Financial data encryption transmission and storage method based on cloud computing

Through hierarchical sharding processing, dual-channel encryption and distributed key management, the security blind spots and single point failure problems of financial data transmission and storage in the cloud computing environment are solved, full-link dynamic encryption and active defense are achieved, and the security and response speed of financial data in the cloud computing environment are improved.

CN120710759APending Publication Date: 2025-09-26BEIJING CREDIT MANAGEMENT CO LTD

Patent Information

Application Number
CN202510934082.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2025-07-07
Publication Date
2025-09-26

AI Technical Summary

Technical Problem

Existing technologies in the transmission and storage of financial data in cloud computing environments have protection blind spots caused by single-dimensional encryption, single point failure problems in centralized key management, and static defense that is difficult to cope with new attacks. They cannot meet the security and response speed requirements of financial regulators.

Method used

A combination of hierarchical sharding processing, dual-channel asymmetric encryption engine, heterogeneous multi-cloud routing module and distributed key management matrix is ​​adopted. Through dynamically updated asymmetric encryption algorithms, homomorphic encryption algorithms that support ciphertext operations, distributed key management matrix and intelligent routing mechanisms, full-link dynamic encryption and active defense are achieved.

Benefits of technology

It achieves full-link dynamic encryption, eliminates security blind spots, reduces the risk of plaintext exposure, reduces the impact of key leakage, improves the defense capability against new attacks, and meets the security and response speed requirements of the financial industry.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure BDA0005487308360000151
    Figure BDA0005487308360000151
  • Figure BDA0005487308360000161
    Figure BDA0005487308360000161
  • Figure BDA0005487308360000181
    Figure BDA0005487308360000181
Patent Text Reader

Abstract

The invention relates to the technical field of cloud computing financial security, and provides a financial data encryption transmission and storage method. The method is characterized by comprising the following steps of: executing sensitivity-driven data grading fragmentation on a user terminal; dynamic elliptic curve encryption is carried out on transmission data through a two-channel encryption engine, and fully homomorphic encryption is carried out on storage data; dynamically distributing the fragments to heterogeneous cloud nodes by the multi-cloud routing based on reinforcement learning; a distributed key management matrix is constructed, key fragments are dispersed and stored in a block chain and a hardware security module, and reconstruction is activated through biological characteristics. The method has the advantages that full-link ciphertext operation is realized, and the plaintext exposure risk is eliminated; ciphertext state financial calculation is supported; single point failure is resisted; the APT attack is defended dynamically; and the quantum security evolution capability is realized. The method is suitable for mobile banks, cross-border payment and other scenes.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of cloud computing financial security technology, and more specifically, to a method for encrypting, transmitting, and storing financial data based on cloud computing. Background Art

[0002] As the financial industry accelerates its migration to the cloud, the cloud adoption rate of global financial institutions will reach 68% in 2023 (IDC data). However, data security incidents in cloud environments have surged by 42% year-on-year, and existing technologies still have certain limitations:

[0003] 1. Single-dimensional encryption leads to protection blind spots

[0004] Existing solutions, such as CN113114689A, focus solely on transmission link encryption, ignoring the security of stored data. Financial services require cloud-based risk assessments and other computations. Traditional AES encryption forces data to be decrypted before computation, creating a "plaintext exposure window" (lasting an average of 17.3 seconds), increasing the likelihood of exploitation by APT attacks by 83%.

[0005] 2. Centralized key management has a single point of failure

[0006] Typical solutions, such as US20220141050A1, use a single KMS (key management system). This can lead to the complete disclosure of all keys if the cloud service provider is attacked through the supply chain (such as the SolarWinds incident). Tests have shown that centralized architectures experience an average response delay of 9.2 minutes in the event of a zero-day attack, failing to meet the three-minute circuit breaker standard required by financial regulations.

[0007] 3. Static defenses are difficult to deal with new attacks

[0008] Current mainstream technologies, such as CN114244515A, rely on fixed encryption protocols and lack the ability to dynamically adjust to protocol vulnerability attacks (e.g., Log4j). When financial data is stored across multiple clouds, attackers have a 61% success rate in restoring datasets through correlation analysis (MITRE ATT&CK testing). Existing obfuscation techniques only increase attackers' time costs by 15%.

[0009] Therefore, to address the above problems, a financial data encryption transmission and storage method based on cloud computing is proposed. Summary of the Invention

[0010] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a financial data encryption transmission and storage method based on cloud computing to solve the problems raised in the above-mentioned background technology.

[0011] To achieve the above objectives, the present invention provides the following technical solution: a method for encrypting, transmitting, and storing financial data based on cloud computing, comprising the following steps:

[0012] Data pre - processing stage: Perform hierarchical sharding processing on the original financial data at the user terminal, and divide the data into shard sets of different security levels according to the preset data sensitivity classification rules;

[0013] Parallel encryption stage: Differentially process the shard set through a dual - channel asymmetric encryption engine. Among them, the transmission channel uses an asymmetric encryption algorithm with dynamic updates to protect the data transmission process, and the storage channel uses a homomorphic encryption algorithm supporting ciphertext operations to process the data to be stored;

[0014] Multi - cloud distribution stage: Dynamically distribute the encrypted data shards to the storage nodes of at least three independent cloud service providers through a heterogeneous multi - cloud routing module. This module evaluates the security status of each cloud node in real - time and optimizes the routing path;

[0015] Key management stage: Build a distributed key management matrix in the cloud, split the main decryption key into N mathematical fragments, and store them dispersedly in the blockchain network and the user - side hardware security module. Only when M fragments (M < N) are obtained can the key be reconstructed.

[0016] Preferably, the hierarchical sharding processing includes a sensitivity recognition sub - step, a redundancy encoding sub - step, and an anti - tampering marking sub - step. The sensitivity recognition sub - step analyzes the financial data structure through a trained neural network model and automatically labels the sensitive levels of three types of data: transaction amount, account information, and identity credentials. The redundancy encoding sub - step applies erasure coding technology to high - sensitivity - level shards to generate check shards, ensuring that the original data can still be fully restored when no more than the preset threshold of shards are lost. The anti - tampering marking sub - step attaches a digital signature label containing a timestamp, geographical coordinates, and shard hash value to each shard, forming a traceable shard metadata chain.

[0017] Preferably, the dual - channel asymmetric encryption engine includes: The transmission - channel encryption unit uses an asymmetric encryption algorithm based on elliptic curves, updates the key pair every 120 seconds in cooperation with a quantum random number generator, and retains the previous generation of keys for 60 seconds after each key update to process the legacy data in transmission. The storage - channel encryption unit constructs a hybrid architecture of fully homomorphic encryption and attribute - based encryption, enabling the cloud storage node to perform balance aggregation and risk - value calculation operations without decrypting, and only returning the decryption result when the user attributes meet the preset policy.

[0018] Preferably, the heterogeneous multi-cloud routing module executes a security assessment submodule, an intelligent routing submodule and a protocol mimicry submodule: the security assessment submodule continuously collects the compliance certification status, historical attack records, and real-time CPU load rate indicators of each cloud service node to generate a dynamic security score; the intelligent routing submodule analyzes the current network threat situation based on a reinforcement learning model, and preferentially routes the highest security level shards to the top three cloud nodes in the score ranking; the protocol mimicry submodule periodically switches TLS1.3, QUIC and private encryption protocols at the data transmission layer, and keeps the two protocols running in parallel for 10 seconds each time to ensure a seamless transition.

[0019] Preferably, the distributed key management matrix implements a biometric binding mechanism, an active defense mechanism, and an audit tracking mechanism. The biometric binding mechanism requires verification of the user's biometrics (iris / fingerprint) during the key fragment reconstruction process, and the biometric template is only stored in the user's local security chip. When a brute force cracking attempt is detected at any key fragment storage point, the active defense mechanism automatically triggers the self-destruction of associated fragments within 200 kilometers of the point. The audit tracking mechanism records all key operations in the alliance blockchain, and the legitimacy of the operation must be verified by three independent audit nodes before being written into an unalterable ledger.

[0020] Preferably, it also includes a dynamic encrypted tunnel switching mechanism: dual tunnels are established between the user terminal and the cloud entry gateway to synchronously create a main tunnel (SM9 algorithm) and a backup tunnel (NTRU algorithm), and intelligent switching control monitors the transmission quality of the main tunnel in real time through dedicated probes. When the packet loss rate exceeds 15% for five consecutive times or the delay suddenly increases by 300%, it automatically switches to the backup tunnel. The state seamless migration uses memory mirroring technology to synchronize the encrypted session state to the backup tunnel in real time to ensure that the switching process does not interrupt business interaction.

[0021] Preferably, the cloud storage layer implements: shard association obfuscation injects pseudo-association identifiers into data shards stored in different cloud platforms, so that external attackers cannot infer data ownership through shard metadata; active defense verification sends simulated attack data packets to storage nodes every 24 hours, and updates their security scores based on node response delays and error rates; ciphertext behavior monitoring analyzes ciphertext access logs through long short-term memory networks (LSTMs), and triggers alarms when unconventional time access and high-frequency abnormal request patterns are detected.

[0022] Preferably, the key lifecycle management system includes a timed rotation unit, a forward security unit and a geo-fence unit. The timed rotation unit adopts a clock service based on a hardware trusted execution environment (TEE) to generate new keys every 72 hours and re-encrypt the affected data fragments. The forward security unit generates a key chain through a key derivation function tree to ensure that the new key is not affected by the leakage of historical keys. The geo-fence unit automatically freezes the activation authority of the relevant key fragments when the user terminal leaves the preset geo-fence area for more than 30 minutes.

[0023] The technical effects and advantages of the present invention are as follows:

[0024] 1. Achieve full-link dynamic encryption to eliminate security blind spots

[0025] To address the "plaintext exposure window" vulnerability in financial data processing, this solution utilizes a collaborative mechanism of dynamic ECC encryption in the transmission channel and fully homomorphic encryption in the storage channel to ensure that data remains encrypted throughout transmission, storage, and cloud computing. Field tests have shown that the latency of ciphertext operations, such as risk assessment, increases by only 18ms (compared to plaintext operations), completely eliminating the 17.3-second plaintext exposure risk associated with traditional solutions. This solution supports the hybrid deployment of the SM4 / SM9 national encryption algorithms and FHE, meeting the requirements of Section 9.3 of the "Financial Data Security Guidelines" (JR / T0197-2020).

[0026] 2. Build a resistant key management system to eliminate single points of failure

[0027] To overcome the response delay issue of centralized KMS, a distributed key matrix of blockchain + HSM is adopted: key fragments are stored in three geographically isolated areas, and a self-destruct mechanism compresses attack response time to within 200ms. Biometric binding reduces the success rate of unauthorized access to 0.003%. Compared with traditional KMS solutions, the scope of key leakage is reduced by 98%, meeting the financial industry's 3-minute circuit breaker standard (the measured average circuit breaker time is 47 seconds).

[0028] 3. Establish an intelligent active defense network to dynamically respond to threats

[0029] Breaking through the limitations of static defense, innovative integration: Protocol mimicry defense switches between TLS / QUIC / private protocols every 90 seconds; cross-cloud shard obfuscation injects pseudo-correlation identifiers to reduce the success rate of dataset restoration from 61% to 4.7%; and active attack sandbox simulates APT attacks every 24 hours to verify node resilience. Effectiveness verification: The success rate of defending against Log4j-type protocol vulnerability attacks increased to 99.2%, while the attacker's correlation analysis cost increased from a 15% time increase to a 300% additional cost.

[0030] 4. Provide quantum security evolution capabilities for future-oriented protection

[0031] Proactively addressing quantum computing threats: Dual-tunnel quantum backup: primary tunnel (SM9) + backup tunnel (NTRU quantum-resistant algorithm); forward-secure key tree: a single key leak only impacts data within 72 hours; dynamic key updates: quantum random key rotation every 120 seconds. Verified by NIST testing standards: resistant to quantum attacks such as Shor's algorithm until at least 2040, reducing the risk of quantum cracking by 89% compared to traditional RSA schemes. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Figure 1 This is a system framework diagram of the present invention. DETAILED DESCRIPTION

[0033] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0034] As attached Figure 1 As shown, (1) a financial data encryption transmission and storage method based on cloud computing includes the following steps:

[0035] Data preprocessing stage: Perform hierarchical sharding on the original financial data at the user terminal, dividing the data into shard sets with different security levels according to preset data sensitivity classification rules;

[0036] Parallel encryption stage: A dual-channel asymmetric encryption engine is used to perform differentiated processing on the shard sets. The transmission channel uses a dynamically updated asymmetric encryption algorithm to protect the data transmission process, and the storage channel uses a homomorphic encryption algorithm that supports ciphertext operations to process the data to be stored.

[0037] Multi-cloud distribution phase: The encrypted data shards are dynamically distributed to the storage nodes of at least three independent cloud service providers through the heterogeneous multi-cloud routing module. This module evaluates the security status of each cloud node in real time and optimizes the routing path.

[0038] Key Management Phase: Build a distributed key management matrix in the cloud, split the master decryption key into N mathematical fragments, and store them dispersedly in the blockchain network and the user-side hardware security module. The key can be reconstructed only when M fragments (M < N) are obtained. Among them, deploy a lightweight sharding engine on the user terminal. First, use a pre-trained ResNet-18 model to classify the sensitivity of the input financial data: the transaction amount is automatically marked as level 5 (highest risk), the account information is marked as level 3, and the product information is marked as level 1. Subsequently, use an improved Reed-Solomon encoder to perform sharding. Level 5 data generates 6 shards (4 data shards + 2 redundant shards), level 3 generates 4 shards (3 + 1), and level 1 remains the original shard. Each shard is attached with a geographical stamp (±5-meter accuracy) generated by the GPS module and an NTP synchronized timestamp. The shard set is processed by a dual-channel encryption engine: the transmission channel is dynamically encrypted every 120 seconds by the ECC module integrated with a quantum random number chip, and the storage channel uses the Microsoft SEAL library to implement BFV fully homomorphic encryption. The encrypted shards are dynamically allocated to Alibaba Cloud, AWS, and Azure nodes by the reinforcement learning routing module deployed in the SD-WAN controller according to the real-time security score. The key management system uses the Shamir secret sharing algorithm to split the master key into 7 fragments, 5 of which can be reconstructed. Among them, 3 fragments are stored in the Hyperledger Fabric blockchain smart contract, and 4 fragments are stored in the user's USB HSM device and are activated and called through fingerprints.

[0039] (2) The hierarchical sharding process includes a sensitivity identification sub-step, a redundant coding sub-step, and a tamper-proof marking sub-step. The sensitivity identification sub-step analyzes the financial data structure through a trained neural network model and automatically marks the sensitive levels of three types of data: transaction amount, account information, and identity credentials. The redundant coding sub-step applies erasure code technology to high-sensitivity level shards to generate check shards, ensuring that the original data can still be completely restored when no more than the preset threshold of shards are lost. The tamper-proof marking sub-step attaches a digital signature label containing a timestamp, geographical coordinates, and shard hash value to each shard, forming a traceable shard metadata chain. Among them, the sensitivity identification model is deployed on the mobile terminal based on the TensorFlow Lite framework. The input layer receives transaction data in JSON format, and the convolutional layer extracts field features (e.g., the "amount" field automatically triggers a level 5 mark). The sharding engine calls the Jerasure library to perform erasure code encoding, configures a (4, 2) encoding scheme for level 5 data, and allows recovery through Lagrange interpolation when any 2 shards are lost. Each shard metadata packet contains a SHA3-256 hash value, Beidou satellite positioning coordinates (WGS84 coordinate system), and an ISO 8601 timestamp provided by the National Time Service Center of the Chinese Academy of Sciences. The SM2 digital signature algorithm is used to sign the metadata packet, and the private key is stored in the terminal SE security chip, and the public key is pre-set in the cloud verification service.

[0040] (3) The dual-channel asymmetric encryption engine includes: the transmission channel encryption unit adopts an asymmetric encryption algorithm based on elliptic curves, cooperates with the quantum random number generator to update the key pair every 120 seconds, and retains the previous generation key for 60 seconds after each key update to process the legacy data in transmission; the storage channel encryption unit constructs a hybrid architecture of fully homomorphic encryption and attribute-based encryption, so that the cloud storage node can perform balance aggregation and risk value calculation operations without decryption, and returns the decryption result only when the user attributes meet the preset policy. Among them, the transmission channel encryption adopts the OpenSSL engine integrated secp384r1 elliptic curve The system uses a curve and cooperates with the IDQ QRNG250 chip to generate true random numbers. The key update cycle is precisely controlled by a hardware timer at 120±0.5 seconds. The old key is retained in memory for 60 seconds to handle legacy data packets in transmission. The storage channel uses the Microsoft SEAL library version 4.1, configured with a polynomial ring dimension n=8192 and a plaintext modulus t=1032193, to implement homomorphic addition operations for balance summation. Attribute-based encryption uses the CP-ABE scheme. The policy compiler converts the logical expression "department=Finance Department AND position level ≥5" into an access tree structure. Private key generation is performed by the cloud-based HSM module.

[0041] (4) The heterogeneous multi-cloud routing module executes a security assessment submodule, an intelligent routing submodule, and a protocol mimicry submodule: the security assessment submodule continuously collects the compliance certification status, historical attack records, and real-time CPU load rate indicators of each cloud service node to generate a dynamic security score; the intelligent routing submodule analyzes the current network threat situation based on a reinforcement learning model and preferentially routes the highest security level shards to the top three cloud nodes in the score ranking; the protocol mimicry submodule periodically switches TLS1.3, QUIC, and private encryption protocols at the data transmission layer, and keeps the two protocols running in parallel for 10 seconds each time the switch is made to ensure a seamless transition; wherein, the security score module obtains the cloud node status through the REST API every 30 seconds: extracts the compliance score (0-100 points) from AWS Trust Advisor, and extracts the compliance score (0-100 points) from Cloudflare. Radar obtains the DDoS attack frequency (times / hour) and Prometheus collects the CPU load rate (%). Reinforcement learning routing is based on the DQN algorithm implemented in TensorFlow, and the reward function is designed to be R = 0.6 × security score + 0.3 × bandwidth utilization - 0.1 × latency. The protocol mimicry engine is implemented in the Linux kernel layer, switching transport protocols through the Netfilter framework. TLS1.3 uses the BoringSSL library, and the private protocol is customized based on the ChaCha20-Poly1305 algorithm. The TCP SA mechanism is used to maintain session state synchronization during protocol switching.

[0042] (5) The distributed key management matrix implements a biometric binding mechanism, an active defense mechanism, and an audit tracking mechanism. The biometric binding mechanism requires the user's biometrics (iris / fingerprint) to be verified during the key fragment reconstruction process, and the biometric template is only stored in the user's local security chip. The active defense mechanism automatically triggers the self-destruction of associated fragments within 200 kilometers of any key fragment storage point when a brute force cracking attempt is detected. The audit tracking mechanism records all key operations in the alliance blockchain and is written into an unalterable ledger after being verified by three independent audit nodes for the legitimacy of the operation. Among them, the key fragment reconstruction terminal integrates the Suprema BioMini fingerprint sensor, and the collected fingerprint template is encrypted with AES-256 and stored only in the local security chip; the fragment self-destruction mechanism is based on geo-fencing technology. When any blockchain node detects an abnormal login, it notifies other nodes within a radius of 200 kilometers in real time through WebSocket, triggering the physical self-destruction circuit of the HSM device to fuse the storage chip; the audit tracking system is deployed on the Fabric blockchain. The smart contract requires three verification nodes from financial institutions, regulatory agencies, and cloud service providers to perform BFT consensus verification on key operations, and all operation records are generated into Merkle tree evidence.

[0043] (6) It also includes a dynamic encryption tunnel switching mechanism: dual tunnels are established between the user terminal and the cloud entry gateway to synchronously create a main tunnel (SM9 algorithm) and a backup tunnel (NTRU algorithm). Intelligent switching control uses a dedicated probe to monitor the transmission quality of the main tunnel in real time. When the packet loss rate exceeds 15% for five consecutive times or the delay suddenly increases by 300%, it automatically switches to the backup tunnel. The state seamless migration uses memory mirroring technology to synchronize the encrypted session state to the backup tunnel in real time to ensure that the switching process does not interrupt business interaction. Among them, the main tunnel uses the SM9 algorithm certified by the National Cryptography Administration and generates the master key through the OP-TEE trusted execution environment; the backup tunnel implements the NTRU-HRSS-701 quantum-resistant algorithm based on the libntru library; the link probe sends a 128-byte probe packet every 500ms. When the packet loss rate is detected to be greater than 15% or the delay is greater than 800ms in five consecutive windows, the routing controller immediately switches the tunnel; session migration is implemented through the DTLS state machine in shared memory, and atomic operations are used to ensure that the message sequence number is not lost during the switching process. The measured switching delay is ≤37ms.

[0044] (7) Implementation of cloud storage layer: Shard association obfuscation injects pseudo-association identifiers into data shards stored in different cloud platforms, so that external attackers cannot infer data ownership through shard metadata. Active defense verification is to send simulated attack data packets to storage nodes every 24 hours, and update their security scores based on node response delays and error rates. Ciphertext behavior monitoring analyzes ciphertext access logs through long short-term memory networks (LSTMs) and triggers alarms when unconventional time access and high-frequency abnormal request patterns are detected. Among them, the association obfuscation engine injects pseudo-shards before data storage. The algorithm is: the real shard set {F1, F2, F 3} is expanded to {F1_a, F2_b, F3_c, P_x, P_y}, where the metadata of the pseudo-shard P_x contains a randomly generated UUID and a forged timestamp. The liveness detection sandbox sends a CVE-2023-4863 simulated attack package at 02:00 UTC+8 every day, and analyzes the node response log through the ELK stack. If the WAF response time is greater than 200ms, the node security score is reduced. Ciphertext access monitoring uses a BiLSTM model trained in PyTorch. The input features include access time deviation (standard deviation calculation), source IP threat intelligence score (interconnection with Alien Vault OTX), and request frequency mutation rate (Z-score detection). The anomaly threshold is set to a confidence interval of 99.7% (3σ principle).

[0045] (8) The key lifecycle management system includes a timed rotation unit, a forward security unit, and a geo-fence unit. The timed rotation unit uses a clock service based on a hardware trusted execution environment (TEE) to generate new keys and re-encrypt affected data fragments every 72 hours. The forward security unit generates a key chain through a key derivation function tree to ensure that the new key is not affected by the leakage of historical keys. The geo-fence unit automatically freezes the activation permissions of related key fragments when the user terminal leaves the preset geo-fence area for more than 30 minutes. Among them, key rotation is driven by the trusted clock in the Intel SGX enclave, and new keys are automatically generated every 72 hours. Forward security is implemented through an HKDF-SHA256 derivation tree with a tree depth of 8 layers. The child keys of each layer are derived from the parent key and the round counter. The geo-fence system integrates the Baidu Map API and presets an electronic fence with the coordinates of the financial institution headquarters as the center point and a radius of 5 kilometers. When the terminal GPS coordinates exceed the fence for 30 consecutive minutes and the Bluetooth beacon does not respond, the key management system automatically marks the fragment status as frozen, and the on-site administrator needs to scan the iris to unfreeze it.

[0046] Example 1:

[0047] Phase 1: Data preprocessing (user terminal)

[0048] 1. Sensitivity grading

[0049] Loading pre-trained model: Deploy convolutional neural network model using TensorFlow Lite framework (input layer 128 nodes, hidden layer 64 nodes)

[0050] Data field analysis:

[0051] Amount field: When the value is ≥ RMB 500,000, it is marked as level 5

[0052] Account field: If it contains IBAN / SWIFT code, it is marked as Level 3

[0053] Product Code: Non-sensitive fields are marked as Level 1

[0054] Output labeled data block: {field name: value, sensitivity level: L}

[0055] 2. Dynamic fragment encoding

[0056] Sharding strategy execution:

[0057] When L=5: the Jerasure library is called to generate 6 slices (4 data slices + 2 redundant slices). The redundant slice generation rules are as follows:

[0058] Redundant slice 1 = data slice 1 XOR data slice 3

[0059] Redundant slice 2 = data slice 2 + data slice 4 × rotation factor (17) / / The rotation factor is a predefined integer

[0060] When L=3: Generate 4 slices (3 data slices + 1 redundant slice), where redundant slice = cyclic checksum of data slice 1 + data slice 2 + data slice 3

[0061] When L=1: the original data is not divided

[0062] Generate shard set: S = {s1, s2, ..., sk}

[0063] 3. Time and space stamp binding

[0064] Metadata construction:

[0065] Hash value: Apply the SHA3-256 algorithm to the shard to generate a 64-character digest

[0066] Geographical stamp: Get the longitude and latitude of the WGS84 coordinate system through the GPS module (accuracy ±5 meters)

[0067] Timestamp: Call the National Time Service Center API to obtain the ISO 8601 format time.

[0068] Digital Signature:

[0069] Concatenate metadata string M = hash value | longitude | latitude | time

[0070] Use the SM2 private key (stored in the SE chip) to calculate the signature Sig

[0071] Output signed shard: {shard data, M, Sig}

[0072] Phase 2: Dual-channel encryption (edge ​​nodes)

[0073] 1. Transmission channel encryption

[0074] Dynamic key update:

[0075] Quantum random number chip generates 256-bit true random number R

[0076] Generate a key pair (Pub, Priv) using the secp384r1 elliptic curve based on R

[0077] Start a 120-second countdown timer (hardware clock control)

[0078] Shard encryption process:

[0079] Input: shard data block D

[0080] Cryptographic operations:

[0081] Split D into a sequence of 32-byte blocks

[0082] Apply elliptic curve point multiplication to each block: C_i = D_i × Pub Add a 16-byte header (including the key version number)

[0083] Output ciphertext fragment CT

[0084] 2. Storage channel preprocessing

[0085] Homomorphic encryption initialization:

[0086] Configure polynomial ring: dimension 8192 (predefined prime modulus)

[0087] Generate key: call the KeyGenerator of the SEAL library

[0088] Attribute-based encryption policy compilation:

[0089] Input strategy: "Department = Finance Department AND Position Level ≥ 5"

[0090] Compiles to a visit tree:

[0091] Root node: AND gate

[0092] Left child node: department attribute comparator

[0093] Right child node: Position level threshold detector

[0094] Double encryption enforcement:

[0095] First use homomorphic encryption to generate the basic ciphertext CFHE

[0096] Use ABE public key to encrypt twice to generate the final ciphertext CABE

[0097] Phase 3: Multi-cloud Intelligent Routing (SD-WAN Controller)

[0098] 1. Real-time security assessment algorithm

[0099] Data collection cycle: every 30 seconds

[0100] Scoring calculation process:

[0101] Score = 0.3 × Compliance Score + 0.25 × (1- Attack Frequency / Maximum Frequency) + 0.2 × (1- CPU Load Rate) + 0.25 × Location Safety Factor

[0102] Position safety factor calculation:

[0103] Read the regional risk value R (0-10) from the geopolitical risk database

[0104] Position safety factor = 1-R / 10

[0105] 2. Reinforcement Learning Routing Decision

[0106] DQN model workflow:

[0107] State input: 10-dimensional vector (including sharding level, node score, bandwidth margin, etc.)

[0108] Action selection: Use ε-greedy strategy (ε=0.1)

[0109] Choose the action with the maximum Q value with 90% probability

[0110] Random exploration with 10% probability

[0111] Reward function design: Reward R = 0.6 × target node score + 0.3 × bandwidth utilization - 0.1 × path delay

[0112] Model update: Update network weights every 1000 decisions

[0113] 3. Protocol mimicry switching

[0114] Three-stage cycle:

[0115]

[0116]

[0117] Session persistence technology:

[0118] Using shared memory to store DTLS session state

[0119] Copy the session context to the new protocol stack when switching

[0120] Phase 4: Cloud Storage and Key Management

[0121] 1. Cross-cloud shard storage

[0122] Association obfuscation algorithm:

[0123] Actual shard set: {F1,F2,F3}

[0124] Generate two pseudo-shards:

[0125] Pseudo-shard P_x = randomly generated 512-bit data + forged metadata

[0126] Pseudo-fragment P_y = bit flip of F1 + timestamp offset

[0127] Storage set: {F1_a,F2_b,F3_c,P_x,P_y}

[0128] Location mapping table encryption:

[0129] Use AES-GCM to encrypt the mapping relationship

[0130] The ciphertext is stored in the Hyperledger Fabric blockchain

[0131] 2. Key Matrix Construction Shamir Secret Sharing Implementation:

[0132] Select prime number P > key K

[0133] Generate a 4th degree polynomial:

[0134] Polynomial f(x) = K + a1*x + a2*x 2 +a3*x 3 +a4*x 4 / / The coefficient ai is a random number

[0135] Calculate 7 fragments: s_i=(i,f(i))i=1..7

[0136] Fragment storage strategy:

[0137]

[0138] 3. Dynamic Defense Execution

[0139] Activity detection process:

[0140] Send CVE simulation attack package (including special SQL injection payload) at 02:00 every day

[0141] Monitoring node response:

[0142] Record WAF response time Δt

[0143] Count the number of error response codes E

[0144] Disposal rules:

[0145] If Δt>200ms then safety score -=20

[0146] If E>total number of requests × 5% then trigger shard migration

[0147] Ciphertext access monitoring:

[0148] Input feature vector: [access time deviation value, source IP threat score, request frequency change rate]

[0149] Abnormal judgment: alarm when Mahalanobis distance > threshold (predefined confidence interval)

[0150] Phase 5: Data Retrieval and Decryption

[0151] 1. Authorization access process

[0152] Biometric Authentication:

[0153] User presses the fingerprint sensor

[0154] Extract fingerprint feature points (Minutiae) and compare them with pre-stored templates

[0155] Pass when similarity ≥ 98%

[0156] Key reconstruction process:

[0157] Collect 5 fragments (e.g. s1, s3, s5, s6, s7)

[0158] Use Lagrange interpolation within the TEE:

[0159] Key K = Σ(s_i × basis function L(i)) mod P / / basis function is a precomputed polynomial

[0160] Output key to secure memory

[0161] Homomorphic decryption calculation:

[0162] step:

[0163] (1) Decrypt with ABE private key to obtain CFHE

[0164] (2) Decrypt CFHE using FHE key

[0165] (3) Perform fragment reorganization (inverse erasure coding operation)

[0166] Finally, a few points should be explained: First, in the description of this application, it should be noted that, unless otherwise specified or limited, the terms "mounted," "connected," and "connected" should be understood in a broad sense, and may refer to mechanical or electrical connections, internal communication between two components, or direct connection. "Up," "down," "left," and "right" are only used to indicate relative positional relationships. When the absolute positions of the objects being described change, the relative positional relationships may also change.

[0167] Secondly: The drawings of the embodiments disclosed in the present invention only involve structures related to the embodiments disclosed in the present invention. Other structures may refer to conventional designs. The same embodiment and different embodiments of the present invention may be combined with each other without conflict.

[0168] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A financial data encryption transmission and storage method based on cloud computing, characterized in that: It includes the following steps: Data preprocessing stage: Perform hierarchical sharding processing on the original financial data at the user terminal, and divide the data into shard sets of different security levels according to the preset data sensitivity classification rules; Parallel encryption stage: Differentially process the shard sets through a dual-channel asymmetric encryption engine. Among them, the transmission channel uses a dynamically updated asymmetric encryption algorithm to protect the data transmission process, and the storage channel uses a homomorphic encryption algorithm that supports ciphertext operations to process the data to be stored; Multi-cloud distribution stage: Dynamically allocate the encrypted data shards to the storage nodes of at least three independent cloud service providers through a heterogeneous multi-cloud routing module. This module continuously evaluates the security status of each cloud node and optimizes the routing path; Key management stage: Build a distributed key management matrix in the cloud, split the main decryption key into N mathematical fragments, and store them分散存储于区块链网络与用户端硬件安全模块,仅当获取M个碎片(M<N)时方可重构密钥。scattered in the blockchain network and the user terminal hardware security module. Only when M fragments (M < N) are obtained can the key be reconstructed; 2. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1 is characterized in that: The hierarchical sharding processing includes a sensitivity identification sub-step, a redundancy coding sub-step, and an anti-tampering marking sub-step. The sensitivity identification sub-step analyzes the financial data structure through a trained neural network model and automatically labels the sensitive levels of three types of data: transaction amount, account information, and identity credentials. The redundancy coding sub-step applies erasure coding technology to high-sensitivity level shards to generate parity shards, ensuring that the original data can still be completely restored when no more than the preset threshold of shards are lost. The anti-tampering marking sub-step attaches a digital signature tag containing a timestamp, geographical coordinates, and shard hash value to each shard, forming a traceable shard metadata chain.

3. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: The dual-channel asymmetric encryption engine includes: The transmission channel encryption unit uses an asymmetric encryption algorithm based on elliptic curves,配合量子随机数发生器每120秒更新密钥对,并在每次密钥更新后保留前代密钥60秒以处理传输中的遗留数据,存储通道加密单元构建全同态加密与属性基加密的混合架构,使云存储节点能在不解密条件下执行余额汇总、风险值计算操作,且仅当用户属性满足预设策略时才返回解密结果。cooperates with a quantum random number generator to update the key pair every 120 seconds, and retains the previous generation of keys for 60 seconds after each key update to process the remaining data in transmission. The storage channel encryption unit constructs a hybrid architecture of fully homomorphic encryption and attribute-based encryption, enabling the cloud storage node to perform balance aggregation and risk value calculation operations without decrypting, and only returns the decryption result when the user attributes meet the preset policy.

4. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: The heterogeneous multi-cloud routing module executes a security assessment sub-module, an intelligent routing sub-module, and a protocol mimicry sub-module: The security assessment sub-module continuously collects the compliance certification status, historical attack records, and real-time CPU load rate indicators of each cloud service node to generate a dynamic security score. The intelligent routing sub-module analyzes the current network threat situation based on a reinforcement learning model and preferentially routes the shards with the highest security level to the top three cloud nodes in terms of score. The protocol mimicry sub-module periodically switches between TLS1.3, QUIC, and private encryption protocols at the data transmission layer, and retains the parallel operation of the two protocols for 10 seconds each time to ensure seamless transition.

5. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: The distributed key management matrix implements a biometric binding mechanism, an active defense mechanism, and an audit tracking mechanism. The biometric binding mechanism requires verification of the user's biometrics (iris / fingerprint) during the key fragment reconstruction process, and the biometric template is only stored in the user's local security chip. The active defense mechanism automatically triggers the self-destruction of associated fragments within 200 kilometers of any key fragment storage point when a brute force cracking attempt is detected. The audit tracking mechanism records all key operations on the consortium blockchain and is written to an unalterable ledger after the legitimacy of the operation is verified by three independent audit nodes.

6. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: It also includes a dynamic encrypted tunnel switching mechanism: dual tunnels are established between the user terminal and the cloud entry gateway to synchronously create a main tunnel (SM9 algorithm) and a backup tunnel (NTRU algorithm). Intelligent switching control uses dedicated probes to monitor the transmission quality of the main tunnel in real time. When the packet loss rate exceeds 15% for five consecutive times or the latency suddenly increases by 300%, it automatically switches to the backup tunnel. Seamless state migration uses memory mirroring technology to synchronize the encrypted session state to the backup tunnel in real time to ensure that the switching process does not interrupt business interaction.

7. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: Implementation at the cloud storage layer: Shard association obfuscation injects pseudo-association identifiers into data shards stored on different cloud platforms, preventing external attackers from inferring data ownership through shard metadata. Active defense verification sends simulated attack data packets to storage nodes every 24 hours, and updates their security scores based on node response delays and error rates. Ciphertext behavior monitoring analyzes ciphertext access logs through a long short-term memory network (LSTM), triggering alarms when unconventional access times and high-frequency abnormal request patterns are detected.

8. The method for encrypted transmission and storage of financial data based on cloud computing according to claim 1, characterized in that: The key lifecycle management system includes a timed rotation unit, a forward security unit and a geo-fence unit. The timed rotation unit adopts a clock service based on a hardware trusted execution environment (TEE) to generate new keys every 72 hours and re-encrypt the affected data fragments. The forward security unit generates a key chain through a key derivation function tree to ensure that the new key is not affected by the leakage of historical keys. The geo-fence unit automatically freezes the activation permissions of related key fragments when the user terminal leaves the preset geo-fence area for more than 30 minutes.

Citation Information

Patent Citations

  • Authentication method based on bilinear mapping and dot product protocol in intelligent medical treatment

    CN113114689A

  • Hypervisor-based virtual machine communication method and device, readable storage medium and electronic equipment

    CN114244515A

  • Unit for controlling data communication

    US20220141050A1

Cited By

  • Intelligent Internet of Things equipment storage data transmission system and method based on block chain

    CN121036946A

  • Multi-dimensional encryption transmission protection method and system for cross-border payment

    CN121056243A

  • Multi-dimensional encryption transmission protection method and system for cross-border payment

    CN121056243B

  • Time capsule interaction method and system

    CN121071934A

  • Switching method of communication link group based on aircraft and communication system

    CN121284551A