Digital human encryption interaction method and interaction system

Abstract

The invention discloses a digital human encryption interaction method and interaction system. According to the method, the reinforced key is constructed through the randomness of a natural action track, the key is enabled to simultaneously bear dual attributes of unique biological characteristics and behavior characteristics of a user, and the randomness is further amplified through cyclic expansion and XOR processing of the temporary key, so that the key is difficult to be cracked by brute force or attacked by mode matching. In combination with multi-factor verification of early-stage secret language recognition and voiceprint collection, multi-layer protection is established from an interaction starting point, hijacking difficulty is increased in a transmission link through subsequent segmented encryption and feature code embedding, all steps are mutually associated to form a closed loop, common encryption cracking means are effectively resisted, an encryption process is completely embedded into a daily interaction scene, and the encryption efficiency is improved. A user does not need to learn complex encryption operation, secret key generation and content encryption can be completed through preset secret words and natural actions, state switching is visual and perceptible through micro expressions and intonation changes of a digital person, and operation confusion of the user is avoided.

Description

Technical Field

[0001] This invention belongs to the field of digital human interaction technology, specifically a digital human encrypted interaction method and system. Background Technology

[0002] Digital human interaction refers to the process by which users exchange information and communicate dynamically with virtual avatars that exist in digital form and possess human-like appearance, behavior, or intelligent characteristics. This process relies on cutting-edge technologies such as computer graphics, artificial intelligence, natural language processing, and speech synthesis. Through various input methods including text, voice, gestures, and even facial expression capture, it enables real-time dialogue, emotional feedback, and task collaboration between users and digital humans. Its core lies in breaking away from the coldness of traditional human-computer interaction by endowing digital humans with realistic appearances and intelligent "souls," enabling them to understand user intentions and provide human-like responses. This provides a more immersive and approachable human-computer interaction experience in fields such as virtual customer service, online education, intelligent navigation, and entertainment idols, and is a key technological link in building the future metaverse ecosystem and intelligent service scenarios.

[0003] However, existing technologies often suffer from a disconnect between security protection and user experience in digital human encrypted interaction scenarios. They either rely on single encryption methods such as static passwords or simple voiceprints, lacking multi-factor closed-loop protection that combines motion trajectories and coded messages, making them vulnerable to brute-force attacks or identity impersonation. Summary of the Invention

[0004] The purpose of this invention is to provide a digital human encrypted interaction method and system in order to solve the problems mentioned above.

[0005] The technical solution adopted in this invention is as follows: a digital human encrypted interaction method, the method comprising the following steps: S1: The user initiates a daily greeting with a preset code word to the digital human. After recognizing the code word, the digital human activates an encrypted interaction mode and provides feedback on its status through micro-expression changes.

[0006] S2: The digital human collects the unique voiceprint frequency features in the user's current speech, combines them with the hash value of the S1 code word keyword, and generates a 128-bit temporary session key; the key is temporarily stored in a local encrypted cache, is only valid for this interaction, and is strongly associated with the S1 code word.

[0007] S3: The digital human guides the user to complete random action commands, captures the action trajectory coordinate sequence and converts it into a binary string, and performs an XOR operation with the temporary key of S2 to obtain the enhancement key.

[0008] S4: The user inputs encrypted content, and the digital human divides the content into paragraphs. Each paragraph is encrypted with AES-256 using the enhanced key from S3, and the corresponding motion trajectory feature code is embedded in the paragraph for subsequent segment verification.

[0009] S5: After sending three ciphertexts, the digital human will confirm the details of the S1 code with the user. Only if the user answers correctly will the transmission continue.

[0010] S6: After receiving the ciphertext, the receiving digital human extracts each segment of the feature code and matches it with the action trajectory of S3 for verification; after the verification is successful, it calls the temporary key template of S2 and waits for the user to rebuild the reinforcement key.

[0011] S7: The receiver guides the user to repeat the action of S3, captures the new trajectory and converts it into a binary string, XORs it with the S2 template to reconstruct the strengthening key; if the action similarity exceeds 90%, it is successful, otherwise the operation is repeated, and the logic of S2 and S3 is associated.

[0012] S8: Decrypt each segment using the reconstructed enhanced key and verify its integrity by comparing the hash value; after all segments have passed, the decrypted content is presented in its entirety, and the decryption logic corresponds one-to-one with the encryption rules of S4.

[0013] S9: After the interaction ends, the digital human immediately deletes the cached key and trajectory data, and subtly changes its tone to indicate the end of the session; ensuring that sensitive data is not retained, and that the closing process is seamlessly connected with the previous interaction.

[0014] In a preferred embodiment, in step S1, the user initiates a daily greeting with a preset code word to the digital human. The code word must contain 3 to 5 keywords and be semantically coherent. The digital human identifies the code word using a keyword matching algorithm in its natural language processing module. When the matching degree reaches 95% or higher, an encrypted interaction mode is activated. At the same time, the digital human's micro-expressions change, the blinking frequency increases from once per second to 1.5 times per second for 2 seconds, and the voice tone slightly decreases by 0.1 pitch units, allowing the user to clearly perceive the state switch. This code word needs to be agreed upon in advance by the user and the digital human in an unencrypted scenario. During the agreement process, the digital human will record the semantic features and pronunciation habits of the code word as a basic verification factor for subsequent steps to ensure the security of subsequent interactions.

[0015] In a preferred embodiment, in step S2, the digital human collects unique voiceprint frequency features from the user's current speech for 1.5 seconds, extracting 128-dimensional features including fundamental frequency formants and Mel frequency cepstral coefficients. This is combined with the hash value of the coded keyword from step S1 to generate a 128-bit temporary session key. The hash value is generated using the SHA-256 algorithm, producing a 256-bit result. The first 128 bits are then weighted and fused with the voiceprint features bit by bit, with a weighting coefficient of 0.6 for the voiceprint and 0.4 for the hash value. The key is temporarily stored in an encrypted cache in local RAM, protected by AES-128 encryption. The cache validity period is the duration of the current interaction plus 5 minutes, after which it is automatically cleared. The key generation process must be completed within 100 milliseconds to ensure the smoothness of the interaction process. Furthermore, the key is strongly correlated with the coded message from step S1 to ensure consistency in subsequent verification.

[0016] In a preferred embodiment, in step S3, the user performs a random action in front of the display interface (coordinate range x∈[0,1920], y∈[0,1080]), and the digital human collects 2 seconds of action data at a frequency of 30 frames / second through the front-facing camera to obtain 60 consecutive finger coordinate points. First, calculate the offset of each point relative to the starting point: , Eliminate differences in starting position; Find all of them again The absolute maximum value | and The absolute maximum value of max|Δy| Normalize the offset to an 8-bit integer range of 0-255 (e.g., ), to obtain the normalized coordinate sequence .

[0017] Next, each normalization point and Convert to an 8-bit binary string, concatenate them into a 16-bit substring, and then connect all the points to form a 960-bit trajectory binary string B. Take the 128-bit temporary key generated in step S2. Expanding the loop to 960 bits yields Finally, for and Bitwise XOR operation yields the enhanced key. This process increases the unpredictability of the key through the randomness of natural actions, thus preventing cracking based on fixed patterns. The normalization formula for coordinate offset is: ; ; In the formula: For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; This is the rounding function; The formula for cyclic expansion of temporary keys is: {K}_{ext}\left [ {j} \right ]=K\left [ {j\, mod\, len(K)} \right ]\, ; In the formula: {K}_{ext}\left [ {j} \right ] For the extended temporary key Bit; The temporary session key generated in step S2; The length of the temporary key; The bit index of the binary string; The formula for generating the enhanced key is: ; {K}_{strength}\left [ {j} \right ] For the strengthening key Bit; B\left [ {j} \right ] The first binary string after the motion trajectory conversion Bit; This is the bitwise XOR operator; {K}_{ext}\left [ {j} \right ] Consistent with the definition in the temporary key cyclic expansion formula.

[0018] In a preferred embodiment, in step S4, after the user inputs encrypted content, the digital human segments the content into segments of 200 characters each, with any segment less than 200 characters being the last segment. Each segment is encrypted using the enhanced key generated in step S3 with AES-256 encryption in CBC mode. The initialization vector IV is generated by converting the ASCII codes of the first 16 characters of the segment and embedding the corresponding motion trajectory feature code. The feature code generation rule is the same as in step S3, i.e., one feature point is taken from every 10 points in the motion trajectory coordinate sequence, converted into a 32-bit binary string, and embedded in the header of the ciphertext. After each ciphertext segment is generated, the SHA-256 hash value of that segment is calculated and stored together with the feature code for subsequent segment verification. The encryption process must ensure that the processing time for each segment does not exceed 50 milliseconds to avoid excessively long waiting times for the user.

[0019] In a preferred embodiment, in step S5, after the digital human sends three ciphertexts, it will confirm the details of the code from step S1 with the user. The question is generated based on the semantic structure of the code and selects the associated information of a keyword in the code, such as the usage scenario or specific collocation of the keyword. The user needs to answer the question by voice. The digital human verifies the accuracy of the answer through the speech recognition and semantic understanding module. The accuracy rate must reach more than 90% before continuing to send subsequent ciphertexts. If the answer is wrong, the digital human will repeat the same question, up to 2 times. If it is still wrong, the interaction will be terminated. This mechanism, combined with the information in step S1, effectively prevents the occurrence of man-in-the-middle hijacking of ciphertexts.

[0020] In a preferred embodiment, in step S6, after receiving the ciphertext, the receiving digital human first parses the feature code in the header of each ciphertext segment, converts the feature code into a coordinate sequence fragment, and then performs a matching verification with the action trajectory feature library stored in step S3. The matching algorithm adopts dynamic time warping, and the similarity threshold is set to 90%. The segments that pass the match are marked as valid, and the segments that fail are discarded directly. Then, the receiving party calls the temporary key template generated in step S2. This template contains voiceprint feature dimensions and hash value location information, and waits for the user to provide the action data required to reconstruct the reinforcement key. The entire verification process must be completed within 1 second after receiving the ciphertext to ensure the real-time nature of the interaction.

[0021] In a preferred embodiment, in step S7, the receiving digital human guides the user to repeat the action of step S3. The guidance uses natural speech to clearly inform the user of the action type. The motion capture uses the same frame rate as step S3 (30 frames / second) for 2 seconds, and collects 60 coordinate points. The method for converting the captured motion trajectory coordinate sequence into a binary string is the same as in step S3, that is, normalizing it to the range of 0-255 and then converting it to binary. Then, it is XORed with the temporary key template in step S2 to reconstruct the strengthening key. The reconstructed key is compared with the strengthening key generated in step S3. The similarity is calculated using Hamming distance. If the Hamming distance is less than 10, it is considered successful. Otherwise, the user is prompted to repeat the operation. A maximum of 3 attempts are allowed. If it still fails, the interaction is terminated. This step is related to the logic of steps S2 and S3 to ensure the accuracy of key reconstruction.

[0022] In a preferred embodiment, in step S8, the receiving digital human decrypts each segment of content using the reconstructed enhanced key. The decryption adopts the AES-256 CBC mode corresponding to step S4. The initialization vector IV is extracted from the ciphertext header. After each segment is decrypted, the SHA-256 hash value of that segment is calculated and compared with the hash value attached to the ciphertext. If they match completely, it is marked as complete. After all segments are decrypted and verified, the digital human reassembles the content in the original order and presents the decrypted content completely via text or voice. If the hash value of a segment does not match, the digital human will prompt that the segment is corrupted and ask the user whether to re-receive the ciphertext. The decryption process must ensure that the processing time for each segment does not exceed 30 milliseconds to ensure that the content is presented to the user quickly.

[0023] In a preferred embodiment, in step S9, the condition for determining the end of the interaction is that the user sends a termination command or there is no operation within 5 minutes. The digital human immediately calls the local cache cleanup module to delete all sensitive information such as cached keys, motion trajectory data, and ciphertext fragments. The cleanup process must be completed within 50 milliseconds. At the same time, the conversation is terminated by a slight change in tone, which is manifested by raising the voice pitch by 0.15 pitch units and adding a brief closing remark. After the cleanup is completed, the digital human returns to the normal interaction mode, and all encryption-related modules are in a dormant state until the next ciphertext is triggered.

[0024] In summary, due to the adoption of the above technical solution, the beneficial effects of the present invention are: 1. In this invention, a strengthened key is constructed using the randomness of natural movement trajectories, allowing the key to simultaneously carry the dual attributes of the user's unique biometric and behavioral characteristics. The cyclic expansion and XOR processing of the temporary key further amplify the randomness, making the key difficult to crack through brute force or pattern matching attacks. Combined with multi-factor verification of coded language recognition and voiceprint collection in the early stages, multi-layered protection is established from the starting point of the interaction. Subsequent segmented encryption and feature code embedding further increase the difficulty of hijacking during transmission. Each step is interconnected, forming a closed loop that effectively resists common encryption cracking methods.

[0025] 2. In this invention, the encryption process is fully embedded in everyday interaction scenarios. Users do not need to learn complex encryption operations; they can generate keys and encrypt content simply by using preset codes and natural actions. The digital human's micro-expressions and tone of voice make state transitions intuitive and perceptible, avoiding user confusion. The key reconstruction stage reuses action trajectory features, so the receiver does not need to memorize keys; simply repeating the same actions quickly restores decryption capabilities. Simultaneously, segmented encryption and real-time verification mechanisms ensure transmission efficiency and content integrity, making encrypted interaction both secure and practical, suitable for high-frequency, private information transmission scenarios. Attached Figure Description

[0026] Figure 1 This is a schematic diagram illustrating the process principle of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention. Example

[0028] Reference Figure 1 A digital human encrypted interaction method and system, the method comprising the following steps: S1: The user initiates a daily greeting with a preset code word to the digital human. After recognizing the code word, the digital human activates an encrypted interaction mode and provides feedback on its status through micro-expression changes.

[0029] S2: The digital human collects the unique voiceprint frequency features in the user's current speech, combines them with the hash value of the S1 code word keyword, and generates a 128-bit temporary session key; the key is temporarily stored in a local encrypted cache, is only valid for this interaction, and is strongly associated with the S1 code word.

[0030] S3: The digital human guides the user to complete random action commands, captures the action trajectory coordinate sequence and converts it into a binary string, and performs an XOR operation with the temporary key of S2 to obtain the enhancement key.

[0031] S4: The user inputs encrypted content, and the digital human divides the content into paragraphs. Each paragraph is encrypted with AES-256 using the enhanced key from S3, and the corresponding motion trajectory feature code is embedded in the paragraph for subsequent segment verification.

[0032] S5: After sending three ciphertexts, the digital human will confirm the details of the S1 code with the user. Only if the user answers correctly will the transmission continue.

[0033] S6: After receiving the ciphertext, the receiving digital human extracts each segment of the feature code and matches it with the action trajectory of S3 for verification; after the verification is successful, it calls the temporary key template of S2 and waits for the user to rebuild the reinforcement key.

[0034] S7: The receiver guides the user to repeat the action of S3, captures the new trajectory and converts it into a binary string, XORs it with the S2 template to reconstruct the strengthening key; if the action similarity exceeds 90%, it is successful, otherwise the operation is repeated, and the logic of S2 and S3 is associated.

[0035] S8: Decrypt each segment using the reconstructed enhanced key and verify its integrity by comparing the hash value; after all segments have passed, the decrypted content is presented in its entirety, and the decryption logic corresponds one-to-one with the encryption rules of S4.

[0036] S9: After the interaction ends, the digital human immediately deletes the cached key and trajectory data, and subtly changes its tone to indicate the end of the session; ensuring that sensitive data is not retained, and that the closing process is seamlessly connected with the previous interaction.

[0037] In step S1, the user initiates a daily greeting with a preset code word to the digital human. The code word must contain 3 to 5 keywords and be semantically coherent. The digital human identifies the code word using a keyword matching algorithm in its natural language processing module. When the matching degree reaches 95% or higher, it activates the encrypted interaction mode. At the same time, the digital human's micro-expressions change, the blinking frequency increases from once per second to 1.5 times per second for 2 seconds, and the voice tone slightly decreases by 0.1 pitch units, allowing the user to clearly perceive the state switch. This code word needs to be agreed upon in advance by the user and the digital human in an unencrypted scenario. During the agreement process, the digital human will record the semantic features and pronunciation habits of the code word as a basic verification factor for subsequent steps to ensure the security of subsequent interactions.

[0038] In step S2, the digital human collects unique voiceprint frequency features from the user's current speech for 1.5 seconds, extracting 128-dimensional features including fundamental frequency formants and Mel frequency cepstral coefficients. This is combined with the hash value of the coded keyword from step S1 to generate a 128-bit temporary session key. The hash value is generated using the SHA-256 algorithm, producing a 256-bit result. The first 128 bits are then weighted and fused with the voiceprint features bit by bit, with a weighting coefficient of 0.6 for the voiceprint and 0.4 for the hash value. The key is temporarily stored in an encrypted cache in local RAM, protected by AES-128 encryption. The cache's validity period is the duration of this interaction plus 5 minutes, after which it is automatically cleared. The key generation process must be completed within 100 milliseconds to ensure smooth interaction. Furthermore, the key is strongly correlated with the coded message from step S1 to ensure consistency in subsequent verification.

[0039] In step S3, The user performs random actions in front of the display interface (coordinate range x∈[0,1920], y∈[0,1080]). The digital human collects 2 seconds of action data at a frequency of 30 frames / second through the front-facing camera, obtaining 60 consecutive finger coordinate points. First, calculate the offset of each point relative to the starting point: , Eliminate differences in starting position; Find all of them again The absolute maximum value | and The absolute maximum value of max|Δy| Normalize the offset to an 8-bit integer range of 0-255 (e.g., ), to obtain the normalized coordinate sequence .

[0040] Next, each normalization point and Convert to an 8-bit binary string, concatenate them into a 16-bit substring, and then connect all the points to form a 960-bit trajectory binary string B. Take the 128-bit temporary key generated in step S2. Expanding the loop to 960 bits yields Finally, for and Bitwise XOR operation yields the enhanced key. This process increases the unpredictability of the key through the randomness of natural actions, thus preventing cracking based on fixed patterns. The normalization formula for coordinate offset is: ; ; In the formula: For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; This is the rounding function; The formula for cyclic expansion of temporary keys is: {K}_{ext}\left [ {j} \right ]=K\left [ {j\, mod\, len(K)} \right ]\, ; In the formula: {K}_{ext}\left [ {j} \right ] For the extended temporary key Bit; The temporary session key generated in step S2; The length of the temporary key; The bit index of the binary string; The formula for generating the enhanced key is: ; {K}_{strength}\left [ {j} \right ] For the strengthening key Bit; B\left [ {j} \right ] The first binary string after the motion trajectory conversion Bit; This is the bitwise XOR operator; {K}_{ext}\left [ {j} \right ] Consistent with the definition in the temporary key cyclic expansion formula.

[0041] In step S4, after the user inputs the encrypted content, the digital human segments the content into segments of 200 characters each, with any segment less than 200 characters being the last segment. Each segment is encrypted using the enhanced key generated in step S3 with AES-256 encryption in CBC mode. The initialization vector IV is generated by converting the ASCII codes of the first 16 characters of the segment and embedding the corresponding motion trajectory feature code. The feature code generation rule is the same as in step S3, that is, one feature point is taken from every 10 points in the motion trajectory coordinate sequence, converted into a 32-bit binary string, and embedded in the header of the ciphertext. After each ciphertext segment is generated, the SHA-256 hash value of that segment is calculated and stored together with the feature code for subsequent segment verification. The encryption process must ensure that the processing time for each segment does not exceed 50 milliseconds to avoid excessively long waiting times for the user.

[0042] In step S5, after sending three ciphertexts, the digital human will confirm the details of the code from step S1 with the user. The question is generated based on the semantic structure of the code and selects the associated information of a keyword in the code, such as the usage scenario or specific collocation of the keyword. The user needs to answer the question by voice. The digital human verifies the accuracy of the answer through the speech recognition and semantic understanding modules. The accuracy rate must reach more than 90% before continuing to send subsequent ciphertexts. If the answer is wrong, the digital human will repeat the same question, up to 2 times. If it is still wrong, the interaction will be terminated. This mechanism, combined with the information from step S1, effectively prevents the occurrence of man-in-the-middle hijacking of ciphertexts.

[0043] In step S6, after receiving the ciphertext, the receiving digital human first parses the feature code in the header of each ciphertext segment, converts the feature code into a coordinate sequence fragment, and then performs a matching verification with the action trajectory feature library stored in step S3. The matching algorithm adopts dynamic time warping, and the similarity threshold is set to 90%. The segments that pass the match are marked as valid, and the segments that fail are discarded directly. Then, the receiving party calls the temporary key template generated in step S2. This template contains voiceprint feature dimensions and hash value location information, and waits for the user to provide the action data required to reconstruct the reinforcement key. The entire verification process must be completed within 1 second after receiving the ciphertext to ensure the real-time nature of the interaction.

[0044] In step S7, the receiving digital human guides the user to repeat the actions of step S3. The guidance uses natural speech to clearly inform the user of the action type. The motion capture uses the same frame rate as step S3, 30 frames / second, for a duration of 2 seconds, and collects 60 coordinate points. The method for converting the captured motion trajectory coordinate sequence into a binary string is the same as in step S3, that is, normalizing it to the range of 0-255 and then converting it to binary. Then, it is XORed with the temporary key template in step S2 to reconstruct the strengthening key. The reconstructed key is compared with the strengthening key generated in step S3. The similarity is calculated using Hamming distance. If the Hamming distance is less than 10, it is considered successful. Otherwise, the user is prompted to repeat the operation. A maximum of 3 attempts are allowed. If it still fails, the interaction is terminated. This step is related to the logic of steps S2 and S3 to ensure the accuracy of key reconstruction.

[0045] In step S8, the receiving digital human decrypts each segment of content using the reconstructed enhanced key. The decryption adopts the AES-256 CBC mode corresponding to step S4. The initialization vector IV is extracted from the ciphertext header. After each segment is decrypted, the SHA-256 hash value of that segment is calculated and compared with the hash value attached to the ciphertext. If they match completely, it is marked as complete. After all segments are decrypted and verified, the digital human reassembles the content in the original order and presents the decrypted content completely via text or voice. If the hash value of a segment does not match, the digital human will prompt that the segment is corrupted and ask the user whether to re-receive the ciphertext. The decryption process must ensure that the processing time of each segment does not exceed 30 milliseconds to ensure that the content is presented to the user quickly.

[0046] In step S9, the conditions for determining the end of the interaction are that the user sends a termination command or there is no operation within 5 minutes. The digital human immediately calls the local cache cleanup module to delete all sensitive information such as cached keys, motion trajectory data, and encrypted fragments. The cleanup process must be completed within 50 milliseconds. At the same time, the conversation is terminated by a slight change in tone, which is manifested by raising the tone of voice by 0.15 pitch units and adding a brief closing remark. After the cleanup is completed, the digital human returns to the normal interaction mode, and all encryption-related modules are in a dormant state until the next coded message is triggered.

[0047] From the above, we can conclude that: In this invention, a strengthened key is constructed using the randomness of natural movement trajectories, allowing the key to simultaneously carry the dual attributes of a user's unique biometric and behavioral characteristics. The cyclic expansion and XOR processing of the temporary key further amplify the randomness, making the key difficult to crack through brute force or pattern matching attacks. Combined with multi-factor verification of coded language recognition and voiceprint acquisition in the early stages, multi-layered protection is established from the very beginning of the interaction. Subsequent segmented encryption and feature code embedding further increase the difficulty of hijacking during transmission. Each step is interconnected, forming a closed loop that effectively resists common encryption cracking methods.

[0048] In this invention, the encryption process is fully embedded in everyday interaction scenarios. Users do not need to learn complex encryption operations; they can generate keys and encrypt content simply by using preset codes and natural actions. The digital human's micro-expressions and tone of voice make state transitions intuitive and perceptible, avoiding user confusion. The key reconstruction stage reuses action trajectory features, so the receiver does not need to memorize keys; simply repeating the same actions quickly restores decryption capabilities. At the same time, segmented encryption and real-time verification mechanisms ensure transmission efficiency and content integrity, making encrypted interaction both secure and practical, suitable for high-frequency private information transmission scenarios.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the term "comprising" or any other variations thereof is intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

[0050] The above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit it. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A digital human encrypted interaction method, characterized in that: The method includes the following steps: S1: The user initiates a daily greeting with a preset code word to the digital human. After recognizing the code word, the digital human activates the encrypted interaction mode and provides feedback on its status through micro-expression changes. S2: The digital human collects the unique voiceprint frequency features in the user's current speech, combines them with the hash value of the S1 code word keyword, and generates a 128-bit temporary session key; the key is temporarily stored in a local encrypted cache, is only valid for this interaction, and is strongly associated with the S1 code word; S3: The digital human guides the user to complete random action instructions, captures the action trajectory coordinate sequence and converts it into a binary string, and performs an XOR operation with the temporary key of S2 to obtain the enhancement key; S4: The user inputs encrypted content, and the digital human divides the content into paragraphs. Each paragraph is encrypted with AES-256 using the enhanced key of S3, and the corresponding motion trajectory feature code is embedded in the paragraph for subsequent segment verification. S5: After sending three ciphertexts, the digital human will confirm the details of the S1 code with the user. Only if the user answers correctly will the transmission continue. S6: After receiving the ciphertext, the receiving digital human extracts each segment of the feature code and matches and verifies it with the action trajectory of S3; after the verification is successful, the temporary key template of S2 is called and the user is asked to rebuild the reinforcement key. S7: The receiver guides the user to repeat the action of S3, captures the new trajectory, converts it into a binary string, and XORs it with the S2 template to reconstruct the strengthening key; if the action similarity exceeds 90%, it is successful; otherwise, the operation is repeated, and the logic of S2 and S3 is associated. S8: Decrypt each segment using the reconstructed enhanced key and verify its integrity by comparing the hash value; after all segments have passed, the decrypted content is presented in its entirety, and the decryption logic corresponds one-to-one with the encryption rules of S4. S9: After the interaction ends, the digital human immediately deletes the cached key and trajectory data, and subtly changes its tone to indicate the end of the session; ensuring that sensitive data is not retained, and that the closing process is seamlessly connected with the previous interaction.

2. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S1, the user initiates a daily greeting with a preset code word to the digital human. The code word must contain 3 to 5 keywords and be semantically coherent. The digital human identifies the code word through the keyword matching algorithm in the natural language processing module. When the matching degree reaches more than 95%, the encrypted interaction mode is activated. At the same time, the digital human's micro-expression changes, the blinking frequency increases from 1 time per second to 1.5 times per second for 2 seconds, and the tone of voice is slightly reduced by 0.1 pitch units.

3. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S2, the digital human collects unique voiceprint frequency features from the user's current speech for 1.5 seconds. It extracts 128-dimensional features, including fundamental frequency formants and Mel frequency cepstral coefficients, from the speech. Combined with the hash value of the coded keyword from step S1, a 128-bit temporary session key is generated. The hash value is generated using the SHA-256 algorithm, resulting in a 256-bit result. The first 128 bits are then used for weighted fusion with the voiceprint features. The weighting coefficient is set to 0.6 for the voiceprint and 0.4 for the hash value. The key is temporarily stored in an encrypted cache in the local RAM. The cache is protected by AES-128 encryption and its validity period is the duration of this interaction plus 5 minutes. It is automatically cleared after the timeout.

4. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S3, the user performs a random action in front of the display interface, and the digital human collects 2 seconds of action data at a frequency of 30 frames per second through the front-facing camera, obtaining 60 consecutive finger coordinate points. First, calculate the offset of each point relative to the starting point. , Eliminate differences in starting position; Find all of them again. The absolute maximum value | and The absolute maximum value of max|Δy| The offsets are normalized to an 8-bit integer range of 0-255 to obtain a normalized coordinate sequence. ; Next, each normalization point and Convert to an 8-bit binary string, concatenate them into a 16-bit substring, and then connect all the points to form a 960-bit trajectory binary string B; take the 128-bit temporary key K generated in step S2, and cyclically expand it to 960 bits to obtain... Finally, for and Bitwise XOR operation yields the enhanced key. This process increases the unpredictability of the key through the randomness of natural actions, thus preventing cracking based on fixed patterns. The normalization formula for coordinate offset is: ; ; In the formula: For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; For the first Each coordinate point relative to the starting point Directional offset; For all The maximum absolute value of the directional offset; This is the rounding function; The formula for cyclic expansion of temporary keys is: {K}_{ext}\left [ {j} \right ]=K\left [ {j\, mod\, len(K)} \right ]\, ; In the formula: {K}_{ext}\left [ {j} \right ] For the extended temporary key Bit; The temporary session key generated in step S2; The length of the temporary key; The bit index of the binary string; The formula for generating the enhanced key is: ; {K}_{strength}\left [ {j} \right ] For the strengthening key Bit; B\left [ {j} \right ] The first binary string after the motion trajectory conversion Bit; This is the bitwise XOR operator; {K}_{ext}\left [ {j} \right ] Consistent with the definition in the temporary key cyclic expansion formula.

5. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S4, after the user inputs encrypted content, the digital human segments the content into segments of 200 characters each, with any segment less than 200 characters being the last segment. Each segment is encrypted using the enhanced key generated in step S3 with AES-256 encryption in CBC mode. The initialization vector IV is generated by converting the ASCII codes of the first 16 characters of the segment and embedding the corresponding motion trajectory feature code. The feature code generation rule is the same as in step S3, that is, one feature point is taken from every 10 points in the motion trajectory coordinate sequence, converted into a 32-bit binary string, and embedded in the header of the ciphertext. After each ciphertext segment is generated, the SHA-256 hash value of that segment is calculated and stored together with the feature code for subsequent segment verification.

6. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S5, after the digital human sends three ciphertexts, it will confirm the details of the code from step S1 with the user. The question is generated based on the semantic structure of the code and selects the associated information of a keyword in the code. The user needs to answer the question by voice. The digital human verifies the accuracy of the answer through the speech recognition and semantic understanding modules. The accuracy rate must reach more than 90% before continuing to send subsequent ciphertexts. If the answer is wrong, the digital human will repeat the same question, up to 2 times. If it is still wrong, the interaction will be terminated.

7. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S6, after receiving the ciphertext, the receiving digital human first parses the feature code in the header of each ciphertext segment, converts the feature code into a coordinate sequence fragment, and then performs a matching verification with the action trajectory feature library stored in step S3. The matching algorithm adopts dynamic time warping, and the similarity threshold is set to 90%. The segments that pass the match are marked as valid, and the segments that fail are discarded directly. Then, the receiving party calls the temporary key template generated in step S2. This template contains voiceprint feature dimensions and hash value location information, and waits for the user to provide the action data required to reconstruct the reinforcement key. The entire verification process must be completed within 1 second after receiving the ciphertext to ensure the real-time nature of the interaction.

8. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S7, the receiving digital human guides the user to repeat the actions of step S3. The guidance uses natural speech to clearly inform the user of the action type. The motion capture uses the same frame rate as step S3 (30 frames / second) for 2 seconds, and collects 60 coordinate points. The method for converting the captured motion trajectory coordinate sequence into a binary string is the same as in step S3, that is, normalizing it to the range of 0-255 and then converting it to binary. Then, it is XORed with the temporary key template in step S2 to reconstruct the strengthening key. The reconstructed key is compared with the strengthening key generated in step S3. The similarity is calculated using Hamming distance. If the Hamming distance is less than 10, it is considered successful. Otherwise, the user is prompted to repeat the operation. A maximum of 3 attempts are allowed. If it still fails, the interaction is terminated. This step is related to the logic of steps S2 and S3 to ensure the accuracy of key reconstruction.

9. The digital human encrypted interaction method as described in claim 1, characterized in that: In step S8, the receiving digital human decrypts each segment of content using the reconstructed enhanced key. The decryption adopts the AES-256 CBC mode corresponding to step S4. The initialization vector IV is extracted from the ciphertext header. After each segment is decrypted, the SHA-256 hash value of that segment is calculated and compared with the hash value attached to the ciphertext. If they match completely, it is marked as complete. After all segments are decrypted and verified, the digital human reassembles the content in the original order and presents the decrypted content completely through text or voice. If the hash value of a certain segment does not match, the digital human will prompt that the content of that segment is corrupted and ask the user whether to receive the ciphertext again. The decryption process must ensure that the processing time of each segment does not exceed 30 milliseconds to ensure that the content is presented to the user quickly.

10. A digital human encrypted interaction system, characterized in that: The system executes the digital human encrypted interaction method as described in any one of claims 1 to 9 during operation.

Images