A method for encrypting information for maritime rescue

Abstract

The present application provides a kind of offshore rescue information encryption method, comprising: step 1, offshore rescue terminal real-time acquisition multi-source feature information;Step 2, multi-source feature information is embedded in the message header of rescue information after being compressed and masked by encoder;Step 3, based on deep learning disturbance model, disturbance parameters are generated according to multi-source feature information;Step 4, using Beidou timing information, key evolution and derivation are carried out on disturbance parameters, to obtain sub-key;Step 5, according to the encryption strength of information content, the message body of rescue information is encrypted according to sub-key, and message encapsulation is completed;Step 6, the encapsulated message is forwarded to emergency command center through Beidou satellite link, and the key is reconstructed through the same model and timing information, and synchronous decryption and integrity verification are carried out.

Description

Technical Field

[0001] This invention relates to an information encryption method, and more particularly to a maritime rescue information encryption method. Background Technology

[0002] In addition to providing high-precision positioning and timing, the BeiDou Navigation Satellite System (BDS) possesses a unique Short Message Communication (SMC) capability, enabling emergency information transmission in maritime, desert, and extreme environments without terrestrial network coverage. This has made it a crucial tool for maritime rescue and emergency communication. However, current BeiDou SMC transmission largely lacks strong dynamic encryption mechanisms, transmitting data in plaintext or with fixed keys, posing security risks of interception, tampering, or replay. Especially in maritime rescue scenarios, communication links are susceptible to signal interference, ionospheric disturbances, and bandwidth limitations. Therefore, achieving secure and reliable encrypted short message transmission while ensuring real-time performance and low power consumption is a pressing issue that needs to be addressed.

[0003] Specifically, the existing technology has the following drawbacks:

[0004] 1. In existing BeiDou short message communication, encryption methods mostly employ fixed keys or periodically updated mechanisms. Due to the low communication frequency, complex channels, and limited terminal computing power in maritime rescue scenarios, fixed keys are easily intercepted or replayed by attackers, and the encryption strength cannot be dynamically adjusted according to the channel environment, resulting in a high risk of information leakage.

[0005] 2. In maritime emergency communication environments, terminal devices are mostly low-power embedded platforms, which are difficult to run computationally complex symmetric or asymmetric encryption algorithms. Although existing encryption algorithms (such as AES and SM4 full-strength mode) have high security, the encryption and decryption processes are time-consuming and energy-intensive, making them unsuitable for the real-time requirements of short message communication. Summary of the Invention

[0006] Purpose of the invention: The technical problem to be solved by the present invention is to provide a method for encrypting maritime rescue information, which addresses the shortcomings of the existing technology.

[0007] To address the aforementioned technical problems, this invention discloses a method for encrypting maritime rescue information, comprising:

[0008] Step 1: The maritime rescue terminal collects multi-source feature information in real time;

[0009] Step 2: After the multi-source feature information is compressed and masked by the encoder, it is embedded into the message header of the rescue information;

[0010] Step 3: Based on the deep learning perturbation model, generate perturbation parameters according to multi-source feature information;

[0011] Step 4: Using BeiDou timing information, perform key evolution and derivation on the perturbation parameters to obtain subkeys;

[0012] Step 5: Select the encryption strength according to the information content, encrypt the message body of the rescue information according to the subkey, and complete the message encapsulation;

[0013] Step 6: The encapsulated message is forwarded to the emergency command center via the BeiDou satellite link. The key is reconstructed using the same model and timing information, and synchronous decryption and integrity verification are performed.

[0014] Furthermore, the multi-source feature information mentioned in step 1 includes:

[0015] BeiDou system timing parameters, including weekday times and week number ;

[0016] Terminal spatial location parameters, including longitude and latitude ;

[0017] Channel state parameters, including carrier-to-noise ratio Signal strength and bit error rate ;

[0018] Marine environmental characteristics, including wind speed High waves and temperature ;

[0019] After normalization and feature standardization, the input feature vector is obtained. , means as follows:

[0020] .

[0021] Furthermore, the message header in step 2, which involves compressing and masking the multi-source feature information using an encoder and then embedding it into the rescue information, includes:

[0022] Step 2-1, using a feature encoder Input feature vector Mapped to latent space codewords , means as follows:

[0023]

[0024] Among them, the feature encoder A lightweight quantization self-encoder structure is adopted;

[0025] Step 2-2, Hidden Space Encoding A fixed-length bit string is obtained after quantization. Combined with the session key from the previous round Generate pseudo-random mask , means as follows:

[0026]

[0027] in, This represents the XOR operation. This represents a pseudo-random generation function. When referring to the week, This represents a concatenation operation that joins multiple data fragments into a long byte sequence. Indicates the message sequence number;

[0028] Steps 2-3: Apply the pseudo-random mask The message header embedded in the BeiDou short message is sent along with the message.

[0029] Furthermore, the generation of perturbation parameters based on multi-source feature information in step 3 includes:

[0030] A perturbation generation model is constructed using a lightweight neural network, consisting of an input layer, two fully connected hidden layers, and an output layer.

[0031] The perturbation generation model is trained offline by maximizing output entropy and minimizing input correlation.

[0032] The perturbation generation model takes the input feature vector as input. As input, the output is a perturbation vector of length 16. Specifically, it is expressed as follows:

[0033]

[0034] in, For the trained perturbation generation model. These are the model parameters.

[0035] Furthermore, the key evolution and derivation of the perturbation parameters described in step 4 includes:

[0036] Step 4-1: Based on the BeiDou timing information, perform key evolution to generate a unique current session key for each time window. , means as follows:

[0037]

[0038] in, For hash functions, This indicates the session key from the previous time. This represents the XOR operation. Indicates the time within a week;

[0039] Step 4-2: Use the HKDF algorithm to derive subkeys for different purposes from the session key. .

[0040] Furthermore, step 4-2 describes using the HKDF algorithm to derive subkeys for different purposes from the session key. It includes an extraction phase and an expansion phase, as detailed below:

[0041] Extraction stage:

[0042]

[0043] in, It is a pseudo-random key. This refers to a message authentication code algorithm based on a hash function. This is an optional random value. The input key material is the session key in step 4-1. ;

[0044] Expansion phase:

[0045]

[0046]

[0047]

[0048]

[0049]

[0050] in, For context identifier strings, For the first step in the expansion process An intermediate derived block, The required output length, i.e., the number of bytes. That is, the subkey that is ultimately derived. This represents a concatenation operation that joins multiple data segments into a long byte sequence. For the first One counter.

[0051] Furthermore, the encryption of the rescue information message body based on the subkey as described in step 5 includes:

[0052] Step 5-1: Divide the message body of BeiDou short message into 3 layers according to importance: core layer, auxiliary layer and control layer.

[0053] Step 5-2: Use different subkeys for different layers of BeiDou short messages. Encryption is performed, specifically including:

[0054] The core layer uses AES-GCM mode under the AEAD mechanism for encryption; the auxiliary layer uses the lightweight ChaCha20-Reduced algorithm for encryption; and the control layer uses a hash algorithm to generate digests for authentication and verification.

[0055] Step 5-3: Encapsulate the encrypted message body.

[0056] Furthermore, step 6, which involves simultaneous decryption and integrity verification, includes:

[0057] Step 6-1: Extract the pseudo-random mask from the message header. , timestamp, device ID, and model hash value;

[0058] Step 6-2, using the session key from the previous round Weekly seconds and message sequence number Reconstruct the mask stream, perform demasking, and obtain a fixed-length bit string. , means as follows:

[0059]

[0060] Step 6-3, convert the fixed-length bit string Input decoder deployed in the command center To obtain reconstructed features , means as follows:

[0061]

[0062] Step 6-4, the reconstructed features As a perturbation generation model Input, Generate key perturbation parameters for the model parameters. , means as follows:

[0063]

[0064] Step 6-5, use key perturbation parameters Rebuild session key , means as follows:

[0065]

[0066] in For hash functions, This indicates the decoding session key from the previous time. This represents the XOR operation. Indicates seconds within a week.

[0067] Step 6-6: Obtain the session key using the same HKDF algorithm as in step 4-2. Derived subkeys, including decryption keys Authentication key With random number generation key ;

[0068] Steps 6-7: Use the decryption key Authentication key With random number generation key Perform decryption and authentication operations;

[0069] Steps 6-8: Repeat steps 6-1 to 6-7 within a ±1 key cycle time window, and select the first successfully authenticated key branch as the final decryption result.

[0070] Furthermore, the use of the decryption key described in steps 6-7 Authentication key With random number generation key Perform decryption and authentication operations, including:

[0071] Step 6-7-1: For the core layer data, perform AEAD decryption and authentication operations, where the decryption operation is represented as follows:

[0072]

[0073] in, Indicates plain text, This function represents the decryption operation with additional authentication data. This represents the decryption key. This represents the key for generating random numbers. This indicates encrypted data. This indicates additional authentication data. Indicates certification label;

[0074] Authentication operation, that is, using an authentication key Additional authentication data Perform integrity verification with the encrypted text and generate an authentication tag. , means as follows:

[0075]

[0076] If the decryption tag Tag′ matches the received authentication tag Tag, authentication is successful and plaintext is output; otherwise, the message is rejected and an error is recorded.

[0077] Step 6-7-2: For auxiliary layer data, use the derived encryption key. The ChaCha20-Reduced algorithm corresponding to step 5-2 is used to decrypt and recover the ciphertext of the auxiliary layer;

[0078] Step 6-7-3: For control layer data, do not perform decryption; use the authentication key. Hash verification is performed on relevant fields to verify device identity, model version, and message integrity.

[0079] Furthermore, the core layer described in step 5-1 includes pre-defined key sensitive information;

[0080] The auxiliary layer includes pre-defined general business information;

[0081] The control layer includes control information for system identification and consistency verification.

[0082] Beneficial effects:

[0083] 1. This invention proposes an adaptive encryption method based on a deep learning perturbation network. It constructs an input feature vector using BeiDou time synchronization information (TOW), terminal location, channel state parameters (such as C / N0, RSSI, and bit error rate), and environmental characteristics (wind speed, wave height, etc.). A lightweight neural network generates a key perturbation quantity ΔK, enabling dynamic evolution of the key over time and with the environment. This method introduces the nonlinear mapping characteristics of a deep model into the key generation process, ensuring that each short message has a unique encryption key. Simultaneously, it achieves self-synchronization between the two ends through BeiDou time synchronization signals, avoiding the dependence on centralized key distribution found in traditional schemes. This significantly improves the anti-eavesdropping, anti-replay, and anti-forgery capabilities of short message communication, realizing a dynamic encryption mechanism of one key per message, ensuring the security and real-time nature of maritime rescue information.

[0084] 2. This invention designs a lightweight layered encryption mechanism that divides messages into a core layer, an auxiliary layer, and a control layer according to their importance, and encrypts each layer using different strengths of the AEAD (Authorized Encryption with Dispute) algorithm. This method enables maritime rescue terminals to achieve millisecond-level encryption and decryption even in complex sea conditions, reduces terminal power consumption, and improves system reliability and practicality. Attached Figure Description

[0085] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments, and the advantages of the present invention in the above and / or other aspects will become clearer.

[0086] Figure 1 Flowchart for encrypting maritime rescue information.

[0087] Figure 2 This is a schematic diagram of one embodiment. Detailed Implementation

[0088] This invention proposes an adaptive encryption method for short messages in maritime rescue based on a deep learning perturbation model. By introducing a lightweight neural network to generate a key perturbation quantity ΔK, the method achieves dynamic evolution of the key over time, location, and channel state. This method utilizes the TOW (Time-of-Way) mechanism of the BeiDou system to achieve self-synchronization of the key between the sender and receiver, eliminating the need for an additional key distribution channel. Simultaneously, it combines channel state parameters and marine environmental characteristics to generate adaptive perturbations, allowing the encryption strength to be adjusted in real time according to communication conditions. Building upon the advantages of the randomness of chaotic mapping and deep learning key generation, this invention overcomes the shortcomings of traditional algorithms, such as difficulty in synchronization, high complexity, and poor real-time performance, achieving dynamic, adaptive, and intelligent short message encryption. This method significantly improves the security and reliability of maritime rescue communications, providing a new technical approach for the application of the BeiDou short message system in emergency situations.

[0089] This method fully integrates the timing characteristics of the BeiDou satellite system with the features of the maritime communication environment. It utilizes a lightweight neural network to dynamically generate key perturbation parameters, achieving a balance between security, real-time performance, and power consumption in short message encryption. This meets the secure transmission requirements for emergency communication in complex maritime environments. The implementation steps of this invention are as follows:

[0090] 1. Multi-source data acquisition and feature extraction.

[0091] First, multi-source feature information is collected in real time at the maritime rescue terminal, including BeiDou system timing parameters (weektime TOW, week number WN), terminal spatial location (latitude Lat, longitude Lon), channel state parameters (carrier-to-noise ratio C / N0, signal strength RSSI, bit error rate BER), and marine environmental characteristics (wind speed V, wave height Hs, temperature T, etc.). The collected parameters are then normalized and feature standardized to form an input feature vector.

[0092]

[0093] The feature vector is input into the deep learning perturbation generation model to generate the key perturbation parameter ΔK, providing dynamic input for subsequent key evolution.

[0094] To address the issue that the receiving end cannot directly acquire complete multi-source features, this invention introduces an encoder-decoder structure between the terminal and the command center. The feature encoder on the terminal side... Map the original feature X to a fixed-length, high-dimensional, irreversible latent space codeword Z:

[0095]

[0096] The encoder employs a lightweight quantized autoencoder structure (reference: Roy A, Vaswani A, Neelakantan A, et al. Theory and experiments on vector quantized autoencoders[J]. arXivpreprint arXiv:1805.11063, 2018.). The encoder output is quantized to obtain a fixed-length bit string code, which is then combined with the session key. Generated pseudo-random mask Perform lightweight encryption:

[0097]

[0098] in This represents the XOR operation. This represents a pseudo-random generation function. The session key is from the previous cycle; TOW indicates the time within the week. This represents the message sequence number. The code' after the pseudo-random mask is embedded in the message header and sent with the message, thus achieving synchronization feature transmission without revealing specific state information.

[0099] 2. Deep learning perturbation generation and model structure.

[0100] This invention employs a lightweight neural network to construct a perturbation generation module. This network includes an input layer, two fully connected hidden layers, and an output layer. The network input is a feature vector X, and the output is a perturbation vector of length 16. The model is trained offline at a ground-based center. By maximizing output entropy and minimizing input correlation, the output results exhibit high randomness and low predictability. After training, the model parameters are distributed to the offshore terminal for deployment. The terminal only needs to perform forward inference once to generate the model. The calculation formula is:

[0101]

[0102] in, For a well-trained deep network, As input features, These are the model parameters. This module has low computational cost and low power consumption, and can generate high random perturbations in milliseconds.

[0103] 3. Dynamic key evolution and derivation.

[0104] Generate disturbance parameters Then, the system uses BeiDou timing information. Perform key evolution to generate a unique session key for each time window. The evolution formula is:

[0105]

[0106] Where H is the hash function. This indicates the session key from the previous time. This represents the XOR operation. This process allows the key to be dynamically updated over time and with environmental characteristics, achieving true one-key-per-report.

[0107] To achieve functional separation and security isolation, this invention employs the HKDF (HMAC-based Key Derivation Function) algorithm to derive subkeys for different purposes from the session key. The HKDF derivation function is defined as a key extraction and expansion process based on HMAC (Hash-based Message Authentication Code), consisting of two stages.

[0108] Extraction stage:

[0109]

[0110] Where salt is an optional random value (in this scheme, it can be the hash value of the previous key cycle or a fixed constant), and Kinput is the input key material, i.e., the session key. , It is a pseudo-random key. This represents a message authentication code algorithm based on a hash function.

[0111] Expansion phase:

[0112]

[0113]

[0114]

[0115]

[0116]

[0117] in This indicates the nth intermediate derived block in the expansion process, where info is the context identifier string, L is the required output length (in bytes), and OKM (Output Keying Material) is the final derived subkey. This represents a concatenation operation that joins multiple data fragments into a long byte sequence. It includes an encryption key for data encryption, an authentication key for authentication and integrity verification, and a random number key for random number generation. 0x01, 0x02, 0x03, etc., are counters used to mark different stages in the expansion process. 0x01 is typically used in the first round of expansion, representing the first derived subkey. 0x02 is used in the second round of expansion, representing the second generated subkey, and so on. This ensures that each expansion generates a different subkey, avoiding duplicate calculations or key leakage. Through hierarchical derivation, this invention effectively avoids the potential risks associated with key reuse.

[0118] 4. Layered adaptive encryption and message encapsulation.

[0119] To balance security and efficiency within limited bandwidth, this invention divides short messages into three layers based on importance: a core layer, an auxiliary layer, and a control layer. Each of the three encryption layers uses an OKM subkey derived from HKDF for encryption, with different encryption strengths and purposes for each layer. The core layer includes critical sensitive information such as rescue location and mission number, with confidentiality, integrity, and replay resistance as its security objectives. The auxiliary layer includes general business information such as timestamps and status parameters, which support business operations but have lower security sensitivity, with confidentiality as its primary security objective. The control layer includes control information such as device number and model version number, used for system identification and consistency verification, with integrity and source trustworthiness as its security objectives.

[0120] Different types of data are assigned to different levels of encryption based on their importance, using different encryption strengths and methods. The core layer data employs a high-strength encryption method based on the AEAD (Authenticated Encryption with Associated Data) mechanism, implemented using AES-GCM (Advanced Encryption Standard in Galois / Counter Mode), a specific implementation of the AEAD mechanism. The AEAD mechanism integrates encryption and authentication, including data encryption and integrity verification of additional authentication data. The encryption formula for the core layer, which has the highest encryption level in the layered encryption, is as follows:

[0121]

[0122]

[0123] in, This indicates the encrypted data. For certification labels, Indicates an authentication encryption function. The key is EKenc(), which is a symmetric encryption function. The text is plaintext, and Nonce is a one-time random number used to generate an independent counter. Initial value. The authentication key will be The system takes the data to be authenticated as input and outputs an authentication tag, where AAD is additional authentication data (including device ID, version number, and timestamp). The final generated encrypted message structure includes the following fields: version number V, device ID, timestamp Epoch, message sequence number Seq, model hash ModelHash, key period KeyEpoch, random number Nonce, encrypted payload Ciphertext, and authentication tag Tag. This structure ensures both integrity verification and adherence to short message length limits.

[0124] The auxiliary layer data is protected using the lightweight ChaCha20-Reduced symmetric encryption algorithm. Through a simplified encryption process and lower computational complexity, the device can reduce computational burden and energy consumption while maintaining security. Compared to traditional encryption algorithms, ChaCha20-Reduced offers faster encryption and decryption speeds and lower resource consumption, making it suitable for maritime rescue equipment with limited terminal computing power to reduce energy consumption. The control layer data is not symmetrically encrypted; instead, an authentication digest is generated using a hash algorithm for integrity verification and provenance verification.

[0125] 5. Synchronous decryption and authentication mechanism.

[0126] Upon receiving the encrypted message, the emergency command center first extracts the encoded features (code′), timestamp, device ID, and model hash value (ModelHash) from the message header. Then, it utilizes the session key from the previous round. Weekday and message sequence number Reconstruct the mask stream Perform demasking:

[0127]

[0128] in, This represents the XOR operation. This represents a pseudo-random generation function.

[0129] The code was then entered into the decoder deployed at the command center. Reconstructed features obtained:

[0130]

[0131] This feature X at the sending end is statistically equivalent and can be used as input to the perturbation model to generate key perturbation parameters. :

[0132]

[0133] in This represents a perturbation model. These represent model parameters. The session key is then reconstructed using the same key evolution formula. :

[0134]

[0135] Where H is the hash function. Represents the XOR operation. This indicates the session key from the previous time. Indicates the time within a week.

[0136] The receiving end uses the reconstructed session key The HKDF algorithm is used to derive a set of subkeys with isolated uses, including the decryption key. Authentication key and random number generation key .

[0137] For core layer data, the receiving end performs the same AEAD decryption and authentication operations as the sending end:

[0138]

[0139] in Indicates plain text, This function represents the decryption operation with additional authentication data. This represents the decryption key. This represents the key for generating random numbers. This indicates encrypted data. This indicates additional authentication data. This indicates an authentication label.

[0140] Meanwhile, the receiving end uses the authentication key Additional authentication data Perform integrity verification with the encrypted text and generate an authentication tag. :

[0141]

[0142] If the Tag′ matches the received authentication tag Tag, the authentication is successful and plaintext is output; otherwise, the message is rejected and an exception is recorded.

[0143] For auxiliary layer data, the receiving end uses a derived encryption key. The auxiliary layer ciphertext is decrypted and recovered using a lightweight streaming decryption algorithm corresponding to the sending end. For control layer data, the receiving end does not perform decryption; instead, it utilizes the authentication key. Hash authentication is performed on relevant fields to verify device identity, model version, and message integrity.

[0144] Through the above-described layered decryption and authentication process, the receiving end can use deterministic decryption and verification methods for data at different levels, achieving a one-to-one correspondence with the layered encryption process of the sending end.

[0145] To avoid key synchronization issues caused by timing errors, the system repeats the reconstruction process within a ±1 key cycle time window, selecting the first successfully authenticated key branch. Since both the terminal and the command center are synchronized based on BeiDou timing, and the encoding-decoding model is the corresponding version, the system can achieve end-to-end consistency and self-synchronizing key updates without additional handshakes or central key distribution.

[0146] 6. System operation process and technical effects.

[0147] The overall workflow of this invention is as follows: Figure 1 As shown, the maritime rescue terminal first collects multi-source feature data (including BeiDou system timing parameters, terminal spatial location, channel state parameters, and marine environmental characteristics). After compression and masking by an encoder, the generated encrypted data 'code' is embedded in the short message header and based on a deep learning perturbation model. Generate disturbance parameters Then, using BeiDou timing information, key evolution is completed to generate a unique session key for each time window. Different subkeys are derived from the encryption. The system selects encryption strength at different levels and completes short message encryption and encapsulation. After the message is forwarded to the emergency command center via the BeiDou satellite link, the command center reconstructs the key using the same model and timing information. This enables synchronous decryption and integrity verification, thereby achieving secure, reliable, and adaptive transmission of short message communication.

[0148] Example:

[0149] Taking a set of maritime state parameters and rescue information as an example, this paper illustrates the specific working process of the encryption method of the present invention. The original feature data includes time information (weektime 345600 seconds, week number 2250), user location information (longitude 118.78°, latitude 31.95°), channel state parameters (carrier-to-noise ratio 43 dB-Hz, signal strength -85 dBm, bit error rate 1.2e-4), marine environmental characteristics (wind speed 6.3 m / s, wave height 1.4 m, temperature 28°C), and mission number. The encryption method proposed in this invention generates a mask encoding during the process. Perturbation vector Evolved key Derived encryption key Authentication key Encrypted ciphertext Certification Labels And the final decrypted plaintext, such as Figure 2 As shown.

[0150] In its specific implementation, this application provides a computer storage medium and a corresponding data processing unit. The computer storage medium is capable of storing a computer program, which, when executed by the data processing unit, can run the invention's content regarding a maritime rescue information encryption method and some or all of the steps in various embodiments. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0151] Those skilled in the art will clearly understand that the technical solutions in the embodiments of the present invention can be implemented using computer programs and their corresponding general-purpose hardware platforms. Based on this understanding, the technical solutions in the embodiments of the present invention, or the parts that contribute to the prior art, can be embodied in the form of computer programs, i.e., software products. These computer program software products can be stored in a storage medium and include several instructions to cause a device containing a data processing unit (which may be a personal computer, server, microcontroller, MCU, or network device, etc.) to execute the methods described in various embodiments or certain parts of the embodiments of the present invention.

[0152] This invention provides a concept and method for encrypting maritime rescue information. Many methods and approaches exist for implementing this technical solution; the above description is merely a preferred embodiment. It should be noted that those skilled in the art can make various improvements and modifications without departing from the principles of this invention, and these improvements and modifications should also be considered within the scope of protection of this invention. All components not explicitly stated in this embodiment can be implemented using existing technologies.

Claims

1. A method for encrypting maritime rescue information, characterized in that, include: Step 1: The maritime rescue terminal collects multi-source feature information in real time; Step 2: After the multi-source feature information is compressed and masked by the encoder, it is embedded into the message header of the rescue information; Step 3: Based on the deep learning perturbation model, generate perturbation parameters according to multi-source feature information; Step 4: Using BeiDou timing information, perform key evolution and derivation on the perturbation parameters to obtain subkeys; Step 5: Select the encryption strength according to the information content, encrypt the message body of the rescue information according to the subkey, and complete the message encapsulation; Step 6: The encapsulated message is forwarded to the emergency command center via the BeiDou satellite link. The key is reconstructed using the same model and timing information, and synchronous decryption and integrity verification are performed. The multi-source feature information mentioned in step 1 includes: BeiDou system timing parameters, including weekday times and week number ; Terminal spatial location parameters, including longitude and latitude ; Channel state parameters, including carrier-to-noise ratio Signal strength and bit error rate ; Marine environmental characteristics, including wind speed High waves and temperature ; After normalization and feature standardization, the input feature vector is obtained. , means as follows: ； Step 2, which involves embedding the multi-source feature information into the message header of the rescue information after compression and masking by the encoder, includes: Step 2-1, using a feature encoder Input feature vector Mapped to latent space codewords , means as follows: ； Among them, the feature encoder A lightweight quantization self-encoder structure is adopted; Step 2-2, Hidden Space Encoding A fixed-length bit string is obtained after quantization. Combined with the session key from the previous round Generate pseudo-random mask , means as follows: ； in, This represents the XOR operation. This represents a pseudo-random generation function. When referring to the week, This represents a concatenation operation that joins multiple data fragments into a long byte sequence. Indicates the message sequence number; Steps 2-3: Apply the pseudo-random mask The message header embedded in the BeiDou short message is sent along with the message; Step 3, which involves generating perturbation parameters based on multi-source feature information, includes: A perturbation generation model is constructed using a lightweight neural network, consisting of an input layer, two fully connected hidden layers, and an output layer. The perturbation generation model is trained offline by maximizing output entropy and minimizing input correlation. The perturbation generation model takes the input feature vector as input. As input, the output is a perturbation vector of length 16. Specifically, it is expressed as follows: ； in, For the trained perturbation generation model. These are the model parameters.

2. The maritime rescue information encryption method according to claim 1, characterized in that, Step 4, which describes key evolution and derivation of the perturbation parameters, includes: Step 4-1: Based on the BeiDou timing information, perform key evolution to generate a unique current session key for each time window. , means as follows: ； in, For hash functions, This indicates the session key from the previous time. This represents the XOR operation. Indicates the time within a week; Step 4-2: Use the HKDF algorithm to derive subkeys for different purposes from the session key. .

3. The maritime rescue information encryption method according to claim 2, characterized in that, Step 4-2 describes using the HKDF algorithm to derive subkeys for different purposes from the session key. It includes an extraction phase and an expansion phase, as detailed below: Extraction stage: ； in, It is a pseudo-random key. This refers to a message authentication code algorithm based on a hash function. This is an optional random value. The input key material is the session key in step 4-1. ; Expansion phase: ； ； …… ； ； in, For context identifier strings, For the first step in the expansion process An intermediate derived block, The required output length, i.e., the number of bytes. That is, the subkey that is ultimately derived. This represents a concatenation operation that joins multiple data segments into a long byte sequence. For the first One counter.

4. The maritime rescue information encryption method according to claim 3, characterized in that, The encryption of the rescue information message body based on the subkey in step 5 includes: Step 5-1: Divide the message body of BeiDou short message into 3 layers according to importance: core layer, auxiliary layer and control layer. Step 5-2: Use different subkeys for different layers of BeiDou short messages. Encryption is performed, specifically including: The core layer uses AES-GCM mode under the AEAD mechanism for encryption; the auxiliary layer uses the lightweight ChaCha20-Reduced algorithm for encryption; and the control layer uses a hash algorithm to generate digests for authentication and verification. Step 5-3: Encapsulate the encrypted message body.

5. A method for encrypting maritime rescue information according to claim 4, characterized in that, Step 6, which involves simultaneous decryption and integrity verification, includes: Step 6-1: Extract the pseudo-random mask from the message header. , timestamp, device ID, and model hash value; Step 6-2, using the session key from the previous round Weekly seconds and message sequence number Reconstruct the mask stream, perform demasking, and obtain a fixed-length bit string. , means as follows: ； Step 6-3, convert the fixed-length bit string Input decoder deployed in the command center To obtain reconstructed features , means as follows: ； Step 6-4, the reconstructed features As a perturbation generation model Input, Generate key perturbation parameters for the model parameters. , means as follows: ； Step 6-5, use key perturbation parameters Rebuild session key , means as follows: ； in For hash functions, This indicates the decoding session key from the previous time. This represents the XOR operation. Indicates the second within the week; Step 6-6: Obtain the session key using the same HKDF algorithm as in step 4-2. Derived subkeys, including decryption keys Authentication key With random number generation key ; Steps 6-7: Use the decryption key Authentication key With random number generation key Perform decryption and authentication operations; Steps 6-8: Repeat steps 6-1 to 6-7 within a ±1 key cycle time window, and select the first successfully authenticated key branch as the final decryption result.

6. A method for encrypting maritime rescue information according to claim 5, characterized in that, Steps 6-7 describe the use of the decryption key. Authentication key With random number generation key Perform decryption and authentication operations, including: Step 6-7-1: For the core layer data, perform AEAD decryption and authentication operations, where the decryption operation is represented as follows: ； in, Indicates plain text, This function represents the decryption operation with additional authentication data. This represents the decryption key. This represents the key for generating random numbers. This indicates encrypted data. This indicates additional authentication data. Indicates certification label; Authentication operation, that is, using an authentication key Additional authentication data Perform integrity verification with the encrypted text and generate an authentication tag. , means as follows: ； If the decryption tag Tag′ matches the received authentication tag Tag, authentication is successful and plaintext is output; otherwise, the message is rejected and an error is recorded. Step 6-7-2: For auxiliary layer data, use the derived encryption key. The ChaCha20-Reduced algorithm corresponding to step 5-2 is used to decrypt and recover the ciphertext of the auxiliary layer; Step 6-7-3: For control layer data, do not perform decryption; use the authentication key. Hash verification is performed on relevant fields to verify device identity, model version, and message integrity.

7. A method for encrypting maritime rescue information according to claim 6, characterized in that, The core layer described in step 5-1 includes pre-defined key sensitive information; The auxiliary layer includes pre-defined general business information; The control layer includes control information for system identification and consistency verification.

Images