Secure multi-party computing-based distributed encryption machine key management method and system

Abstract

The invention relates to the field of network security and cryptography, and discloses a scheme for realizing distributed generation, storage, collaborative derivation and verification of keys in a distributed encryption machine cluster by using secure multi-party computing (MPC). The method is applied to a computing cluster composed of n peer-to-peer encryptors configured with device unique identifiers. In the initialization stage, identifiers of all devices are mapped into abscissas to generate key fragments, and a verifiable secret sharing technology is utilized to shield fault nodes. And each encryption machine performs homomorphic summation on the legal sub-fragments to generate a unique root key fragment physically bound with local hardware. In the derivation stage, multiple machines dynamically calculate interpolation coefficients through blind processing in combination with device identifiers participating in calculation, and collaboratively derive consistent session keys on the premise that a complete root key is not restored. According to the method, deep anchoring of an algorithm and physical hardware is realized, node cloning attacks are effectively prevented, single-point fault risks are eliminated, and it is ensured that network transmission data cannot be decoded by a man-in-the-middle.

Description

Technical Field

[0001] This invention relates to the fields of network security and cryptography, specifically to a scheme for implementing distributed key generation, storage, collaborative derivation, and verification in a distributed encryption machine cluster using secure multi-party computation (MPC). Background Technology

[0002] Traditional key management architectures typically employ centralized storage, with keys stored in a central server (KMS) in plaintext or single-machine encrypted form. This architecture has the following drawbacks: (1) Single point of failure risk: Once the central server is compromised, the encryption keys of all downlinks will be at risk of being leaked. (2) Internal threats: High-privilege operation and maintenance personnel may have direct access to the plaintext of the key. (3) Distribution risk: Even with encryption tunnel protection, there is still a risk of protocol-level cracking during the distribution of the key to the encryption machine. (4) Computing power avalanche effect: In a cluster environment, the traditional centralized synchronization mechanism is prone to single-point performance bottlenecks when faced with large-scale concurrent requests, which can lead to system paralysis. Summary of the Invention

[0003] This invention proposes a peer-to-peer encryption machine key collaborative management system based on an MPC architecture. The core innovations include: (1) Distributed Key Generation (DKG): The key exists in fragments from the moment it is generated, and no single encryption machine possesses the complete key. (2) Built-in Byzantine fault-tolerant verification (VSS): Using verifiable secret sharing technology, faulty encryption machine nodes that provide erroneous data can be identified and shielded in real time. (3) Collaborative derivation mechanism: Through blinding processing, multiple encryption machines can collaboratively derive a consistent session key without restoring the root key. (4) Coordinate mapping mechanism based on hardware fingerprint: abandoning the traditional polynomial sequential allocation of coordinates, the physical device unique identifier (Device ID) of each encryption machine is mapped to the horizontal coordinate point in a finite field through a hash algorithm. The beneficial effects of this invention are as follows: (1) The root key is never stored on the ground: The root key is stored in fragments in the security modules of each encryption machine, which realizes the security isolation between the physical level and the logical level. (2) Decentralized peer-to-peer architecture: eliminates the risk of traditional single point of failure and improves the cluster's resilience. (3) Anti-interception capability: Through blinding and deblinding mechanisms, it ensures that the network transmission data in the derivation process cannot be deciphered by man-in-the-middle. (4) Deep anchoring of algorithm and physical hardware: Since the generation and derivation of root key fragments strongly depend on the physical device ID of the encryption machine itself, even if an attacker steals the fragment data in memory, the key cannot be restored without a valid hardware device ID participating in the Lagrange interpolation calculation. This achieves dual anti-counterfeiting measures of "algorithm-hardware". (See attached diagram for details.) Figure 1 This is a flowchart of a distributed encryption machine key management method based on secure multi-party computation according to the present invention. Figure 2 A flowchart of the collaborative derivation process of a single encryption machine's internal key provided in this embodiment of the invention. Figure 3 The complete key collaborative derivation timing diagram provided in the embodiments of the present invention. Detailed Implementation Example 1: Distributed Root Key Initialization (1) Cluster configuration: The encryption machine cluster is configured to consist of n peer nodes, denoted as EM1 to EM2. n Let the threshold value be t (1 < t ≤ n), and let any encryption machine in the cluster be denoted as EM. i (1 ≤ i ≤ n), and each encryption machine has a unique identifier (denoted as DID) from the factory. i Each node in the cluster publicly discloses its own DID in advance. (2) Polynomial generation: Arbitrary encryption machine EM i Locally generate t-1 order polynomials: f i (x) =a i + a (i,1) x + a (i,2) x 2 +…+a (i,t-1) x t-1 Where the constant term a i It is made by EM i A secret seed that is generated and contributed independently, a (i,1) to a (i,t-1) It is EM i Randomly generated coefficients. (3) Fragment calculation and distribution based on hardware ID: Encryption machine EM i Use hash function H to hash the EM of each node in the cluster j The device ID (1 ≤ j ≤ n) is mapped to the x-coordinate in a finite field. j = H(DID j ), and calculate the EM provided to each node. j The sub-fragment f i (x jThis is then distributed to ensure that the fragments acquired by each node are vertically aligned on the mathematical coordinate axis. (Note: When j=i, it is EM) i (Less fragments are left for local use). Each encryption machine performs this distribution operation synchronously and broadcasts coefficient commitments. (4) Commitment verification: The receiving encryption machine verifies whether the received sub-fragments are on the correct curve based on the broadcast elliptic curve mapping value (commitment), and identifies and blocks Byzantine fault nodes that provide erroneous data in real time. Example 2: Local Root Key Sharding Aggregation (1) Fragment collection: After verification, any encryption machine EM i The set of legitimate sub-fragments {f1(x)} sent by each node in the cluster to itself based on its own hardware identifier mapping will be collected. i ), f2(x i ),…,f n (x i )}. (2) Summation calculation: Encryption machine EM i Homomorphically sum the local and received legitimate sub-fragments: Summation calculation: Encryption machine EM i Homomorphically sum the local and received valid sub-fragments: (3) Application of results: S i That is, the encryption machine EM i The final locally unique root key fragment, used as the private input for subsequent derivation operations, is permanently anchored in the mathematical coordinate system to the physical hardware feature x of the encryption machine. i = H(DID i On the other hand, physical-level binding of data and hardware is achieved. Example 3: Blinding Collaborative Derivation Process When the business side requires a session key: (1) Blinding request: The encryption machine that initiates the request generates the elliptic curve point P after the session identifier is mapped, and randomly selects the blinding factor k, and sends P' = k * P to other encryption machines. (2) Homomorphic computation: The encryption machines involved in the computation (at least t responding) use their own root key to fragment and return the blinded derived component W. j ' = S j * P'. At this point, since P' is blinded, the computing nodes cannot know the original session identifier, nor can they deduce the root key fragment. (3) Hardware coordinate-based deblinding synthesis: The request initiator uses the inverse of k to execute W. j = k -1 * W j' =S j * P, and finally, based on the device IDs of the t encryption machines involved in the calculation, the session key is obtained by aggregating and calculating the hash using the Lagrange interpolation method.

Claims

1. A distributed encryption machine key management method based on secure multi-party computation, characterized in that, The method is applied to a computing cluster consisting of n peer-to-peer encryption machines, each configured with a unique device identifier. Includes the following steps: S1: Distributed root key initialization and distribution: Each encryption machine locally generates a t-1 order random polynomial f i (x) maps the device unique identifier of each encryption machine in the cluster to the horizontal axis, calculates and distributes sub-fragments for other encryption machines in the cluster, and broadcasts the coefficient commitment. S2: Commitment-based Byzantine fault-tolerant verification: The receiver's encryption machine verifies the legitimacy of sub-fragments based on broadcast commitments, and identifies and blocks faulty nodes in real time; S3: Local Root Key Fragment Synthesis: Each encryption machine performs a homomorphic summation on its local and received legitimate sub-fragments to generate a unique root key fragment S physically bound to the local device's unique identifier. i ; S4: Session Key Cooperative Derivation: The requesting machine generates and blinds the session identifier. At least t cryptographic machines use their respective root key fragments to return blinded derived components. The requesting machine combines the device unique identifiers of each participating cryptographic machine to perform deblinding aggregation hashing to generate the session key.

2. The method according to claim 1, characterized in that: In S1, the encryption machine EM i Send to encryption machine EM j The sub-fragment is f i (x j ), where x j For encryption machine EM j The device's unique identifier is hashed to obtain the horizontal coordinate value, thereby ensuring that the fragments obtained by each node are vertically aligned on a mathematical coordinate axis based on hardware characteristics.

3. The method according to claim 1, characterized in that: The blinding process includes generating a random blinding factor k by the encryption machine, and performing a scalar multiplication operation on the elliptic curve point P mapped by the session identifier using k to obtain the blinding request value P' = k * P.

Images