Encryption and decryption method and device for multi-modal data such as voice text

Abstract

The invention relates to the technical field of information security, in particular to a voice text and other multi-mode data encryption and decryption method and device. The method comprises the following steps: decomposing a medical record containing multiple data types into independent composition modes; for each mode, using a unique and disposable symmetric key to encrypt large-volume data of the mode through a symmetric algorithm; defining an independent attribute-based access strategy for each mode, and encrypting a corresponding symmetric key by using a ciphertext strategy based on an attribute-based encryption scheme; and packaging the symmetric ciphertexts of all the modes and the corresponding encrypted symmetric keys into a structured multi-mode encryption container. The apparatus includes a plurality of functional modules for performing encryption and decryption steps. According to the scheme, the user can only decrypt the symmetric key corresponding to the data mode of which the authority meets the access strategy by virtue of the attribute private key, so that the problems of coarse access control granularity, high calculation overhead and easiness in leakage of strategy privacy in the prior art can be effectively solved.

Description

Technical Field

[0001] This invention relates to the field of information security technology, specifically to a method and apparatus for encrypting and decrypting multimodal data such as voice and text. Background Technology

[0002] With the deep integration of information technology into the medical field, modern medicine increasingly relies on the comprehensive analysis of massive and heterogeneous medical data. This data exhibits typical multimodal characteristics; a complete electronic health record or personal health record usually contains multiple data types: structured text (such as medical records and lab reports), unstructured text (such as medical orders and progress notes), high-resolution medical images (such as MRI, CT scans, and X-rays), one-dimensional physiological signals (such as ECG and EEG), and real-time health monitoring data from wearable devices and medical IoT sensors (such as heart rate and blood pressure). This multimodal nature and high sensitivity of the data present unprecedented challenges to data security and privacy protection.

[0003] Existing data protection solutions reveal significant limitations when dealing with complex healthcare scenarios. Traditional encryption methods, such as encrypting the entire patient record file, while providing basic confidentiality, offer overly coarse-grained access control. In this "all or nothing" model, once a user obtains the decryption key, they can access all information in the record. This does not meet the needs of modern collaborative healthcare work. In clinical practice, different members of a multidisciplinary team need access to different parts of the same medical record: radiologists need to view imaging data but do not have access to the patient's mental health assessment; pharmacists need to view medication history but do not need access to the patient's raw genetic sequencing data; and billing staff only need access to information related to insurance and costs. Traditional encryption schemes cannot achieve this fine-grained access control based on role, responsibility, and data type.

[0004] To address the problem of fine-grained access control, the academic community has proposed attribute-based encryption (ABE) techniques, particularly ciphertext-policy attribute-based encryption (CP-ABE). In CP-ABE, data encryption is tied to an access policy, which consists of a series of attributes and their logical relationships. A user can only successfully decrypt the ciphertext if the set of attributes contained in their private key satisfies the access policy. This allows data owners to autonomously and flexibly control who can access which data without needing to know the identities of all potential accessors in advance.

[0005] However, despite the enormous potential of CP-ABE in theory, it still faces the following challenges when directly applied in real-world medical settings: High computational overhead: Most existing CP-ABE schemes are based on bilinear pairing operations, which are extremely computationally intensive. The computational complexity of encryption and decryption is usually proportional to the complexity of the access policy. This is acceptable on the server side, but it constitutes an insurmountable performance bottleneck for some resource-constrained devices.

[0006] Privacy breaches in access policies: In many standard CP-ABE schemes, access policies describing access permissions are stored and transmitted in plaintext along with ciphertext. This can lead to serious privacy breaches. Summary of the Invention

[0007] To address the problems in the prior art, this invention provides a method and apparatus for encrypting and decrypting multimodal data such as voice and text. This solution enables ultra-fine-grained access control at the modality level for complex medical records containing various data types such as text, images, and physiological signals, while significantly reducing the computational burden on terminal devices and protecting the privacy of access policies.

[0008] The technical solution of this invention mainly includes the following steps: Data modality decomposition: Decomposing a complete medical record containing multiple data types (e.g., a patient file containing medical history text, MRI images, and ECG signals) into its independent constituent modalities.

[0009] Hybrid encryption: An independent encryption process is performed on each decomposed data modality, specifically including: For each data modality (e.g.) , and Generate a unique, one-time symmetric encryption key, called the Data Encryption Key (DEK), for example... , and .

[0010] Using a symmetric cryptographic algorithm that supports Authentication, Encryption and Data Accreditation (AEAD) (such as AES-256-GCM), the data mode is encrypted using the corresponding DEK to generate symmetric ciphertext. .

[0011] The data owner defines a specific access policy for this data modality. (For example, The strategy is "role = radiologist and department = radiology department".

[0012] The CP-ABE scheme, based on elliptic curve cryptography (ECC), is used, employing the system public key (PK) and a modality-specific access policy. The corresponding DEK is encrypted to generate ABE ciphertext. .

[0013] Encryption encapsulation: symmetric ciphertext of all data modalities Its corresponding ABE ciphertext Along with the necessary metadata, it is encapsulated in a structured, parsable multimodal cryptographic container (MEC) file.

[0014] Decryption and Access: When a data user (such as a doctor) requests access to the MEC file, their client application uses the user's attribute private key SK to attempt to decrypt the ABE ciphertext of each data block in the MEC one by one. Only when the user's attribute set satisfies the access policy of a certain data block can the corresponding DEK be successfully decrypted. The DEK is then used to decrypt the symmetric ciphertext of that data block, ultimately obtaining the plaintext data. For data blocks that do not meet the policy, decryption will fail, and the user will not be able to access their content.

[0015] Another aspect of the present invention provides an apparatus for implementing the above-described method. This apparatus can be a standalone server, a personal computing device, or an embedded system. The apparatus integrates multiple functional modules, including: a decomposition module for separating multimodal medical records into independent data modalities; a key generation module responsible for generating a one-time symmetric encryption key for each data modality; a symmetric encryption module for high-performance encryption of each modality's data; an attribute encryption module for encrypting the symmetric key using the CP-ABE scheme; and an encapsulation module for assembling all encrypted components into a MEC container. These modules work together to implement the secure encryption process described in this invention.

[0016] Compared with the prior art, the advantages of this invention are: This invention implements "modal" access control within a single logical record. This allows users with different permissions to securely access different parts of the same medical record, which is well-suited to the collaborative work patterns of modern medicine.

[0017] Furthermore, by employing a hybrid encryption strategy, computationally intensive large file encryption tasks are handled by high-performance symmetric algorithms, while the computationally complex CP-ABE algorithm is used only for encrypting small DEKs. By replacing the traditional bilinear pairing scheme with the ECC-based CP-ABE scheme, the computational resource consumption on the client side is significantly reduced.

[0018] Furthermore, by embedding the access policy into the ABE ciphertext instead of transmitting it in plaintext, the risk of inferring sensitive patient information by analyzing the access policy can be prevented. Attached Figure Description

[0019] Figure 1 This is a flowchart of a method for encrypting and decrypting multimodal data such as voice and text proposed in this invention. Detailed Implementation

[0020] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0021] First, the overall system architecture of this solution will be described. This system architecture mainly includes the following four types of entities: Trusted Authorization Center: The trusted authorization center is responsible for system initialization, generation and management of public-private key pairs, and management of user attributes. In a centralized embodiment, there can be a single global trusted authorization center. However, in a preferred embodiment, a multi-authorization center architecture can be adopted, assigning different categories of attributes (such as organizational attributes and role attributes) to different authorization agencies to avoid single points of failure. The trusted authorization center executes the following algorithm: System initialization algorithm. The trusted authorization center takes a security parameter (e.g., key length) and a complete set of system-wide attributes as input, and outputs the system public key and master key. The system public key is public to all entities, while the master key is kept by the trusted authorization center.

[0022] User private key generation algorithm. When a user registers, the trusted authorization center assigns a set of attributes to them based on their identity and proof of identity. Then, the trusted authorization center uses the master key and the attribute set to generate a unique private key for that user. This private key, bound to its attributes, is the sole credential for decryption authorization.

[0023] Data Owner: The data owner is typically the patient or a medical institution authorized by the patient. The data owner holds the raw, unencrypted multimodal medical data and is responsible for defining the policies for accessing this data. During the encryption process, the data owner uses the system's public key and their own defined access policies to encrypt the data. The device used by the data owner, such as a personal computer or server, serves as the physical carrier of the encryption device in this invention.

[0024] Data users: The data users are entities that need to access medical data, such as doctors, nurses, and researchers. Each data user possesses a private key issued by the trusted authorization center and associated with their attribute set. They use this key to decrypt encrypted data within their authorized scope. The devices used by the data users, such as tablets or workstations, are the physical carriers of the decryption device in this invention.

[0025] Cloud Platform: The cloud platform is a third-party platform that provides data storage and computing services. This platform is used to execute storage and data transmission commands. It should be noted that, to protect user data privacy, all data stored on this platform must be strongly encrypted; the platform itself cannot decrypt any content of MEC files.

[0026] The present invention also provides an encryption device and a decryption device, which are physical entities for implementing the method of the present invention.

[0027] Encryption device: This device is operated by the data owner and is responsible for executing the encryption process. It includes a decomposition module, a key generation module, a symmetric encryption module, an attribute encryption module, and a packaging module. The decomposition module is used to receive a multimodal medical record and decompose it into independent, individually processable data modalities according to data type (such as text, image, signal).

[0028] The key generation module calls the cryptographically secure pseudo-random number generator for each data mode to generate a unique, one-time symmetric key (DEK).

[0029] The symmetric encryption module acquires each modal data and its corresponding DEK, and encrypts the data using a symmetric encryption algorithm (such as AES-GCM) to generate symmetric ciphertext.

[0030] The attribute encryption module receives the DEK corresponding to each modality and the access policy defined by the data owner, encrypts the DEK using the CP-ABE scheme and the system public key, and generates ABE ciphertext.

[0031] The encapsulation module collects the symmetric ciphertext and ABE ciphertext of all modalities, along with metadata, and assembles them into a single, structured data container file according to the MEC specification.

[0032] Decryption Device: This device is operated by the data user and is responsible for executing the decryption process. It includes a parsing module, an attribute decryption module, a symmetric decryption module, and a data presentation module. The parsing module is used to receive and parse the structure of the MEC file, and identify the various data blocks and their metadata contained therein.

[0033] The attribute decryption module is used to use the user's attribute private key to traverse each data block in the MEC and attempt to decrypt the ABE ciphertext in its block header.

[0034] The symmetric decryption module is used to decrypt the symmetric ciphertext in the corresponding data block using the DEK after the attribute decryption module successfully recovers the DEK of a certain data block, thereby recovering the plaintext data.

[0035] The data presentation module is used to display successfully decrypted plaintext data to the user, and to display a message indicating restricted access for data that the user does not have permission to access.

[0036] Reference Figure 1 The multimodal data encryption method proposed in this invention is described in detail. This method is completed collaboratively by various modules of the encryption device: S101. Receive multimodal data. The decomposition module of the encryption device receives a medical record belonging to the same patient to be encrypted from a data source (such as a HIS system or wearable device). This record contains multiple data modalities, such as medical record text. MRI images and ECG signal The decomposition module identifies and separates them.

[0037] S102, Traverse each data modality. The system iterates through each data component. Start a separate encryption subprocess.

[0038] S103, Generate a data encryption key. For the currently processed data mode... The key generation module calls a cryptographically secure pseudo-random number generator to generate a high-strength symmetric key. (For example, 256 bits).

[0039] By using a unique and one-time DEK for each data modality, cryptographic isolation between modalities is achieved. This means that even if a DEK is leaked, only the corresponding data modality will be compromised, without affecting the security of other parts of the record.

[0040] S104. Perform symmetric encryption. The symmetric encryption module uses... And the data modality of the preset symmetric encryption algorithm (such as AES-256-GCM) Encryption is performed to obtain symmetric ciphertext. And related encryption parameters.

[0041] For large volumes of medical data (such as images and signals), hardware-accelerated AES encryption is much faster than asymmetric encryption, which can greatly reduce the computational overhead and energy consumption of encryption devices, and is especially suitable for resource-constrained mobile devices.

[0042] S105. Define the access policy. The data owner is the current data modality. Define an access policy This strategy is represented in the form of an access tree and is passed to the attribute encryption module.

[0043] S106. Execute CP-ABE encryption. The attribute encryption module calls the CP-ABE encryption algorithm, using the system public key and access policy. For the newly generated Encryption is performed to obtain the ABE ciphertext. .

[0044] S107. Determine if all modalities have been processed. If there are still unprocessed data modalities, the process returns to step S102.

[0045] S108. Assemble the MEC file. This is done after all data modes have been encrypted and the corresponding... and Then, the encapsulation module assembles these components and global metadata (such as patient ID hashes, indexes, etc.) into a complete binary file.

[0046] S109. Store MEC. Upload the final generated MEC file to the cloud platform for storage. The encryption process is now complete.

[0047] Next, the multimodal data decryption method proposed in this invention will be described in detail. This method is completed collaboratively by various modules of the decryption device: S201. Receive access request. A data user (e.g., a cardiologist) wishes to access a patient's medical records. The user holds a private key SK issued by the trusted authorization center, containing their attribute set.

[0048] S202. Obtain the MEC file. Download the target patient's MEC file from the cloud platform.

[0049] S203. Parse the MEC and traverse the data blocks. The parsing module of the decryption device first parses the global header of the MEC to obtain the data block index. Based on the index, it is determined that the MEC file contains, for example, three data blocks: text, image, and signal. These data blocks are then processed one by one.

[0050] S204, Attempt to decrypt the block header. For the currently processed data block. The attribute decryption module extracts the ABE ciphertext from its block header. .

[0051] S205. Perform CP-ABE decryption. The attribute decryption module calls the CP-ABE decryption algorithm and attempts to decrypt using the user's private key SK. .

[0052] S206. Determine if decryption was successful. The CP-ABE decryption algorithm internally verifies whether the user's attribute set meets the requirements. Hidden access policies .

[0053] If the conditions are met, the algorithm will successfully return the plaintext. .

[0054] If the conditions are not met, the algorithm will fail and return an error flag.

[0055] S207. If DEK recovery is successful.

[0056] The symmetric decryption module then uses the data that was just recovered. The symmetric encryption parameters in the block header affect the block payload. Decryption is performed.

[0057] Decrypted plaintext data The data is presented to the user by the data presentation module. For example, displaying medical record text or rendering medical images on the screen.

[0058] S208. If DEK recovery fails, the user does not have access to the data block. The application displays "Access Restricted" to the user and continues processing the next data block.

[0059] S209. Determine if all data blocks have been processed. If there are still unprocessed data blocks, the process returns to step S204. Once all data blocks have been processed, the process ends.

[0060] Through this process, users will ultimately only be able to see all data modalities within their authorized scope.

[0061] To enable the present invention to be efficiently applied in resource-constrained environments such as mobile healthcare and the Internet of Things for healthcare, a preferred embodiment supporting computational outsourcing is further proposed. This embodiment introduces a computational proxy server with strong computing power, capable of assisting terminal devices in completing the most time-consuming cryptographic operations without acquiring any sensitive plaintext information. In this embodiment, the encryption and decryption processes can be partially outsourced, specifically including: Encrypted outsourcing: The client device (such as the patient's mobile phone) performs a lightweight "pre-encryption": For each DEK, the client does not directly perform full CP-ABE encryption, but instead generates a temporary secret value s, uses s to perform a simple encryption of the DEK, and applies the access policy based on s. The various attribute components are transformed.

[0062] The client sends this preprocessed data (encrypted DEK and transformed policy components) to the computation proxy server.

[0063] The computational proxy server performs the CP-ABE operation to generate the final ABE ciphertext. .

[0064] It should be noted that during this process, the proxy server can neither obtain the plaintext of the DEK nor tamper with the access policy attributes preset by the data owner.

[0065] After that, the proxy server will complete the calculation. It is returned to the client, which then completes the final MEC encapsulation.

[0066] Decryption outsourcing: When a user equipment needs to decrypt a data block, it does not perform a complete CP-ABE decryption locally. Instead, it uses its own private key to generate a specific ABE ciphertext. The "conversion key".

[0067] The device will use this conversion key and They are sent together to the computing proxy server.

[0068] The proxy server uses a conversion key pair Perform cryptographic operations to convert it into a "partially decrypted" intermediate ciphertext.

[0069] The proxy server returns the converted intermediate ciphertext to the user device.

[0070] The user equipment performs local operations to recover the plaintext DEK from the intermediate ciphertext.

[0071] By introducing a computational outsourcing mechanism, a large amount of cryptographic computation burden can be transferred from resource-constrained terminal devices to cloud proxy servers, greatly improving the system's response speed and availability, reducing the energy consumption of mobile devices, and making it possible to deploy complex access control systems based on ABE in real medical scenarios.

[0072] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0073] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to the specific implementations described. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A method for encrypting and decrypting multimodal data such as voice and text, characterized in that, Includes the following steps: Receive the patient's medical records and decompose the medical records into a first data modality and a second data modality; Generate a first data encryption key for the first data mode and a second data encryption key for the second data mode; The first data modality is encrypted using a symmetric encryption algorithm and the first data encryption key to generate a first symmetric ciphertext. The second data modality is encrypted using the symmetric encryption algorithm and the second data encryption key to generate a second symmetric ciphertext; Define a first access policy associated with the first data modality, and a second access policy associated with the second data modality, which is different from the first access policy, wherein the access policy is defined based on access attributes; The first data encryption key is encrypted using an attribute-based encryption scheme with a ciphertext policy, the system public key, and the first access policy to generate a first attribute-based encrypted ciphertext. The second data encryption key is encrypted using the attribute-based encryption scheme of the ciphertext policy, the system public key, and the second access policy to generate a second attribute-based encrypted ciphertext. The first symmetric ciphertext, the second symmetric ciphertext, the first attribute-based encrypted ciphertext, and the second attribute-based encrypted ciphertext are encapsulated into a single, structured data container.

2. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, The symmetric encryption algorithm is the AES encryption algorithm, and it adopts an authentication encryption additional data mode, such as the counter mode.

3. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, The step of encrypting the first data modality using a symmetric encryption algorithm and the first data encryption key to generate the first symmetric ciphertext specifically includes: Generate a unique initialization vector for the encryption operation of the first data mode; The first data modality is used as plaintext, along with the first data encryption key and the initialization vector, and input into the symmetric encryption algorithm to generate the first symmetric ciphertext. The first symmetric ciphertext contains encrypted data and an authentication tag, which is used to verify the integrity and authenticity of the data during decryption.

4. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, The attribute-based encryption scheme using the ciphertext policy, the system public key, and the first access policy are used to encrypt the first data encryption key to generate the first attribute-based encrypted ciphertext, specifically including: Generate a random root secret value; The first data encryption key is encrypted based on the system public key and the root secret value; The root secret value is recursively shared from top to bottom along the access tree of the first access strategy, thereby generating a corresponding encrypted secret shared value for each leaf node of the access tree; The encrypted first data encryption key and all encrypted secret shared values ​​are combined to form the first attribute-based encrypted ciphertext.

5. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, The data container includes a global header and one or more data blocks; the global header contains indexes pointing to the respective data blocks; each data block corresponds to a data modality and includes a block payload for storing the symmetric ciphertext of that modality, and a block header for storing the attribute-based cryptographic ciphertext of that modality.

6. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, It also includes decryption steps: Receive a request from a data user who holds an attribute-based encrypted private key for a ciphertext policy associated with their attribute set; Use the private key to decrypt the first attribute-based encrypted ciphertext and the second attribute-based encrypted ciphertext in the data container; When the user's attribute set satisfies the first access policy, the first attribute-based encrypted ciphertext is successfully decrypted to recover the first data encryption key, and then the first data encryption key is used to decrypt the first symmetric ciphertext. When the user's attribute set satisfies the second access policy, the second attribute-based encrypted ciphertext is successfully decrypted to recover the second data encryption key, and then the second data encryption key is used to decrypt the second symmetric ciphertext.

7. The method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, The first access policy and the second access policy are hidden in the first attribute-based encrypted ciphertext and the second attribute-based encrypted ciphertext, making it impossible to determine the content of the policy without possessing the corresponding valid private key.

8. A method for encrypting and decrypting multimodal data such as voice and text according to claim 1, characterized in that, It also includes outsourcing the computationally intensive operations of attribute-based encryption or decryption of ciphertext policies to a computing proxy server to reduce the computational overhead of end devices.

9. A device for encrypting and decrypting multimodal data such as voice and text, characterized in that, It includes a decomposition module, a key generation module, a symmetric encryption module, an attribute encryption module, and a packaging module; The decomposition module is used to decompose the medical record into a first data modality and a second data modality; The key generation module is used to generate a first data encryption key for the first data mode and a second data encryption key for the second data mode; The symmetric encryption module is used to encrypt the first data mode using a symmetric encryption algorithm and the first data encryption key to generate a first symmetric ciphertext, and to encrypt the second data mode using the symmetric encryption algorithm and the second data encryption key to generate a second symmetric ciphertext. The attribute encryption module is used to encrypt the first data encryption key using an attribute-based encryption scheme of the ciphertext policy, a system public key, and a first access policy associated with the first data modality to generate a first attribute-based encrypted ciphertext, and to encrypt the second data encryption key using the attribute-based encryption scheme of the ciphertext policy, a system public key, and a second access policy associated with the second data modality to generate a second attribute-based encrypted ciphertext. The encapsulation module is used to encapsulate the first symmetric ciphertext, the second symmetric ciphertext, the first attribute-based encrypted ciphertext, and the second attribute-based encrypted ciphertext into a single, structured data container.

10. A computer-readable storage medium storing computer instructions thereon, characterized in that, When the instruction is executed, it implements the method for encrypting and decrypting multimodal data such as voice and text as described in any one of claims 1 to 8.

Images