relationship, verification failure-retransmission relationship, and acknowledgement feedback-next round send relationship are combined according to the session round identifier to generate a session round causal chain; The generated fragment causal state representation is as follows: For each segment node in the causal chain of the session round, extract the segment sequence number, session round identifier, node type identifier, predecessor node type identifier, successor node type identifier, and feedback status identifier. The node type identifier can be the segment sending node identifier, segment receiving node identifier, acknowledgment feedback node identifier, or retransmission feedback node identifier. A fragment position code is generated based on the position of the fragment node in the causal chain of the session round. The fragment position code is the ratio of the sorting number of the fragment node in the causal chain of the corresponding session round to the total number of nodes in the corresponding session round. The fragment sequence number, session round identifier, node type identifier, predecessor node type identifier, successor node type identifier, feedback status identifier, and fragment position code are concatenated in a fixed order to generate a fragment causal state vector. Arrange the fragment causal state vectors corresponding to each fragment node according to the session round identifier and fragment sequence number to generate a fragment causal state representation; The trusted neighborhood shrinking layer for ciphertext fragments, based on session round-based causal chains and fragment causal state representations, divides adjacent fragment nodes in the heterogeneous graph of ciphertext transmission into neighborhoods. Adjacent fragment nodes with consistent session rounds, consistent link identifiers, and matching feedback states are assigned to the trusted fragment neighborhood. Adjacent fragment nodes with inconsistent session rounds, missing feedback states, or mismatched link identifiers are assigned to the neighborhood of fragments awaiting verification. The layer then shrinks the propagation range of the neighborhood of fragments awaiting verification, generating a trusted neighborhood fragment graph, where: The trusted neighborhood shrinking layer for ciphertext fragments includes: Fragment Neighborhood Cache Queue: Stores the set of neighboring fragment nodes of the current fragment node; Session consistency table: records the session round matching results of adjacent segment nodes; Link consistency determination table: records the link identifier matching results of adjacent segment nodes; Feedback State Matching Register: Records the feedback state matching results and feedback missing states of adjacent segment nodes; Trusted Neighborhood Labeling Table: Labels adjacent segment nodes that meet the requirements of consistent session rounds, consistent link identifiers, and matching feedback states; Neighborhood tagging table to be verified: tags adjacent segment nodes with inconsistent session rounds, missing feedback status, or mismatched link identifiers; Neighborhood Shrinkage Controller: Preserves propagation connections in the neighborhood of trusted fragments and handles cross-round propagation connections, feedback missing propagation connections, and link mismatch propagation connections in the neighborhood of fragments to be verified; Fragment graph output buffer unit: buffers fragment nodes and propagation connections after neighborhood shrinkage, and outputs a reliable neighborhood fragment graph; In the trusted neighborhood shrinking layer of the encrypted fragment, the fragment neighborhood cache queue passes the set of adjacent fragment nodes to the session consistency discrimination table, the link consistency discrimination table, and the feedback state matching register, which respectively generate session round matching results, link identifier matching results, and feedback state matching results. The trusted neighborhood label table and the neighboring neighborhood label table generate trusted fragment neighborhood labels and neighboring fragment labels to be verified according to the matching results, and pass the labeling results to the neighborhood shrinking controller. The neighborhood shrinking controller adjusts the propagation connection between fragment nodes according to the labeling results, and passes the fragment nodes and propagation connection after neighborhood shrinking to the fragment graph output cache unit. The fragment graph output cache unit outputs the trusted neighborhood fragment graph. The neighborhood partitioning of adjacent segment nodes in the heterogeneous graph of encrypted transmission is performed as follows: Using the current fragment node as the center node, based on the fragment association edge, session round edge, link transmission edge and feedback status edge in the encrypted transmission heterogeneous graph, extract the adjacent fragment nodes that have a connection relationship with the current fragment node and generate a candidate neighborhood fragment set. For each neighboring segment node in the candidate neighborhood segment set, compare the session round identifier, link identifier, and feedback status identifier of the neighboring segment node with the current segment node; When the session round identifier, link identifier and feedback status identifier of the adjacent fragment node are consistent with those of the current fragment node, the adjacent fragment node is included in the trusted fragment neighborhood. When the session round identifier of the adjacent fragment node is inconsistent with that of the current fragment node, the link identifier does not match, or the feedback status identifier is missing, the adjacent fragment node will be included in the neighborhood of the fragment to be verified. Associate the current fragment node, the trusted fragment neighborhood, and the fragment neighborhood to be verified to generate a fragment neighborhood partitioning record; Shrink the propagation range of the neighborhood of the fragment to be verified, specifically as follows: Based on the segment neighborhood partitioning record, extract the node of the segment to be verified and the corresponding propagation connection in the neighborhood of the segment to be verified. The propagation connection includes cross-round propagation connection, feedback missing propagation connection and link mismatch propagation connection. Perform truncation processing on cross-round propagation connections, and delete the propagation connections between the unverified fragment node with inconsistent session round identifiers and the current fragment node; Isolate the propagation connection for missing feedback, keep the node of the segment to be verified with missing feedback status in the neighborhood of the segment to be verified, and stop the segment to be verified node from propagating the segment status to the neighborhood of trusted segments. Perform shrinkage processing on the link mismatch propagation connection, delete the link transmission edge between the link identification mismatched segment node and the current segment node, and retain the feedback state edge corresponding to the segment node to be verified. The trusted fragment nodes in the neighborhood, the preserved trusted propagation connections, and the fragment nodes to be verified that have been isolated and marked are combined to generate a trusted neighborhood fragment graph. The task boundary state recycling layer, based on the trusted neighborhood fragment graph, session round causal chain, and feedback state relationship, marks the start fragment, end fragment, retransmission termination fragment, and cross-task residual fragment of the current communication task. It performs recycling processing on the cross-task residual fragment state, generates a task boundary recycling graph, and generates a ciphertext fragment transmission state packet based on the task boundary recycling graph, wherein: The task boundary state recycling layer includes: Task boundary cache queue: stores fragment nodes, session round causal chains, and feedback state relationships in the trusted neighborhood fragment graph; Boundary Identifier Register: Records the task start identifier, task end identifier, start segment number, and end segment number of the current communication task; Fragment Attribution Table: Records the communication data packet identifier, session round identifier, and task attribution identifier corresponding to the fragment node; Retransmission Termination Status Table: Records the number of retransmissions of a retransmission segment, the retransmission feedback status, and the retransmission termination identifier; Residual Fragment Screening Table: Marks cross-task residual fragments whose session rounds do not belong to the current communication task, whose arrival time is later than the task end fragment, or whose task ownership identifier does not match; State reclamation controller: Performs state reclamation on cross-task residual fragments and removes cross-task residual fragments from the valid fragment states of the current communication task; Task boundary graph caching unit: caches fragment nodes, task boundary relationships, and feedback state relationships after the completion of state reclamation, and outputs a task boundary reclamation graph; Transmission status packet output unit: Generates encrypted fragment transmission status packets based on the task boundary reclamation graph; In the task boundary state reclamation layer, the task boundary cache queue passes the trusted neighborhood fragment graph, session round causal chain, and feedback state relationship to the boundary identifier register, fragment ownership discrimination table, and retransmission termination state table. The boundary identifier register outputs the boundary identifier of the current communication task, the fragment ownership discrimination table outputs the fragment task ownership result, and the retransmission termination state table outputs the retransmission termination state. The residual fragment screening table generates cross-task residual fragment tags based on the boundary identifier, fragment task ownership result, and retransmission termination state, and passes the cross-task residual fragment tags to the state reclamation controller. After completing the cross-task residual fragment state reclamation, the state reclamation controller passes the updated fragment nodes and task boundary relationships to the task boundary graph cache unit. The task boundary graph cache unit outputs the task boundary reclamation graph, and the transmission state packet output unit generates encrypted fragment transmission state packets based on the task boundary reclamation graph. The start segment, end segment, retransmission termination segment, and cross-task residual segment are marked as follows: Based on the task start identifier, locate the first segment node corresponding to the current communication task in the trusted neighborhood segment graph, and mark the first segment node as the start segment; Based on the task end identifier, locate the end segment node corresponding to the current communication task in the trusted neighborhood segment graph, and mark the end segment node as the end segment; Based on the verification failure-retransmission relationship and feedback status relationship in the causal chain of the session rounds, the number of retransmissions and retransmission feedback status corresponding to the same communication data segment are counted. When the retransmission feedback status is confirmed or the number of retransmissions reaches the retransmission threshold of 3 times, the corresponding communication data segment is marked as a retransmission termination segment. The fragment nodes in the trusted neighborhood fragment graph are classified according to the communication data packet identifier, session round identifier, and task ownership identifier. When the task ownership identifier of a fragment node is inconsistent with the current communication task, or when the arrival time of a fragment node is later than the receiving timestamp of the end fragment, the corresponding fragment node is marked as a cross-task residual fragment. The process of reclaiming residual state fragments across tasks is as follows: Remove cross-task residual fragments from the valid fragment set of the current communication task and write the cross-task residual fragments to the residual fragment record; Retain the communication data packet identifier, session round identifier, fragment sequence number, arrival time, and corresponding task ownership identifier for the cross-task residual fragments; Disconnect the cross-task residual fragment from the fragment association edge, session round edge, and link transmission edge in the current communication task, while retaining the feedback state edge corresponding to the cross-task residual fragment; A task boundary reclamation map is generated based on the set of valid fragments after removal and disconnection, task boundary relationships, and feedback status relationships. Based on the valid fragment status, retransmission termination fragment status, and residual fragment records in the task boundary recycling graph, generate a ciphertext fragment transmission status packet. The improved MH-Net multi-view heterogeneous graph network was trained using a combination of session round causal chain reconstruction error, trusted neighborhood partitioning error, task boundary state retrieval error, and ciphertext fragment transmission state prediction error as the joint optimization objective. The network parameters of each layer were continuously optimized. When the change in the joint loss value over five consecutive training rounds was less than 0.001, the improved MH-Net multi-view heterogeneous graph network was considered to have completed convergence training. The improved MH-Net multi-view heterogeneous graph network is trained as follows: The encrypted transmission heterogeneous graph, session round real chain, trusted neighborhood real label, task boundary real label, and encrypted fragment transmission state label are extracted from the training sample set. The training sample set includes 12,000 communication session samples. Each communication session sample contains 64 to 256 fragment nodes, 2 to 8 session rounds, 1 to 4 link nodes, and 4 to 6 types of feedback state nodes. The encrypted transmission heterogeneous graph is input into the improved MH-Net multi-view heterogeneous graph network. The session round causal chain reconstruction layer outputs the predicted session round causal chain and fragment causal state representation. The encrypted fragment trusted neighborhood shrinking layer outputs the predicted trusted neighborhood fragment graph. The task boundary state recycling layer outputs the predicted task boundary recycling graph and the encrypted fragment transmission state prediction result. The session round causal chain reconstruction error is obtained by subtracting the number of causal edges in the predicted session round causal chain from the number of real causal edges in the actual session round chain, squaring the differences, and taking the average. The trusted neighborhood segmentation error is obtained by subtracting the trusted segment label value and the trusted neighborhood real label value in the predicted trusted neighborhood segment graph segment by segment, squaring the differences, and taking the average. The task boundary state recovery error is obtained by subtracting the start segment label, end segment label, retransmission termination segment label, and cross-task residual segment label in the predicted task boundary recovery graph from the actual task boundary label, squaring the differences, and taking the average. The encrypted segment transmission state prediction result is obtained by subtracting the encrypted segment transmission state label, squaring the state differences, and taking the average. The session round causal chain reconstruction error is multiplied by 0.30, the trusted neighborhood partitioning error by 0.25, the task boundary state retrieval error by 0.20, and the ciphertext fragment transmission state prediction error by 0.25, and then summed to generate a joint loss value. The rate of change of the joint loss value relative to each trainable parameter in the session round causal chain reconstruction layer, the ciphertext fragment trusted neighborhood shrinkage layer, and the task boundary state retrieval layer is calculated to obtain the gradient value. The gradient value is multiplied by the learning rate of 0.001 and then subtracted from the current value of the corresponding trainable parameter to obtain the updated network parameters. The updated network parameters are written into the corresponding layer. The input, error calculation, joint loss value calculation, and parameter update are repeated until the change of the joint loss value in 5 consecutive training rounds is less than 0.001. This indicates that the improved MH-Net multi-view heterogeneous graph network has completed convergence training.

[0023] In this embodiment, the generation of the ciphertext fragment scheduling packet includes: The encrypted segment transmission status representation in the encrypted segment transmission status packet is organized according to the session round identifier, communication data packet identifier and segment sequence number to generate a scheduling candidate segment sequence; Based on the causal state of the segments, trusted neighborhood identifiers, task boundary reclamation identifiers, and link transmission states in the scheduling candidate segment sequence, the scheduling state of each communication data segment is determined. The scheduling states include transmission state, delay state, retransmission state, and stripping state, where: The scheduling status of each communication data segment is determined as follows: Extract the causal state, trusted neighborhood identifier, task boundary reclamation identifier, and link transmission state for each communication data segment in the scheduling candidate segment sequence; When the task boundary recycling identifier of a communication data segment is a cross-task residual identifier, the corresponding communication data segment is determined to be in a stripped state. When the feedback status of a communication data segment is a verification failure or a missing status, and the task boundary recycling identifier is not a cross-task residual identifier, the corresponding communication data segment will be determined as a retransmission status. When the causal state of a communication data segment is the unacknowledged preceding segment state, or the link transmission state is the link switching state, the corresponding communication data segment will be determined as a delayed state. When the trusted neighborhood identifier of a communication data segment is a trusted segment neighborhood identifier, the segment causal state is the previous segment confirmed state, the link transmission state is the link available state, and the feedback state is not the verification failure state or the missing state, the corresponding communication data segment is determined to be in the sending state. Associate each communication data segment with its corresponding scheduling status to generate a segment scheduling status record; Communication data segments in the stripped scheduling state are removed from the scheduling candidate segment sequence. The remaining communication data segments are sorted according to session round identifier, segment causal order, and link transmission state to generate a sequence of segments to be encrypted, wherein: Generate the sequence of fragments to be encrypted, specifically as follows: Remove communication data segments with a scheduling status of stripping from the scheduling candidate segment sequence, and retain communication data segments with scheduling status of sending, delayed, and retransmission. The retained communication data segments are grouped according to the session round identifier to generate session round segment groups; Within each session round fragment group, the communication data fragments are first sorted in ascending order according to the causal order of the fragments, and then the communication data fragments in the retransmission scheduling state are inserted after the corresponding original sent fragments. When two communication data segments have the same causal order, the communication data segments with the link transmission status of "link available" are arranged first, and then the communication data segments with the link transmission status of "link switching" are arranged. The fragment groups of each session round are concatenated in ascending order of session round identifier to generate a sequence of fragments to be encrypted; Based on the causal order of the fragments in the sequence of fragments to be encrypted, the link transmission status, and the transmission time window, a fragment transmission order table is generated, and the transmission round identifier and transmission time window identifier are configured. Based on the fragment sequence number, transmission round identifier, transmission time window identifier, and receiver reassembly position, an out-of-order mapping table and an index association table are generated, where: Generate an out-of-order mapping table and an indexed association table, specifically as follows: Based on the original fragment sequence number in the fragment sequence to be encrypted and the actual transmission sequence number in the fragment transmission order table, establish the correspondence between the original fragment sequence number and the actual transmission sequence number, and generate an out-of-order mapping record; Arrange the out-of-order mapping records in ascending order of the actual sent sequence number to generate an out-of-order mapping table; Based on the segment sequence number, transmission round identifier, transmission time window identifier, and receiver reassembly location, establish an index-related record for each communication data segment; The index-associated records include the fragment sequence number, the actual transmission sequence number, the transmission round identifier, the transmission time window identifier, and the receiver's reassembly position; Arrange the index-related records in ascending order of their reorganization positions at the receiving end to generate an index-related table; The sequence of fragments to be encrypted, the fragment sending order table, the out-of-order mapping table, and the index association table are associated to generate a ciphertext fragment scheduling packet.

[0024] In this embodiment, generating the key to encapsulate the encryption packet includes: The receiving end generates an ML-KEM encapsulation public key and an ML-KEM decapsulation private key, and sends the ML-KEM encapsulation public key to the sending end; The sending end performs key encapsulation based on the ML-KEM encapsulated public key, generates a session shared key and key encapsulation data, and associates them with the ciphertext fragment scheduling packet to establish a session key derivation extension record for the ML-KEM algorithm, wherein: The session key derivation extension record for the ML-KEM algorithm is established as follows: The sending end performs key encapsulation based on ML-KEM encapsulation of the public key, generating a session shared key and key encapsulation data; Configure a scheduling packet identifier for the ciphertext segment scheduling packet, and extract the identifier of the segment sequence to be encrypted, the segment transmission order table identifier, the out-of-order mapping table identifier, and the index association table identifier from the ciphertext segment scheduling packet; Associate the session shared key, key encapsulation data, scheduling packet identifier, segment sequence identifier to be encrypted, segment transmission order table identifier, out-of-order mapping table identifier, and index association table identifier to generate a session key derivation extension record; Write the session key derivation extension record into the key derivation extension cache of the ML-KEM algorithm; Based on the sequence of fragments to be encrypted, the fragment transmission order table, the out-of-order mapping table, and the index association table in the encrypted fragment scheduling packet, generate a fragment path record including session round identifier, fragment sequence number, field marker, link identifier, transmission time window identifier, and index association identifier; In the session key derivation extension record, fragment path key derivation processing is performed on the session shared key and fragment path record to generate an initial fragment key sequence, where: Perform fragment path key derivation processing, specifically as follows: Extract the session shared key and scheduling packet identifier from the session key derivation extension record, and call the fragment path record generated based on the ciphertext fragment scheduling packet to process each fragment path record in the order of the fragment sending sequence table; The session round identifier, fragment sequence number, field marker, link identifier, sending time window identifier, and index association identifier in each fragment path record are converted into fixed-length byte strings, where the session round identifier occupies 2 bytes, the fragment sequence number occupies 4 bytes, the field marker occupies 1 byte, the link identifier occupies 2 bytes, the sending time window identifier occupies 4 bytes, and the index association identifier occupies 4 bytes. The fragment path string is generated by concatenating the fixed-length byte strings in the order of session round identifier, fragment sequence number, field marker, link identifier, sending time window identifier, and index association identifier. Perform key derivation operation based on SHA3-256 hash function on the fragment key derivation input, input the fragment key derivation input into SHA3-256 hash function to obtain digest result, and perform iterative digest concatenation when the digest result length is less than 256 bits to generate a derived output byte string with a length of not less than 256 bits; Extract 256 bits from the beginning of the derived output byte string as the initial segment key for the current segment, and establish a correspondence between the initial segment key and the segment sequence number; Combine the initial fragment keys according to the fragment arrangement order in the fragment transmission order list to generate an initial fragment key sequence, and write the initial fragment key sequence into the session key derivation extension record; In the session key derivation extension record, based on the fragment transmission order table and the out-of-order mapping table, out-of-order fragment key backfilling processing is performed on the initial fragment key sequence to generate a backfilled fragment key sequence, wherein: Perform out-of-order fragment key backfilling processing, specifically as follows: Extract the fragment transmission order table identifier, out-of-order mapping table identifier, and initial fragment key sequence from the session key derivation extension record, and call the corresponding fragment transmission order table and out-of-order mapping table; Extract the original fragment sequence number and the actual transmission sequence number from the disordered mapping table one by one, and establish the key backfill position relationship from the original fragment sequence number to the actual transmission sequence number; Extract the corresponding initial fragment key from the initial fragment key sequence according to the original fragment sequence number, and determine the writing position of the corresponding initial fragment key in the backfill fragment key sequence according to the actual sending sequence number; When the position corresponding to the actual sent sequence number is empty, the initial fragment key is written to the corresponding position. When the position corresponding to the actual sent sequence number already has a fragment key, the fragment sequence number is compared according to the index association identifier, and the fragment key that is consistent with the out-of-order mapping table is retained. For retransmitted segments in the segment transmission sequence list, the backfill position is regenerated based on the original segment sequence number, transmission round identifier, and transmission time window identifier corresponding to the retransmitted segment, and the initial segment key corresponding to the retransmitted segment is written into the retransmission backfill position. Arrange the written fragment keys in ascending order of the actual sent sequence number to generate a backfill fragment key sequence; The key-encapsulated data, the sequence of fragments to be encrypted, the fragment sending order table, the fragment path record, the backfilled fragment key sequence, and the index association table are bound together to generate a key-encapsulated encryption package.

[0025] In this embodiment, generating the encrypted transmission packet includes: Based on the sequence of fragments to be encrypted, the fragment transmission order table, the backfilled fragment key sequence, and the index association table in the key-encapsulated encryption packet, determine the correspondence between the fragment encryption keys and the fragment transmission order; According to the segment encryption key correspondence, each communication data segment in the segment sequence to be encrypted is matched with the corresponding backfilled segment key to generate a segment encryption task table, where: Each communication data segment is matched with its corresponding backfill segment key, specifically as follows: Extract the segment number, transmission round identifier, transmission time window identifier, and index association identifier of the communication data segment from the sequence of segments to be encrypted; The actual transmission sequence number corresponding to the communication data segment is determined based on the index association table, and the backfill segment key corresponding to the actual transmission sequence number is located in the backfill segment key sequence; The communication data segment, the actual transmission sequence number, the backfill segment key, and the index association identifier are matched to generate a single segment encryption task record; When the same actual transmission sequence number corresponds to two or more communication data segments, the communication data segments that match are retained according to the reassembly position of the receiving end in the index association table. Arrange the individual segment encryption task records in ascending order of actual transmission sequence number to generate a segment encryption task table; Each communication data segment is encrypted segment by segment according to the segment encryption task table, generating a set of ciphertext segments, specifically: Extract single-fragment encryption task records one by one from the fragment encryption task table. Each single-fragment encryption task record includes the communication data fragment, fragment sequence number, actual sent sequence number, backfill fragment key, and index association identifier. The actual sent sequence number and index association identifier are converted into a fragment encryption auxiliary field, and the fragment encryption auxiliary field is associated with the communication data fragment; Using the backfilled fragment key as the fragment encryption key for the current communication data fragment, the fragment encryption auxiliary field and the communication data fragment are input together into the encryption operation process to generate a ciphertext fragment; Associate the ciphertext fragment, fragment sequence number, actual transmission sequence number, fragment encryption auxiliary field, and index association identifier to generate a single-fragment ciphertext record; Arrange the individual ciphertext records in ascending order of their actual transmission sequence number to generate a set of ciphertext fragments; An integrity identifier is generated based on the ciphertext fragments, fragment sequence numbers, transmission time window identifiers, and index association identifiers in the ciphertext fragment set. This integrity identifier is then bound to the corresponding ciphertext fragment, generating a bound ciphertext fragment set. Generate an integrity identifier and bind it to the corresponding ciphertext fragment, specifically as follows: Extract individual ciphertext records one by one from the ciphertext fragment set. Each individual ciphertext record includes the ciphertext fragment, fragment sequence number, actual transmission sequence number, fragment encryption auxiliary field, and index association identifier. Determine the transmission time window identifier corresponding to the ciphertext segment based on the segment transmission sequence table, and concatenate the ciphertext segment, segment number, transmission time window identifier and index association identifier in a fixed order to generate integrity verification input; Perform a digest calculation based on the SHA3-256 hash function on the integrity verification input to generate a 256-bit integrity digest value, and extract 128 bits from the beginning of the integrity digest value as an integrity identifier; Write the integrity identifier into the corresponding single-segment ciphertext record and establish the correspondence between the integrity identifier, the segment sequence number, and the actual transmission sequence number; Arrange the single-fragment ciphertext records after writing the integrity identifier in ascending order of the actual sending sequence number to generate a set of bound ciphertext fragments. The encrypted fragment set is sent in the order of fragment transmission, and the encrypted transmission packet is generated by associating the key-encapsulated data, fragment path record, index association table and transmission record.

[0026] In this embodiment, generating recovery communication data and transmission feedback records and performing feedback updates includes: The receiving end parses the encrypted transmission packet to obtain the bound encrypted fragment set, key encapsulation data, fragment path record, index association table and sending record; The receiving end decapsulates the key-encapsulated data based on the ML-KEM decapsulation private key, recovers the session shared key, and generates a fragment decryption key sequence based on the session shared key, fragment path record, index association table, and sending record, wherein: Generate a fragment decryption key sequence, specifically: The receiving end decapsulates the key-encapsulated data based on the ML-KEM decapsulation private key, restores the session shared key, and extracts the scheduling packet identifier from the key-encapsulated encrypted packet associated with the key-encapsulated data; Based on the session round identifier, fragment sequence number, field marker, link identifier, sending time window identifier, and index association identifier in the fragment path record, generate the fragment path string according to the same fixed-length byte conversion rules as the fragment path key derivation processing at the sending end; Concatenate the session shared key, the scheduling packet identifier, and the fragment path string in sequence to generate the fragment decryption key derived input; Perform a key derivation operation based on the SHA3-256 hash function on the derivation input of the fragment decryption key to generate a 256-bit fragment decryption key; Based on the index association table and the sending record, determine the actual receiving location corresponding to the segment decryption key, arrange the segment decryption keys according to the actual receiving location, and generate a segment decryption key sequence. Integrity verification is performed based on the integrity identifier in the bound ciphertext fragment set. Ciphertext fragments that pass verification are written to the decryption fragment set, while ciphertext fragments that fail verification are written to the transmission feedback record. Perform integrity verification, specifically as follows: Extract the bound ciphertext fragments one by one from the set of bound ciphertext fragments. Each bound ciphertext fragment includes a ciphertext fragment, a fragment sequence number, a sending time window identifier, an index association identifier, and an integrity identifier. The encrypted fragment, fragment number, transmission time window identifier, and index association identifier are concatenated in a fixed order to generate the integrity verification input for the receiving end. Perform a digest calculation based on the SHA3-256 hash function on the integrity verification input of the receiving end to generate a 256-bit integrity digest value of the receiving end, and extract 128 bits from the beginning of the integrity digest value of the receiving end as the integrity identifier of the receiving end; The integrity identifier of the receiving end is compared with the integrity identifier in the bound ciphertext segment. When the two match, the corresponding ciphertext segment is written into the set of segments to be decrypted. When the two are inconsistent, the segment number, sending time window identifier, index association identifier and verification failure status of the corresponding ciphertext segment are written into the transmission feedback record. The decryption process proceeds segment by segment according to the decryption key sequence. Then, based on the index association table and transmission records, the decrypted communication data segments are sorted and reassembled to generate the recovered communication data. Perform segment-by-segment decryption and segment reconstruction, specifically as follows: Extract the ciphertext fragments, fragment numbers, and index association identifiers one by one from the set of fragments to be decrypted, and determine the actual reception location of the corresponding ciphertext fragments based on the sending records and the index association table; Based on the actual receiving location, the corresponding segment decryption key is extracted from the segment decryption key sequence, and the corresponding segment decryption key is used to perform decryption operation on the ciphertext segment to generate a decrypted communication data segment. Determine the receiving end reassembly position corresponding to the decrypted communication data segment based on the index association table, and write the decrypted communication data segment into the corresponding reassembly position. When two or more decrypted communication data segments correspond to the same reassembly position at the receiving end, the decrypted communication data segment that passes the verification and matches the actual sending sequence number in the sending record is retained according to the actual sending sequence number and integrity verification result in the sending record. The decrypted communication data segments in each reassembly position are spliced ​​together in ascending order of the reassembly position at the receiving end to generate the recovered communication data. Based on the verification failure records, missing fragment records, retransmission records, and reassembly status records in the transmission feedback records, update the ciphertext transmission heterogeneous graph, improve the MH-Net multi-view heterogeneous graph network parameters, and the key derivation parameters of the ML-KEM algorithm.

[0027] Example 1: To verify the feasibility of this invention in practice, it was applied to an encrypted communication scenario between an edge terminal, an edge forwarding node, and a central service node. This method receives a batch of simulated communication data, containing 4800 rounds of session records, including 3200 training rounds, 800 verification rounds, and 800 test rounds. In each round, the edge terminal sends a service payload to the central service node and receives confirmation feedback from the central service node. The original communication data length ranges from 96KB to 1536KB, with an average length of 684KB. Transmission locations include the edge terminal side, the edge forwarding node side, and the central service node side. The links include a primary transmission link and a backup transmission link. The average round-trip time of the primary transmission link is 41ms, and the average round-trip time of the backup transmission link is 58ms. The test samples include out-of-order arrival, duplicate fragments, verification failures, delayed confirmations, and cross-task residual fragments. The out-of-order arrival rate is 14.6%, the duplicate fragment rate is 3.8%, the verification failure rate is 2.7%, and the cross-task residual fragment rate is 1.9%.

[0028] Using one round of sessions as a specific simulation sample, the original communication data to be transmitted was 786,432 bytes, including an identity field of 3,072 bytes, a service field of 24,576 bytes, a payload field of 737,280 bytes, a time field of 1,024 bytes, and a checksum field of 480 bytes. This method performs format standardization, invalid record removal, duplicate record deletion, and timestamp alignment on the basic data of this round of communication. Three invalid records and two duplicate records were removed, reducing the average timestamp deviation between the sender's record, receiver's record, and feedback record from 6.4 ms to 1.1 ms. Subsequently, field parsing is performed on the communication data to be transmitted, generating identity field markers, service field markers, payload field markers, time field markers, and checksum field markers. The data is then divided into 192 communication data segments according to the field markers, data length, and session round record. Each segment is 4,096 bytes long, and each segment is configured with a communication data packet identifier, segment sequence number, field marker, session round identifier, link identifier, and feedback status identifier, forming a standardized communication segment dataset.

[0029] Based on this standardized communication fragment dataset, this method extracts fragment association relationships, session round relationships, link transmission relationships, and feedback status relationships, and maps them into a heterogeneous graph for encrypted transmission. Fragment numbers 001 to 006 correspond to identity and service fields, fragment numbers 007 to 186 correspond to payload fields, and fragment numbers 187 to 192 correspond to time and verification fields. This heterogeneous graph for encrypted transmission contains 192 fragment nodes, 3 session round nodes, 2 link nodes, and 5 types of feedback status nodes, forming 191 fragment association edges, 18 session round edges, and 216 link and feedback status edges, for a total of 425 edges. During simulated transmission, fragment 057 experienced an arrival delay of 136ms on the main transmission link, fragment 064 arrived repeatedly, fragment 119 failed integrity verification, and one residual fragment from the previous communication task entered the current receiving window.

[0030] After the ciphertext transmission heterogeneous graph enters the improved MH-Net multi-view heterogeneous graph network, this method performs multi-view relation encoding on the ciphertext transmission heterogeneous graph. The session round causal chain reconstruction layer establishes 3 main causal chains and 6 retransmission feedback branches based on fragment transmission, fragment reception, acknowledgment feedback, and retransmission feedback. Fragment 057 is marked as delayed arrival, fragment 064 as a duplicate fragment, and fragment 119 as a failed verification pending retransmission. The ciphertext fragment trusted neighborhood shrinking layer divides the neighborhoods of adjacent fragment nodes. Out of 192 fragment nodes, 188 fragments are assigned to the trusted fragment neighborhood, and 4 fragments are assigned to the unverified fragment neighborhood, reducing the average number of propagation adjacent nodes of the unverified fragment from 7 to 2. The task boundary state reclamation layer identifies one cross-task residual fragment and performs state reclamation on this fragment, preventing it from entering the effective scheduling range of the current task, and finally outputs the ciphertext fragment transmission state packet.

[0031] During the training phase, the improved MH-Net multi-view heterogeneous graph network was trained using 3200 training session samples. The initial joint loss was 0.692, decreasing to 0.184 by the 20th session, 0.061 by the 40th session, and then decreasing to 0.043, 0.042, 0.041, 0.041, and 0.040 by the 52nd to 56th sessions, respectively, with the change in each of the five sessions being less than 0.001. In the validation samples, the accuracy rates for identifying out-of-order segments were 98.1%, for identifying duplicate segments 97.4%, for identifying cross-task residual segments 96.8%, and for identifying verification failure and retransmission status 98.6%. For this round of simulated samples, the encrypted segment transmission status packet accurately retained 188 normally transmitted segments, 2 delayed segments, and 1 retransmitted segment, and removed 1 cross-task residual segment.

[0032] This method generates a ciphertext fragment scheduling packet based on the ciphertext fragment transmission status packet. In this session round, after scheduling status determination, 188 of the 192 communication data fragments enter the transmission state, 2 enter the delay state, 1 enter the retransmission state, and 1 enter the stripping state. The remaining 191 valid fragments are sorted according to session round identifier, fragment causal order, and link transmission status to generate a sequence of fragments to be encrypted. Fragment 057 is moved to the second transmission time window, fragment 119 is configured as a retransmission fragment, and a duplicate copy of fragment 064 is not included in the sequence of fragments to be encrypted. The generated out-of-order mapping table contains 191 mapping records, and the index association table contains 191 index records.

[0033] After generating an ML-KEM encapsulation public key and an ML-KEM decapsulation private key, the receiving end sends the ML-KEM encapsulation public key to the sending end. The sending end performs key encapsulation based on the ML-KEM encapsulation public key, generating a session shared key and key encapsulation data, and associates them with the ciphertext fragment scheduling packet to establish a session key derivation extension record. This method generates 191 fragment path records based on the fragment sequence to be encrypted, the fragment transmission order table, the out-of-order mapping table, and the index association table. The fragment path key derivation process generates 191 initial fragment keys based on the session shared key and the fragment path records. The out-of-order fragment key backfilling process generates a backfilled fragment key sequence based on the fragment transmission order table and the out-of-order mapping table. The key position for fragment 057 is backfilled from the original 57th bit to the 84th bit. The retransmission key for fragment 119 is different from the initial transmission key to avoid reusing the original key in retransmitted fragments.

[0034] When generating the encrypted transmission packet, this method matches 191 communication data fragments with their corresponding backfill fragment keys, forming 191 fragment encryption tasks. After fragment-by-fragment encryption, 191 encrypted fragments are generated, and an integrity identifier is generated based on the fragment sequence number, sending time window identifier, and index association identifier. The encrypted fragments with bound integrity identifiers are sent in the fragment sending order. The total transmission time for this round of sessions is 238ms, with an average single fragment transmission interval of 1.24ms. After parsing the encrypted transmission packet, the receiving end decapsulates the key-encapsulated data, recovers the session shared key, and generates a fragment decryption key sequence based on the session shared key, fragment path record, index association table, and sending record. During integrity verification, 190 fragments pass the verification on the first attempt, 1 retransmitted fragment passes the verification after retransmission, and 0 cross-task residual fragments enter the set of fragments to be decrypted. After the receiving end completes fragment-by-fragment decryption, sorting, and fragment reassembly, the recovered communication data size is 786432 bytes, consistent with the original communication data to be transmitted.

[0035] In the same batch of 800 test sessions, this method was compared with the traditional whole-packet session key encryption method. The success rate of fragment reassembly was 93.8% for the traditional method and 99.2% for this method; the key correspondence error rate was 2.8% for the traditional method and 0.3% for this method; the false entry rate of cross-task residual fragments was 3.5% for the traditional method and 0.4% for this method; the proportion of duplicate fragments entering the reassembly candidate set was 5.2% for the traditional method and 0.7% for this method; the missed identification rate of failed verification fragments was 4.7% for the traditional method and 0.6% for this method; the average reception recovery time was 276ms for the traditional method and 188ms for this method; the average recovery delay of out-of-order fragments was 318ms for the traditional method and 226ms for this method. As can be seen from Example 1, this method has a clear data change process in the processes of standardized communication segment generation, ciphertext transmission heterogeneous graph construction, improved MH-Net multi-view heterogeneous graph network state identification, ciphertext segment scheduling, ML-KEM algorithm key encapsulation, segment path key derivation, out-of-order segment key backfilling, segment-by-segment encryption, and receiver verification and reassembly. It can improve the reliability of encrypted transmission of communication data under conditions of out-of-order transmission, link fluctuation, segment retransmission, and cross-task residual interference.

[0036] The above description is only a preferred embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the scope of protection of the present invention.

Claims

1. A communication data transmission method based on data encryption, characterized in that, include: Collect basic communication data, perform preprocessing, field labeling and fragment segmentation on the basic communication data, and generate a standardized communication fragment dataset; Based on the standardized communication segment dataset, segment association, session round relationship, link transmission relationship and feedback status relationship are extracted, and various relationships are mapped into a graph structure to construct a heterogeneous graph for encrypted transmission. An improved MH-Net multi-view heterogeneous graph network is constructed, multi-view relationship encoding is performed on the ciphertext transmission heterogeneous graph, a causal chain between fragment sending, fragment receiving, acknowledgment feedback and retransmission feedback is established based on the session round causal chain reconstruction layer, the propagation neighborhood of the fragment to be verified is shrunk through the ciphertext fragment trusted neighborhood shrinking layer, and the state of the fragments left across tasks is recycled through the task boundary state recycling layer, and the ciphertext fragment transmission state packet is output. The transmission order, transmission time window, and index association status of communication data segments are determined based on the encrypted segment transmission status packet, and an encrypted segment scheduling packet is generated. The key encapsulation is performed based on the ML-KEM algorithm to generate a session shared key and key encapsulation data. Combined with the ciphertext fragment scheduling packet, the key encapsulation encryption packet is generated through fragment path key derivation processing and out-of-order fragment key backfilling processing. The sequence of fragments to be encrypted and the order of fragment transmission are determined based on the key encapsulation encryption packet. Segment-by-segment encryption and integrity identifier binding are performed on the sequence of fragments to be encrypted, and the bound ciphertext fragments are sent in the order of fragment transmission to generate ciphertext transmission packets. The receiving end performs key recovery, verification decryption, and fragment reconstruction based on the encrypted transmission packet, generates recovered communication data and transmission feedback records, updates the encrypted transmission heterogeneous graph, improves the MH-Net multi-view heterogeneous graph network parameters, and the key derivation parameters of the ML-KEM algorithm.

2. The communication data transmission method based on data encryption according to claim 1, characterized in that, The generation of the standardized communication fragment dataset includes: Collect basic communication data, which includes communication data to be transmitted, sender identity identifier, receiver identity identifier, session round record, link status record, and historical feedback record; Standardized communication basic data is generated by performing format unification, invalid record removal, duplicate record deletion, and timestamp alignment on the basic communication data. Perform field parsing on the communication data to be transmitted in the standardized communication basic data, and generate field tags according to the field position and field type. The field tags include identity field tags, business field tags, payload field tags, time field tags, and verification field tags. The communication data to be transmitted is segmented according to the field tags, data length, and session round record. Each communication data segment is configured with a communication data packet identifier, segment number, field tags, session round identifier, link identifier, and feedback status identifier to generate a communication data segment sequence. The communication data fragment sequence is associated with the sender's identity identifier, receiver's identity identifier, link status record, and historical feedback record to generate a standardized communication fragment dataset.

3. The communication data transmission method based on data encryption according to claim 1, characterized in that, The construction of the ciphertext transmission heterogeneous graph includes: The standardized communication segment dataset is organized according to communication data packet identifier, segment sequence number, field label, session round identifier, link identifier, and feedback status identifier to generate a segment index record set; Based on the sequence number continuity of fragments in the fragment index record set, the field tag attribution relationship, and the relationship between fragments within the same communication data packet, a fragment association relationship is generated; The session round relationship is generated based on the session round identifier, sender identity identifier, receiver identity identifier, acknowledgment feedback order, and retransmission identifier in the fragment index record set; Extract the link identifier, sending timestamp, receiving timestamp, and segment arrival status from the link status records associated with the segment index record set, and generate the link transmission relationship. Generate feedback status relationships based on historical feedback records, verification status, missing status, duplicate status, and retransmission status in the fragment index record set; The communication data fragments, session rounds, transmission links, and feedback status are mapped as graph nodes, and various relationships are mapped as graph edges. Field tags, link status records, and feedback status identifiers are written to construct a heterogeneous graph for encrypted transmission.

4. The communication data transmission method based on data encryption according to claim 1, characterized in that, The output ciphertext fragment transmission status packet includes: An improved MH-Net multi-view heterogeneous graph network is constructed, which includes a session round causal chain reconstruction layer, a ciphertext fragment trusted neighborhood shrinking layer, and a task boundary state recycling layer. The session round causal chain reconstruction layer receives various relationships in the heterogeneous graph of encrypted transmission, arranges the segment sending node, segment receiving node, acknowledgment feedback node and retransmission feedback node according to the session round identifier, establishes the sending-receiving relationship, receiving-acknowledgment relationship, verification failure-retransmission relationship and acknowledgment feedback-next round sending relationship, and generates the session round causal chain and segment causal state representation; The trusted neighborhood shrinking layer of ciphertext fragments is based on the causal chain of session rounds and the causal state representation of fragments. It divides the neighboring fragment nodes in the heterogeneous graph of ciphertext transmission into neighborhoods. The neighboring fragment nodes with consistent session rounds, consistent link identifiers, and matching feedback states are included in the trusted fragment neighborhood. The neighboring fragment nodes with inconsistent session rounds, missing feedback states, or mismatched link identifiers are included in the neighborhood of fragments to be verified. The propagation range of the neighborhood of fragments to be verified is shrunk to generate a trusted neighborhood fragment graph. The task boundary state recycling layer, based on the trusted neighborhood fragment graph, session round causal chain, and feedback state relationship, marks the start fragment, end fragment, retransmission termination fragment, and cross-task residual fragment of the current communication task. It performs recycling processing on the state of cross-task residual fragments, generates a task boundary recycling graph, and generates ciphertext fragment transmission state packets based on the task boundary recycling graph. The improved MH-Net multi-view heterogeneous graph network was trained using a combination of session round causal chain reconstruction error, trusted neighborhood partitioning error, task boundary state recovery error, and ciphertext fragment transmission state prediction error as the joint optimization objective. The network parameters of each layer were continuously optimized. When the change of the joint loss value in 5 consecutive training rounds was less than 0.001, the improved MH-Net multi-view heterogeneous graph network was considered to have completed convergence training.

5. The communication data transmission method based on data encryption according to claim 1, characterized in that, The generated ciphertext fragment scheduling packet includes: The encrypted segment transmission status representation in the encrypted segment transmission status packet is organized according to the session round identifier, communication data packet identifier and segment sequence number to generate a scheduling candidate segment sequence; Based on the causal state of the segments, trusted neighborhood identifier, task boundary reclamation identifier, and link transmission state in the scheduling candidate segment sequence, the scheduling state of each communication data segment is determined. The scheduling state includes the sending state, the delayed state, the retransmission state, and the stripping state. The communication data segments in the stripped state are removed from the scheduling candidate segment sequence. The remaining communication data segments are sorted according to the session round identifier, segment causal order and link transmission state to generate a segment sequence to be encrypted. Based on the causal order of the fragments in the sequence of fragments to be encrypted, the link transmission status, and the transmission time window, a fragment transmission order table is generated, and the transmission round identifier and transmission time window identifier are configured. Based on the fragment sequence number, transmission round identifier, transmission time window identifier, and receiver reassembly position, generate an out-of-order mapping table and an index association table; The sequence of fragments to be encrypted, the fragment sending order table, the out-of-order mapping table, and the index association table are associated to generate a ciphertext fragment scheduling packet.

6. The communication data transmission method based on data encryption according to claim 1, characterized in that, The generated key encapsulation encryption package includes: The receiving end generates an ML-KEM encapsulation public key and an ML-KEM decapsulation private key, and sends the ML-KEM encapsulation public key to the sending end; The sending end performs key encapsulation based on the ML-KEM encapsulated public key, generates a session shared key and key encapsulation data, and associates them with the ciphertext fragment scheduling packet to establish a session key derivation extension record of the ML-KEM algorithm; Based on the sequence of fragments to be encrypted, the fragment transmission order table, the out-of-order mapping table, and the index association table in the encrypted fragment scheduling packet, generate a fragment path record including session round identifier, fragment sequence number, field marker, link identifier, transmission time window identifier, and index association identifier; In the session key derivation extension record, the session shared key and the fragment path record are processed to generate the initial fragment key sequence. In the session key derivation extension record, the initial fragment key sequence is backfilled with out-of-order fragment keys according to the fragment transmission order table and the out-of-order mapping table to generate a backfilled fragment key sequence. The key-encapsulated data, the sequence of fragments to be encrypted, the fragment sending order table, the fragment path record, the backfilled fragment key sequence, and the index association table are bound together to generate a key-encapsulated encryption package.

7. The communication data transmission method based on data encryption according to claim 1, characterized in that, The generation of the encrypted transmission packet includes: Based on the sequence of fragments to be encrypted, the fragment transmission order table, the backfilled fragment key sequence, and the index association table in the key-encapsulated encryption packet, determine the correspondence between the fragment encryption keys and the fragment transmission order; According to the segment encryption key correspondence, each communication data segment in the segment sequence to be encrypted is matched with the corresponding backfilled segment key to generate a segment encryption task table; Each communication data segment is encrypted segment by segment according to the segment encryption task table, generating a set of ciphertext segments; An integrity identifier is generated based on the ciphertext fragments, fragment sequence numbers, sending time window identifiers, and index association identifiers in the ciphertext fragment set. The integrity identifier is then bound to the corresponding ciphertext fragment, generating a bound ciphertext fragment set. The encrypted fragment set is sent in the order of fragment transmission, and the encrypted transmission packet is generated by associating the key-encapsulated data, fragment path record, index association table and transmission record.

8. The communication data transmission method based on data encryption according to claim 1, characterized in that, The process of generating recovery communication data and transmission feedback records and performing feedback updates includes: The receiving end parses the encrypted transmission packet to obtain the bound encrypted fragment set, key encapsulation data, fragment path record, index association table and sending record; The receiving end decapsulates the key-encapsulated data based on the ML-KEM decapsulation private key, recovers the session shared key, and generates a fragment decryption key sequence based on the session shared key, fragment path record, index association table, and sending record; Integrity verification is performed based on the integrity identifier in the bound ciphertext fragment set. Ciphertext fragments that pass the verification are written into the fragment set to be decrypted, and ciphertext fragments that fail the verification are written into the transmission feedback record. The decryption process is performed segment by segment according to the decryption key sequence. The decrypted communication data segments are then sorted and reassembled based on the index association table and the sending record to generate the recovered communication data. Based on the verification failure records, missing fragment records, retransmission records, and reassembly status records in the transmission feedback records, update the ciphertext transmission heterogeneous graph, improve the MH-Net multi-view heterogeneous graph network parameters, and the key derivation parameters of the ML-KEM algorithm.