Single-server privacy information retrieval method and system for lightweight client
By generating ciphertext for queries on the server side, the problem of excessive client-side computational burden in traditional single-server PIR is solved, enabling lightweight privacy information retrieval that is suitable for mobile devices and browsers, reducing latency and computational burden, and supporting a variety of application scenarios.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-01-09
- Publication Date
- 2026-04-14
AI Technical Summary
In traditional single-server PIR schemes, the computational burden on the client is too heavy, especially when generating query ciphertext, which requires a large number of FHE encryption operations, resulting in high latency and unsuitability for mobile devices and browsers. Furthermore, the ciphertext cannot be reused, and it must be regenerated for each query, which increases the computational pressure.
By generating several pre-set ciphertexts and storing them on the server, the client decomposes the index into sub-indexes and randomly selects slot values to generate requests. The server then rotates and reconstructs the query vector, reducing client computation. By using one-hot encoding and offset calculation, query generation is completely transferred to the server side.
It significantly reduces the computational and communication burden on the client, has extremely low latency, small request size, supports mobile devices and browsers, database updates do not affect the pre-built encrypted pool, and the protocol security is suitable for mobile devices, medical data queries, and cloud privacy databases.
Smart Images
Figure CN121858631A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of privacy protection technology, and in particular to a lightweight client-side single-server privacy information retrieval method and system. Background Technology
[0002] This invention belongs to the field of privacy information retrieval (PIR) technology, and particularly relates to a system and method for protecting user query privacy in a single-server environment using homomorphic encryption technology.
[0003] In traditional single-server PIR schemes, fully homomorphic encryption (FHE) is typically required, necessitating numerous FHE encryption operations by the client when generating query vectors. For example, in SealPIR, OnionPIR, and XPIR, the client needs to generate O(√N) query ciphertexts, each containing high-dimensional polynomial coefficients. This results in a massive computational burden and latency typically ranging from 200ms to several seconds, making it unsuitable for mobile devices, browsers, or devices with limited computing power.
[0004] Furthermore, traditional PIRs often employ coefficient encoding, resulting in non-reusable ciphertext. Each query requires the generation of a new set of 0 / 1 ciphertext, further exacerbating the computational burden on the client. Summary of the Invention
[0005] In view of the above problems, this invention proposes a lightweight client-side single-server privacy information retrieval method and system, which enables the client's query generation to be completely transferred to the server for execution, thereby significantly reducing the client's computational and communication burden and providing structural conditions for parallel optimization on the server side.
[0006] To achieve the above objectives, the technical solution of the present invention includes the following:
[0007] A lightweight, single-server privacy information retrieval method using a client application, the method comprising: Several pre-generated ciphertexts are generated, sent, and stored on the server. When a user needs to query an index At that time, the index It is decomposed into several sub-indexes, and the index is generated by randomly selecting a slot value p of a pre-set ciphertext for each sub-index. Request So that the server can respond to the request Rotation to reconstruct the index Encrypted query vector Then, homomorphic computation is performed with the database to obtain encrypted query results without being able to predict the client's query intent; Decrypt the encrypted query result to obtain the index. The plaintext.
[0008] Furthermore, several pre-defined ciphertexts are generated, including: Randomly select slot ,in, To preset the number of slots in the encrypted message; Constructing plaintext vectors based on one-hot encoding , where the plaintext vector Middle slot The value of is 1, and the value of other slot positions is 0; plaintext vector Encryption is performed to obtain the pre-set ciphertext. .
[0009] Furthermore, a request corresponding to each sub-index is generated by randomly selecting a slot value p of a pre-defined ciphertext for each sub-index. ,include: Randomly select a pre-set ciphertext Among them, the pre-set encrypted text Middle slot The value of is 1, and the value of other slot positions is 0; Calculate offset ;in, Indicates the first Sub-indexes of each dimension, To preset the number of slots in the encrypted message; Generate this index Request .
[0010] Furthermore, according to the request Rotation to reconstruct the index Encrypted query vector ,include: Request Each sub-request included Perform rotation operations separately to obtain subqueries. The rotation operation includes: automorphism operation and key exchange operation. All subqueries Concatenate to form an encrypted query vector .
[0011] Furthermore, the encrypted query result ;in, Indicates an encryption operation. Indicates based on encrypted query vector The query results obtained from the database This represents the first generation generated after homomorphic plaintext encoding of the database. Homomorphic plaintext.
[0012] Furthermore, the database's organizational structure includes: data buckets - data blocks - homomorphic plaintext.
[0013] A lightweight client-based single-server privacy information retrieval system, the system comprising: The client is used to pre-generate several pre-set ciphertexts; when a user needs to query an index... At that time, the index It is decomposed into several sub-indexes, and the index is generated by randomly selecting a slot value p of a pre-set ciphertext for each sub-index. Request Decrypt the encrypted query results to obtain the index. The plaintext; The server is used to store pre-set ciphertext, according to the request. Reconstructing the index by rotating the pre-set ciphertext Encrypted query vector Then, homomorphic computation is performed with the database to obtain encrypted query results without being able to predict the client's query intent.
[0014] An electronic device includes: a processor and a memory storing computer program instructions; the processor, when executing the computer program instructions, implements the single-server privacy information retrieval method for a lightweight client as described above.
[0015] A computer-readable storage medium storing computer program instructions, which, when executed by a processor, implement the single-server privacy information retrieval method for a lightweight client as described above.
[0016] A computer program product, when run on a computer device, causes the computer device to execute the lightweight client single-server privacy information retrieval method described in any of the preceding claims.
[0017] Compared with the prior art, the present invention has at least the following beneficial effects.
[0018] 1. Zero encryption overhead on the client side: No encryption computation is required at all.
[0019] 2. Extremely low latency: Client-generated requests take approximately 0.01ms.
[0020] 3. Extremely small request size: approximately 0.02KB, which is 4-5 orders of magnitude smaller than common PIR systems.
[0021] 4. Supports mobile devices and browsers.
[0022] 5. Database updates do not affect the pre-configured ciphertext pool. Attached Figure Description
[0023] Figure 1 A flowchart of a single-server privacy information retrieval method for a lightweight client. Detailed Implementation
[0024] The system will now be described in further detail with reference to the accompanying drawings. The examples given are for illustrative purposes only and are not intended to limit the scope of the system.
[0025] This invention employs a two-stage PIR architecture of offline + online, enabling the server-side reconstruction of the query ciphertext through pre-set ciphertext (offline transmission) and offset (online transmission). The core idea is that the client no longer generates the query ciphertext, but instead uploads several one-hot pre-set ciphertexts in advance. The server then converts the pre-set ciphertexts into the query ciphertext through a rotation operation, thereby transferring the high-cost operation to the server.
[0026] Specifically, the lightweight client-side single-server privacy information retrieval method of the present invention, such as... Figure 1 As shown, it includes the following steps.
[0027] I. Offline Stage: The client generates pre-set ciphertext.
[0028] 1. The client randomly selects a slot. ,in The number of slots in a homomorphic ciphertext is typically selected from 2048, 4096, 8192, etc.
[0029] 2. Construct a one-hot plaintext vector v, where The rest are 0.
[0030] 3. Execute Encrypt(v) to obtain the pre-set ciphertext. Here, Encrypt represents the encryption operation.
[0031] 4. Record the number of each preset ciphertext. , and the corresponding position p to the index table TC.
[0032] 5. Upload pre-set encrypted text and its pre-set ciphertext number To the server index table .
[0033] It should be noted that the pre-set ciphertext is unrelated to any query and can be generated in batches of 500 to 10,000 records for long-term use.
[0034] II. Online Phase: Client-side lightweight offset request.
[0035] 1. User query index Decomposed into Sub-dimensions: .in, .
[0036] 2. The client randomly selects the k-th pre-set ciphertext, and its slot is... .
[0037] 3. Calculate the offset: .
[0038] 4. The final client generates a request: The request size is approximately 20 bytes, therefore it can be considered a very lightweight request. III. Online Phase: Server-side Rotation and Query Reconstruction.
[0039] Server completes: Subquery calculation , Indicates rotation operation. This indicates that ciphertext is selected from a preset ciphertext pool. This rotation operation... Depend on The structure consists of several parts, where automorphism represents the permutation of the ciphertext polynomial coefficients, and key-switch represents the transformation of the permutation result from the galois key field back to the original key field. Finally, all subqueries are processed. The concatenation results in the final ciphertext Q for the query.
[0040] IV. Online Phase: Server response calculation.
[0041] The server performs a homomorphic dot product and obtains the encrypted query result. , This represents the first generation generated after homomorphic plaintext encoding of the database. A homomorphic plaintext. Returns the encrypted query result to the client. The database is configured as follows: The structure is organized so that each block is aligned with a query vector slot, facilitating efficient homomorphic matching.
[0042] V. Online Phase: Client Decryption.
[0043] The client's job is to decrypt the query using its private key and then decode it according to pre-agreed encoding rules to obtain the final readable query answer. In this way, the server can only operate on the ciphertext throughout the process, and only the client possessing the private key can ultimately "see" the plaintext result, thus ensuring the privacy of the query result content.
[0044] In summary, because the Offset distribution of this invention is random and determined by... and It was jointly decided that the server could not derive the value from the offset. The one-hot location of the pre-prepared ciphertext is randomized, making it impossible for the server to distinguish the semantics of different pre-prepared ciphertexts. The IND-CPA security of the entire protocol relies on RLWE and is suitable for mobile encrypted retrieval, medical data querying, cloud privacy databases, IoT scenarios, etc.
[0045] Although specific embodiments of the system have been disclosed for illustrative purposes to aid in understanding and implementing the system, those skilled in the art will understand that various substitutions, variations, and modifications are possible without departing from the spirit and scope of the system and the appended claims. Therefore, the system should not be limited to the content disclosed in the preferred embodiments, and the scope of protection claimed by the system is determined by the scope defined in the claims.
Claims
1. A lightweight client-side single-server privacy information retrieval method, characterized in that, The method, which involves applying a client, includes: Several pre-defined ciphertexts are generated, sent, and stored on the server. When a user needs to query an index At that time, the index It is decomposed into several sub-indexes, and the index is generated by randomly selecting a slot value p of a pre-set ciphertext for each sub-index. Request So that the server can respond to the request Rotation to reconstruct the index Encrypted query vector Then, homomorphic computation is performed with the database to obtain encrypted query results without being able to predict the client's query intent; Decrypt the encrypted query result to obtain the index. The plaintext.
2. The method according to claim 1, characterized in that, Generate several pre-defined ciphertexts, including: Randomly select slot ,in, To preset the number of slots in the encrypted message; Constructing plaintext vectors based on one-hot encoding , where the plaintext vector Middle slot The value of is 1, and the value of other slot positions is 0; plaintext vector Encryption is performed to obtain the pre-set ciphertext. .
3. The method according to claim 1, characterized in that, The request corresponding to each sub-index is generated by randomly selecting a pre-defined ciphertext slot value p for each sub-index. ,include: Randomly select a pre-set ciphertext Among them, the pre-set encrypted text Middle slot The value of is 1, and the value of other slot positions is 0; Calculate offset ;in, Indicates the first Sub-indexes of each dimension, To preset the number of slots in the encrypted message; Generate this index Request .
4. The method according to claim 3, characterized in that, According to the request Rotation to reconstruct the index Encrypted query vector ,include: For the request Each sub-request included Perform rotation operations separately to obtain subqueries. The rotation operation includes: automorphism operation and key exchange operation; All subqueries Concatenate to form an encrypted query vector .
5. The method according to claim 4, characterized in that, The encrypted query result ;in, Indicates an encryption operation. Indicates based on encrypted query vector The query results obtained from the database This represents the first generation generated after homomorphic plaintext encoding of the database. Homomorphic plaintext.
6. The method according to claim 1, characterized in that, The database's organizational structure includes: data buckets - data blocks - homomorphic plaintext.
7. A lightweight client-side single-server privacy information retrieval system, characterized in that, The system includes: The client is used to pre-generate several pre-set ciphertexts; when a user needs to query an index... At that time, the index It is decomposed into several sub-indexes, and the index is generated by randomly selecting a slot value p of a pre-set ciphertext for each sub-index. Request Decrypt the encrypted query results to obtain the index. The plaintext; The server is used to store pre-set ciphertext, according to the request. Reconstructing the index by rotating the pre-set ciphertext Encrypted query vector Then, homomorphic computation is performed with the database to obtain encrypted query results without being able to predict the client's query intent.
8. An electronic device, characterized in that, The electronic device includes: a processor and a memory storing computer program instructions; when the processor executes the computer program instructions, it implements the single-server privacy information retrieval method for a lightweight client as described in any one of claims 1-6.
9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer program instructions, which, when executed by a processor, implement the single-server privacy information retrieval method for a lightweight client as described in any one of claims 1-6.
10. A computer program product, characterized in that, When the computer program product is run on a computer device, the computer device performs the single-server privacy information retrieval method for a lightweight client as described in any one of claims 1-6.