Multi-party collaborative signature management method and system based on key segmentation technology
By deploying AI exception detection model and dynamic management of key sharding in the encryption system, the problem of key sharding fixed and undynamic adaptation in the collaborative signature of multiple participants is solved, and high security and reliability signature management is achieved.
Patent Information
- Application Number
- CN202510413642.2
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-04-03
- Publication Date
- 2025-06-13
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
When multiple participants sign together, the key sharding is fixed for a long time, which is prone to failure of historical signatures due to sharding leakage or cracking, and it is impossible to dynamically adapt to organizational structure changes or identify malicious nodes.
By initializing data in the encryption system, collecting historical key abnormal behavior data, deploying AI exception detection models, and using the exception detection model for training, dynamically adjusting signature thresholds and rotation keys, realizing dynamic management and security monitoring of keys.
Improve the security of keys and the reliability of signatures, dynamically adapt to organizational structure changes and identify malicious nodes, reducing the risk of keys being cracked or leaked.
Smart Images

Figure CN120150946A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of sharding encryption, and specifically to a multi-party collaborative signature management method and system based on key splitting technology. Background Art
[0002] With the wide application of mobile intelligent terminals, their communication content is transmitted through public wireless channels, which is vulnerable to attacks. Moreover, the operating systems are complex and the applications are diverse, making security issues increasingly prominent. Traditional hardware cryptographic devices such as USBs and KEYs have problems such as being difficult to carry and inconvenient to connect, and it is difficult to meet the security requirements of mobile intelligent terminals. In an encryption system, the secure storage of keys is crucial. If the keys are not stored properly, they are easily obtained by attackers, leading to information leakage and system attacks. Splitting the key into multiple components and storing them separately can reduce the risk of key leakage. In some application scenarios, such as e-government, e-commerce, finance and other fields, multiple parties need to jointly sign to ensure the security, integrity and non-repudiation of transactions. Through key splitting technology, multi-party collaborative signature can be realized, so that only multiple parties can complete the signature operation through joint cooperation, improving the security and reliability of the signature.
[0003] However, when multiple parties participate in the signature collaboratively, the traditional key shards are fixed for a long time. Once a shard is leaked or cracked, all historical signatures may become invalid, and it cannot dynamically adapt to changes in the organizational structure. Moreover, the existing systems rely on static policies and cannot dynamically identify malicious nodes or adaptively adjust the signature rules. Summary of the Invention
[0004] The purpose of the present invention is to provide a multi-party collaborative signature management method and system based on key splitting technology to solve the problems raised in the prior art.
[0005] To achieve the above purpose, the present invention provides the following technical solutions:
[0006] A multi-party collaborative signature management method based on key splitting technology, the method includes the following steps:
[0007] S100. Initialize data in the encryption system, collect the behavior data when key anomalies occur in history, deploy an AI anomaly detection model, and train the anomaly detection model using the behavior data of historical anomaly occurrences;
[0008] Further, the specific steps of training the anomaly detection model using the behavior data of historical anomaly occurrences are:
[0009] S101. In the encryption system, set the initial threshold to k, the number of shard private keys required for signature is greater than or equal to k, and the initial time window is T 0Collect the behavior data when sharding node anomalies occur in the history of the computer encryption system, clean the collected behavior data, remove duplicate and incomplete data. For missing data, calculate the average value of the two adjacent data on both sides of the missing part as the filling value, and use the filling value to supplement the missing data; normalize the cleaned behavior data;
[0010] S102. Deploy an anomaly detection model in the encryption system using the Isolation Forest algorithm. Divide the abnormal behavior data collected from each node into a training set and a test set according to a ratio of 7:3, where the ratio 7 is the training set and the ratio 3 is the test set; use the training set to train the anomaly detection model, and use the test set to evaluate the trained model, specifically by calculating the accuracy of the model using the test set;
[0011] S103. When training the model using the training set, extract the abnormal scores output by the model during each training, obtain the abnormal scores of the behavior data when anomalies occur in each node in the history, calculate the average value and standard deviation of the abnormal scores of each node respectively, and calculate the abnormal score threshold of each node using the average value and standard deviation. The formula is:
[0012] Ex i =Sp i +β i ×Sb i ;
[0013] In the formula, Ex i represents the anomaly score threshold of the i-th node, Sp i represents the average value of the anomaly scores of the i-th node, Sb i represents the standard deviation of the anomaly scores of the i-th node, and β i represents the threshold coefficient of the i-th node; calculate the anomaly score thresholds of all nodes in sequence.
[0014] Initializing the data provides a definite starting state for the encryption system, enabling all parameters and configurations to be in a known initial state when the system starts running. This helps to ensure the stability and reliability of the system and avoid system errors or anomalies caused by uninitialized or improperly initialized conditions.
[0015] Collecting the behavior data when key anomalies occur in the history can help the system record various possible abnormal situations, such as key leakage, key tampering, abnormal key usage frequency, etc. These data can serve as important clues for discovering potential security threats, helping to detect vulnerabilities or signs of being attacked in the encryption system in a timely manner, so as to take corresponding measures for prevention and repair.
[0016] The AI anomaly detection model can monitor key-related behaviors in the encryption system in real time. Once it detects abnormal behaviors deviating from the normal mode, it can quickly issue an alarm, enabling the administrator to take timely measures for response and minimizing the losses caused by potential security incidents. Compared with traditional rule-based detection methods, the AI model can more sensitively capture complex and non-traditional abnormal patterns, improving the accuracy and timeliness of detection.
[0017] S200. Generate the master key through the distributed key generation protocol, use the hierarchical deterministic algorithm to derive time, split the master key to form key shards, and distribute each shard to different nodes;
[0018] Furthermore, the specific steps for distributing each shard to different nodes are as follows:
[0019] S201. Generate random polynomials at each node in the encryption system, generate node private key shares at each node using the distributed key generation protocol, sum up all the node private key shares to obtain the common master key of all nodes, and set the master private key as SK master ;
[0020] S202. Initialize the encryption system, set the initialization time window as T 0 , the shard rotation period as △T, and calculate the time window number t. The formula is:
[0021]
[0022] In the formula, T now represents the current time, and the formula represents taking the floor of the result after dividing by the rotation period;
[0023] Calculate the shard private key using the generated master private key and time window t. The formula is:
[0024] SK t,i = HMAC - SHA256(SKmaste r,t ‖i) mod p;
[0025] In the formula, SK t,i represents the shard private key of the i-th node at time window t, mod represents taking the remainder, HMAC - SHA256 represents the hash algorithm, p represents the prime field that maps the hash value to the elliptic curve group, and t||i represents concatenating the time window and the shard index; take the modulus of the HMAC output hash value with respect to the size p of the elliptic curve prime field to make SK t,i ∈[1, p - 1];
[0026] Send the calculated SK t,i to the corresponding node through a secure channel, and the node stores the shard private key in the trusted execution environment after receiving it.
[0027] The master key is split into multiple shards and distributed across different nodes, avoiding the risks associated with storing the complete master key on a single node. Even if a certain node is attacked, the attacker cannot obtain the complete master key, thus greatly enhancing the security of the key.
[0028] When it is necessary to update the master key or revoke the permissions of a certain node, operations can be performed directly on the corresponding node without the need to regenerate and redistribute the keys for the entire system. This can reduce the complexity of key management and, at the same time, minimize the security risks brought about by key updates or revocations.
[0029] By using a hierarchical deterministic algorithm to derive keys, a series of related keys can be generated according to time or other specific rules. This approach makes the generation and management of keys more orderly and predictable. Meanwhile, the hierarchical structure also facilitates the classification and permission management of keys. Different levels of keys can correspond to different operations or data access permissions, enabling administrators to configure them flexibly according to actual requirements.
[0030] S300. Analyze the times when key anomalies occurred in history, use the linear regression algorithm to fit and obtain the anomaly occurrence frequency function based on the time series, calculate the rotation period on the time series using the anomaly occurrence frequency, and formulate a periodic rotation trigger mechanism;
[0031] Further, the specific steps for formulating the periodic rotation trigger mechanism are as follows:
[0032] S301. Let the time series be {ti 1 、ti 2 、ti 3 、...、ti n}, and collect the number of anomalies that occurred in the corresponding time series in history as {y 1 、y 2 、y 3 、...、y n}; Use the linear regression algorithm to fit the time series and the number of anomalies to obtain the anomaly occurrence frequency function as y = α 0 +α 1 ×ti. In the function, α 0 represents the constant term, and α 1 represents the coefficient of the independent variable; Calculate the values of α 0 and α 1 using the least squares method;
[0033] S302. Calculate the rotation period using the anomaly occurrence frequency function. The formula is: △T ti =1 / y ti , where △T tiDenote the rotation period when the time series is \(t_i\), \(y\) ti Denote the abnormal occurrence frequency when the time series is \(t_i\);
[0034] S303. The encryption system calculates the corresponding rotation period according to the time series to generate a new shard private key, uses the new shard private key to replace the old shard private key, and updates the hash value using a forward-secure hash chain during rotation. The formula is:
[0035] H t = H(H t-1 ‖ Sign(SK t , H t-1 ));
[0036] In the formula, H t Denotes the hash value at time window \(t\), H t-1 Denotes the hash value at time window \(t - 1\), Sign(SK t , H t-1 ) Denotes the result of signing H t using the private key SK t-1 .
[0037] By analyzing the historical key abnormal time, fitting the abnormal occurrence frequency function, the potential law of key abnormality occurrence can be discovered. The rotation period calculated based on this can cause the key to be rotated before an abnormality may occur, control the risk at the budding state, avoid the security hidden dangers accumulated due to the long-term use of the same key, and actively reduce the risk of the key being cracked or leaked.
[0038] Periodically rotating the key can effectively resist long-term attacks against the key. Even if the attacker obtains some key information within a period of time, due to the timely rotation of the key, the information they have will soon become invalid and it is difficult to cause substantial damage to the system, thus enhancing the security and stability of the entire encryption system.
[0039] S400. Use the AI anomaly detection model to perform real-time detection on the behavior data of each node, construct feature vectors, calculate the anomaly score of each node, and determine whether each node has an abnormal risk and mark it;
[0040] Furthermore, the specific steps for determining whether each node has an abnormal risk and marking it are as follows:
[0041] S401. When the encryption system performs real-time shard encryption, collect the behavior data of each node, convert the collected behavior data into a feature vector \(x\), and use the anomaly detection model to calculate the real-time anomaly score of each node. The formula is:
[0042]
[0043] In the formula, s i represents the anomaly score of the i-th node, and E(h(x i )) represents the average path length of the feature vector x of the i-th node among multiple trees; c(m) represents the normalization factor; i
[0044] S402. Use the anomaly threshold of each node to judge the real-time anomaly score. When s i > Ex i , it is judged that the i-th node has an anomaly risk; when s i ≤ Ex i , it is judged that the i-th node has no anomaly risk; after judging that the node has an anomaly risk, mark it.
[0045] S500. Establish an emergency rotation mechanism to re-allocate shards for rotation for the marked nodes with anomaly risks; and dynamically adjust the threshold according to the anomaly score;
[0046]
[0047] Further, the specific steps for dynamically adjusting the shard weight and threshold according to the anomaly score are as follows:
[0048]
[0049] S501. Use the anomaly score to adjust the signature threshold. The formula is: In the formula, k new represents the new adjusted threshold, k represents the old threshold before adjustment, θ represents the adjustment coefficient, m represents the total number of nodes; max(s i - Ex i , 0) represents the summation of the anomaly scores exceeding the anomaly threshold;
[0050] S502. After detecting an abnormal node, the encryption system generates a temporary time window t', re-allocates shards, and performs an emergency rotation of the shard private keys.
[0051] When detecting the marked nodes with anomaly risks, the emergency rotation mechanism can be quickly started to re-allocate shards for rotation. In this way, the nodes that may have security risks can be promptly isolated from other parts of the system, preventing the further spread of potential attacks or key leakage incidents, and minimizing the impact on the entire system.
[0052] Dynamically adjust the threshold according to the anomaly score, enabling the system to flexibly adjust the sensitivity to anomalies based on the actual security situation. If the anomaly score is high, it indicates that the system faces a greater security risk. At this time, lowering the threshold can allow the system to more keenly detect potential abnormal behaviors and take measures in a timely manner; conversely, if the anomaly score is low, appropriately increasing the threshold can reduce false alarms, avoid unnecessary emergency rotations, and improve the stability and operation efficiency of the system.
[0053] As the operating environment of the system changes, such as the update of network attack methods and the change of system load, the occurrence of anomalies in nodes will also be different. The mechanism of dynamically adjusting the threshold enables the system to automatically adapt to these changes, always maintaining the effective monitoring and response capabilities for anomalies, and ensuring that the system can maintain a high level of security under different environmental conditions.
[0054] S600. The client initiates a signature request, excludes abnormal nodes, filters and aggregates shard nodes, and performs collaborative signature.
[0055] Furthermore, the specific steps for performing collaborative signature are as follows:
[0056] S601. The client initiates a signature request, attaches the time window identifier t, removes the shard private keys of nodes with abnormal risks, and extracts the shard private keys of more than the new threshold k new aggregate and sign the shard private keys of nodes, and use the Schnorr aggregate signature protocol for signature.
[0057] A multi-party collaborative signature management system based on key splitting technology. The multi-party collaborative signature management system includes a data collection module, an initialization module, a training module, a shard private key module, a shard rotation module, an anomaly detection module, a dynamic adjustment module, and a signature module;
[0058] The data collection module is used to collect node behavior data when anomalies occur in the encryption system history;
[0059] The initialization module is used to initialize and define the shard-encrypted data;
[0060] The training module is used to train the anomaly detection model using the node behavior data when anomalies occur in the collected history;
[0061] The shard private key module is used for the system to generate the master private key and calculate the shard private keys using the hash algorithm;
[0062] The shard rotation module is used to construct a function of the anomaly occurrence frequency based on time series using the number of anomalies in the history, and calculate the rotation period of the corresponding time series; when it is determined that a node has an abnormal risk, an emergency rotation is immediately performed;
[0063] The anomaly detection module is used to calculate the anomaly score during real-time encryption using the anomaly detection module, and determine whether there is an anomaly risk for the node by using the anomaly threshold;
[0064] The dynamic adjustment module is used to dynamically adjust the signature threshold using the anomaly score;
[0065] The signature module is used to perform signatures using the adjusted new threshold.
[0066] The shard rotation module includes a periodic rotation unit and an emergency rotation unit;
[0067] The periodic rotation unit is used to construct a function of the anomaly occurrence frequency based on time series using the number of anomalies in history, calculate the rotation period corresponding to the time series, and perform shard rotation using the rotation period;
[0068] The emergency rotation unit is used to immediately perform an emergency rotation when it is determined that the node has an anomaly risk.
[0069] The training module includes a model training unit and an anomaly threshold calculation unit;
[0070] The model training unit is used to train the anomaly detection model using the node behavior data during the historical anomaly occurrence collected;
[0071] The anomaly threshold calculation unit is used to extract the anomaly scores during model training and calculate the anomaly threshold for each node.
[0072] Compared with the prior art, the beneficial effects of the present invention are:
[0073] 1. By analyzing the historical key anomaly time, the present invention fits the anomaly occurrence frequency function, and can discover the potential law of key anomaly occurrence. The rotation period calculated based on this can rotate the key before an anomaly may occur, control the risk in the budding state, avoid the security hidden dangers accumulated due to the long-term use of the same key, and actively reduce the risk of the key being cracked or leaked.
[0074] 2. The present invention dynamically adjusts the threshold according to the anomaly score, enabling the system to flexibly adjust the sensitivity to abnormal situations according to the actual security situation. If the anomaly score is high, it indicates that the system faces a greater security risk. At this time, reducing the threshold can make the system more sensitive to capture potential abnormal behaviors and take measures in a timely manner; conversely, if the anomaly score is low, appropriately increasing the threshold can reduce false alarms, avoid unnecessary emergency rotations, and improve the stability and operation efficiency of the system. Brief Description of the Drawings
[0075] Figure 1 It is the module distribution diagram of the multi-party collaborative signature management system based on the key splitting technology of the present invention;
[0076] Figure 2 This is a schematic diagram of the steps of the multi-party collaborative signature management method based on the key splitting technology of the present invention. Specific embodiments
[0077] Next, the technical solutions in the embodiments of the present invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.
[0078] Embodiment: As Figure 1 - Figure 2 shown, the present invention provides a technical solution,
[0079] A multi-party collaborative signature management method based on the key splitting technology, the method comprising the following steps:
[0080] S100. Initialize data in the encryption system, collect the behavior data when key anomalies occur in history, deploy an AI anomaly detection model, and train the anomaly detection model using the behavior data of historical anomaly occurrences;
[0081] The specific steps of training the anomaly detection model using the behavior data of historical anomaly occurrences are:
[0082] S101. In the encryption system, set the initial threshold to k, the number of sharded private keys required during signature should be greater than or equal to k, and the initial time window is T 0 ; Collect the behavior data when sharded node anomalies occur in the history of the computer encryption system, clean the collected behavior data, remove duplicate and incomplete data, for the missing data, calculate the average value of the two adjacent data on both sides of the missing part as the filling value, and use the filling value to supplement the missing data; Normalize the cleaned behavior data;
[0083] S102. Deploy an anomaly detection model in the encryption system using the Isolation Forest algorithm, divide the anomaly behavior data collected from each node into a training set and a test set according to a ratio of 7:3, where the ratio 7 is the training set and the ratio 3 is the test set; Use the training set to train the anomaly detection model, and use the test set to evaluate the trained model, specifically calculate the accuracy rate of the model using the test set;
[0084] S103. When training the model using the training set, extract the anomaly scores output by the model during each training, obtain the anomaly scores of the behavior data when anomalies occur in each node in history, calculate the average value and standard deviation of the anomaly scores of each node respectively, and calculate the anomaly score threshold of each node using the average value and standard deviation. The formula is:
[0085] Ex i = Sp i + β i × Sb i ;
[0086] In the formula, Ex i represents the anomaly score threshold of the i-th node, Sp i represents the average anomaly score of the i-th node, Sb i represents the standard deviation of the anomaly score of the i-th node, and β i represents the threshold coefficient of the i-th node; the anomaly score thresholds of all nodes are calculated sequentially.
[0087] Initializing the data provides a definite starting state for the encryption system, enabling all parameters and configurations to be in a known initial state when the system starts running. This helps ensure the stability and reliability of the system and avoid system errors or anomalies caused by uninitialized or improperly initialized conditions.
[0088] Collecting the behavior data when key anomalies occur in the collection history can help the system record various possible anomaly situations, such as key leakage, key tampering, abnormal key usage frequencies, etc. These data can serve as important clues for discovering potential security threats, helping to promptly detect vulnerabilities or signs of being attacked in the encryption system, so as to take corresponding measures for prevention and repair.
[0089] The AI anomaly detection model can monitor the key-related behaviors in the encryption system in real time. Once it detects abnormal behaviors deviating from the normal mode, it can quickly issue an alarm, enabling the administrator to take timely measures for response and minimizing the losses caused by potential security incidents. Compared with traditional rule-based detection methods, the AI model can more acutely capture complex and non-traditional anomaly patterns, improving the accuracy and timeliness of detection.
[0090] S200. Generate the master key through the distributed key generation protocol, derive the time using the hierarchical deterministic algorithm, split the master key to form key shards, and distribute each shard to different nodes;
[0091] The specific steps for distributing each shard to different nodes are as follows:
[0092] S201. Generate a random polynomial at each node in the encryption system, generate the node private key share at each node using the distributed key generation protocol, sum up all the node private key shares to obtain the common master key of all nodes, and let the master private key be SK master ;
[0093] S202. Initialize the encryption system, and let the initialization time window be T0 , the sharding rotation period is △T, and the time window number t is calculated using the formula:
[0094]
[0095] In the formula, T now represents the current time, and the formula means taking the floor of the result after dividing by the rotation period;
[0096] Calculate the sharding private key using the generated master private key and time window t. The formula is:
[0097] SK t,i = HMAC-SHA256(SK master , t||i) mod p;
[0098] In the formula, SK t,i represents the sharding private key of the i-th node at time window t, mod represents taking the remainder, HMAC-SHA256 represents the hash algorithm, p represents the prime field that maps the hash value to the elliptic curve group, and t||i represents concatenating the time window and the sharding index; take the modulus of the HMAC output hash value with the size p of the elliptic curve prime field to make SK t,i ∈[1, p - 1];
[0099] Send the calculated SK t,i to the corresponding node through a secure channel, and the node stores it in a trusted execution environment after receiving the sharding private key.
[0100] Dividing the master key into multiple shards and distributing them on different nodes avoids the risk brought by storing the complete master key on a single node. Even if a certain node is attacked, the attacker cannot obtain the complete master key, thus greatly improving the security of the key.
[0101] When it is necessary to update the master key or revoke the permission of a certain node, only operations need to be performed on the corresponding node, without the need to regenerate and distribute the keys of the entire system. This can reduce the complexity of key management and at the same time reduce the security risks brought by key update or revocation.
[0102] Deriving keys using a hierarchical deterministic algorithm can generate a series of related keys according to time or other specific rules. This method makes the generation and management of keys more orderly and predictable. At the same time, the hierarchical structure is also convenient for classifying and managing permissions of keys. Keys at different levels can correspond to different operations or data access permissions, facilitating flexible configuration by administrators according to actual needs.
[0103] S300. Analyze the time when the key was abnormal in history, use the linear regression algorithm to fit and obtain the abnormal occurrence frequency function based on the time series, calculate the rotation period on the time series using the abnormal occurrence frequency, and formulate a periodic rotation trigger mechanism;
[0104] The specific steps for formulating the periodic rotation trigger mechanism are as follows:
[0105] S301. Let the time series be {ti 1 、ti 2 、ti 3 、...、ti n}, and collect the number of times of abnormality occurring in the corresponding time series in history as {y 1 、y 2 、y 3 、...、y n}; Use the linear regression algorithm to fit the time series and the number of abnormal times to obtain the abnormal occurrence frequency function as y = α 0 +α 1 ×ti. In the function, α 0 represents the constant term, and α 1 represents the independent variable coefficient; Calculate the values of α 0 and α 1 using the least squares method;
[0106] S302. Calculate the rotation period using the abnormal occurrence frequency function. The formula is: △T ti = 1 / y ti , where △T ti represents the rotation period when the time series is ti, and y ti represents the abnormal occurrence frequency when the time series is ti;
[0107] S303. The encryption system calculates the corresponding rotation period according to the time series to generate a new shard private key, uses the new shard private key to replace the old shard private key, and updates the hash value using a forward-secure hash chain during rotation. The formula is:
[0108] H t = H(H t-1 ||Sign(SK t , H t-1 ));
[0109] In the formula, H t represents the hash value at time window t, H t-1 represents the hash value at time window t - 1, and Sign(SK t , H t-1 ) represents the result of signing H t with the private key SK t-1 .
[0110] By analyzing the historical key anomaly times and fitting the anomaly occurrence frequency function, the potential pattern of key anomalies can be discovered. The rotation period calculated based on this can cause the key to be rotated before anomalies may occur, control the risk at the budding stage, avoid the security risks accumulated due to the long-term use of the same key, and proactively reduce the risk of the key being cracked or leaked.
[0111] Periodically rotating the key can effectively resist long-term attacks against the key. Even if the attacker obtains partial key information within a certain period of time, due to the timely rotation of the key, the information they have will soon become invalid and it is difficult to cause substantial damage to the system, thereby enhancing the security and stability of the entire encryption system.
[0112] S400. Use the AI anomaly detection model to detect the behavior data of each node in real time, construct feature vectors, calculate the anomaly scores of each node, and determine whether there is an anomaly risk for each node and mark it;
[0113] The specific steps for determining whether there is an anomaly risk for each node and marking it are as follows:
[0114] S401. When the encryption system performs real-time sharding encryption, collect the behavior data of each node, convert the collected behavior data into a feature vector x, and use the anomaly detection model to calculate the real-time anomaly score of each node. The formula is:
[0115]
[0116] In the formula, s i represents the anomaly score of the i-th node, and E(h(x i )) represents the average path length of the feature vector x of the i-th node in multiple trees; c(m) represents the normalization factor; i c(m) = 2H(m - 1) - 2(m - 1) / m;
[0117] c(m) = 2H(m - 1) - 2(m - 1) / m;
[0118] S402. Use the anomaly threshold of each node to judge the real-time anomaly score. When s i > Ex i , it is judged that the i-th node has an anomaly risk; when s i ≤ Ex i , it is judged that the i-th node does not have an anomaly risk; after judging that the node has an anomaly risk, mark it.
[0119] S500. Formulate an emergency rotation mechanism, reassign shards for rotation to the marked nodes with anomaly risks; and dynamically adjust the threshold according to the anomaly scores;
[0120] The specific steps for dynamically adjusting the shard weight and threshold according to the anomaly score are as follows:
[0121] S501. Adjust the signature threshold using the anomaly score. The formula is:
[0122]
[0123] In the formula, k new represents the new adjusted threshold, k represents the old threshold before adjustment, represents the adjustment coefficient, m represents the total number of nodes; max(s i -Ex i , 0) represents the summation of the anomaly scores exceeding the anomaly threshold;
[0124] S502. After detecting an abnormal node, the encryption system generates a temporary time window t', reallocates the shards, and performs an emergency rotation of the shard private keys.
[0125] When an abnormal risk node is found marked, the emergency rotation mechanism can be quickly activated to reallocate the shards for rotation. This can promptly isolate the nodes that may have security risks from the rest of the system, preventing the further spread of potential attacks or key leakage incidents and minimizing the impact on the entire system.
[0126] Dynamically adjusting the threshold according to the anomaly score enables the system to flexibly adjust its sensitivity to abnormal situations based on the actual security status. If the anomaly score is high, it indicates that the system faces a greater security risk. At this time, lowering the threshold can allow the system to more keenly detect potential abnormal behaviors and take measures promptly; conversely, if the anomaly score is low, appropriately increasing the threshold can reduce false alarms, avoid unnecessary emergency rotations, and improve the stability and operating efficiency of the system.
[0127] As the operating environment of the system changes, such as the update of network attack methods and the change of system load, the situation of node anomalies will also be different. The mechanism of dynamically adjusting the threshold enables the system to automatically adapt to these changes, always maintain the ability to effectively monitor and respond to abnormal situations, and ensure that the system can maintain a high level of security under different environmental conditions.
[0128] S600. The client initiates a signature request, excludes abnormal nodes, filters and aggregates shard nodes, and performs collaborative signing.
[0129] The specific steps for performing collaborative signing are as follows:
[0130] S601. The client initiates a signature request, attaches the time window identifier t, removes the shard private keys of the nodes with abnormal risks, and extracts those greater than the new threshold k newAggregate signatures are performed on the sharded private keys of the nodes, and the Schnorr aggregate signature protocol is used for signing.
[0131] A multi-party collaborative signature management system based on key splitting technology. The multi-party collaborative signature management system includes a data collection module, an initialization module, a training module, a sharded private key module, a sharded rotation module, an anomaly detection module, a dynamic adjustment module, and a signature module;
[0132] The data collection module is used to collect node behavior data when an anomaly occurs in the history of the encryption system;
[0133] The initialization module is used to perform initialization definitions on the sharded encrypted data;
[0134] The training module is used to train the anomaly detection model using the node behavior data during anomaly occurrences in the collected history;
[0135] The sharded private key module is used for the system to generate a master private key and calculate sharded private keys using a hash algorithm;
[0136] The sharded rotation module is used to construct a function of the anomaly occurrence frequency based on time series using the number of anomalies in the history, and calculate the rotation period of the corresponding time series; when it is determined that a node has an anomaly risk, an emergency rotation is immediately performed;
[0137] The anomaly detection module is used to calculate the anomaly score during real-time encryption using the anomaly detection module, and determine whether a node has an anomaly risk using an anomaly threshold;
[0138] The dynamic adjustment module is used to dynamically adjust the signature threshold using the anomaly score;
[0139] The signature module is used to perform signatures using the adjusted new threshold.
[0140] The sharded rotation module includes a periodic rotation unit and an emergency rotation unit;
[0141] The periodic rotation unit is used to construct a function of the anomaly occurrence frequency based on time series using the number of anomalies in the history, calculate the rotation period of the corresponding time series, and perform sharded rotation using the rotation period;
[0142] The emergency rotation unit is used to immediately perform an emergency rotation when it is determined that a node has an anomaly risk.
[0143] The training module includes a model training unit and an anomaly threshold calculation unit;
[0144] The model training unit is used to train the anomaly detection model using the node behavior data during anomaly occurrences in the collected history;
[0145] The abnormal threshold calculation unit is used to extract the abnormal scores during model training and calculate the abnormal threshold for each node.
[0146] Embodiment: Now, in a certain computer encryption system, sharded private keys are implemented to determine whether there is an abnormal risk for its nodes.
[0147] The behavior data of a certain node is collected in real time, including initiating 10 signature requests within the past 1 hour, with an average delay dA = 150 ms, a location offset ΔgA = 500 km, and a shard utilization rate uA = 20%.
[0148] The behavior data is converted into a feature vector x = [2.5, 2.5, 4.5, -3].
[0149] The calculated abnormal score is 0.81. Assuming the abnormal threshold is 0.7, it is determined that there is an abnormal risk.
[0150] If the current threshold k = 5, then:
[0151] k new = 5 + 0.1×(0.81 - 0.7) = 5.011 ≈ 5;
[0152] The threshold value is adjusted to 5, but the sharded private key of node A may be temporarily frozen.
[0153] For those skilled in the art, it is obvious that the present invention is not limited to the details of the above exemplary embodiments, and the present invention can be implemented in other specific forms without departing from the spirit or basic characteristics of the present invention. Therefore, from any point of view, the embodiments should be regarded as exemplary and non-restrictive. The scope of the present invention is defined by the appended claims rather than the above description. Therefore, all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present invention. Any reference signs in the claims should not be construed as limiting the claimed rights.
Claims
1. A multi-party collaborative signature management method based on key splitting technology, characterized in that: The method comprises the following steps: S100, initialize data in the encryption system, collect historical behavior data when key anomalies occur, deploy an AI anomaly detection model, and train the anomaly detection model using historical behavior data when anomalies occur; S200, generating a master key through a distributed key generation protocol, deriving time using a hierarchical deterministic algorithm, dividing the master key into key shards, and distributing each shard to different nodes; S300, analyzing the time when the key anomalies occurred in history, fitting the anomaly frequency function based on the time series using a linear regression algorithm, calculating the rotation period on the time series using the anomaly frequency, and formulating a periodic rotation trigger mechanism; S400, use the AI anomaly detection model to perform real-time detection on the behavior data of each node, build a feature vector, calculate the anomaly score of each node, determine whether each node has an abnormal risk and mark it; S500. Develop an emergency rotation mechanism to reallocate shards for rotation for nodes marked as having abnormal risks; and dynamically adjust the threshold based on the abnormality score; S600: The client initiates a signature request, excludes abnormal nodes, screens shard node aggregation, and performs collaborative signing.
2. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of using the historical abnormal behavior data to train the anomaly detection model in S100 are: S101. In the encryption system, set the initial threshold to k, the number of shard private keys required for signing is greater than or equal to k, and the initial time window is T0; collect the behavior data when the shard node abnormality occurs in the history of the computer encryption system, clean the collected behavior data, remove duplicate and incomplete data, and for missing data, calculate the average value of the adjacent data on both sides of the missing part as the filling value, and use the filling value to supplement the missing data; normalize the cleaned behavior data; S102, deploying an anomaly detection model in the encryption system using the isolation forest algorithm, dividing the abnormal behavior data collected from each node into a training set and a test set in a ratio of 7:3, where the ratio 7 is the training set and the ratio 3 is the test set; using the training set to train the anomaly detection model, and using the test set to evaluate the trained model, specifically using the test set to calculate the accuracy of the model; S103. When the model is trained using the training set, the model output anomaly score is extracted during each training, and the anomaly score of the behavior data when an anomaly occurs at each node in the history is obtained. The average value and standard deviation of the anomaly score of each node are calculated respectively, and the anomaly score threshold of each node is calculated using the average value and standard deviation. The formula is: From i =Sp i +β i ×Sb i ; In the formula, Ex i represents the abnormal score threshold of the i-th node, Sp i represents the average abnormal score of the i-th node, Sb i represents the standard deviation of the abnormal score of the i-th node, β i Represents the threshold coefficient of the i-th node; the abnormal score thresholds of all nodes are calculated in sequence.
3. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of distributing each shard to different nodes in S200 are: S201. Generate a random polynomial at each node in the encryption system, use a distributed key generation protocol to generate a node private key share in each node, sum up all node private key shares to obtain a common master key for all nodes, and set the master private key to be SK master ; S202, initialize the encryption system, set the initialization time window to T0, the shard rotation period to △T, calculate the time window number t, the formula is: In the formula, T now Indicates the current time, and the formula indicates that the result after the rotation period is rounded down; The generated master private key and time window t are used to calculate the shard private key. The formula is: SK t,i =HMAC-SHA256(SK master ,t||i))mod p; In the formula, SK t,i represents the shard private key of the i-th node in time window t, mod represents remainder, HMAC-SHA256 represents the hash algorithm, p represents mapping the hash value to the prime field of the elliptic curve group, and t||i represents concatenating the time window and the shard index; the HMAC output hash value is modulo the size of the elliptic curve prime field p, so that SK t,i ∈[1,p-1]; The calculated SK is transmitted through a secure channel t,i Sent to the corresponding node, the node stores the sharded private key in the trusted execution environment after receiving it.
4. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of formulating the periodic rotation trigger mechanism in S300 are: S301, let the time series be {ti1, ti2, ti3, ..., ti n }, the number of abnormalities that occurred in the corresponding time series in the collection history is {y1, y2, y3, ..., y n }; Use linear regression algorithm to fit the time series and the number of anomalies, and get the anomaly frequency function as y=α0+α1×ti, where α0 represents the constant term and α1 represents the independent variable coefficient; use the least squares method to calculate the values of α0 and α1; S302, using the abnormal frequency function to calculate the rotation period, the formula is: △T ti =1 / y ti , where △T ti represents the rotation period when the time series is ti, y ti Indicates the frequency of abnormal occurrence when the time series is ti; S303, the encryption system calculates the corresponding rotation period according to the time series to generate a new shard private key, uses the new shard private key to replace the old shard private key, and uses the forward secure hash chain to update the hash value during the rotation. The formula is: H t =H(H t-1 |||Sign(SK t ,H t-1 )); In the formula, H t Represents the hash value at time window t, H t-1 Indicates the hash value at time window t-1, Sign(SK t , H t-1 ) indicates the use of private key SK t For H t-1 The result of signing.
5. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of determining whether each node has an abnormal risk and marking it in S400 are: S401. When the encryption system performs real-time shard encryption, collect the behavior data of each node, convert the collected behavior data into a feature vector x, and use the anomaly detection model to calculate the anomaly score of each node in real time. The formula is: In the formula, s i represents the abnormal score of the i-th node, E(h(x i )) represents the feature vector x of the i-th node i The average path length in multiple trees; c(n) represents the normalization factor; S402: Use the abnormal threshold of each node to judge the real-time abnormal score. i >Ex i When , it is judged that the i-th node has abnormal risk; when s i ≤Ex i When , it is judged that the i-th node does not have abnormal risk; after judging that the node has abnormal risk, it is marked.
6. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of dynamically adjusting the shard weight and threshold according to the abnormality score in S500 are: S501. Use the abnormal score to adjust the signature threshold. The formula is: In the formula, k new represents the new threshold after adjustment, k represents the old threshold before adjustment, represents the adjustment coefficient, m represents the total number of nodes; max(s i -Ex i , 0) means summing the anomaly scores exceeding the anomaly threshold; S502: After detecting an abnormal node, the encryption system generates a temporary time window t', reallocates shards, and performs an emergency rotation of shard private keys.
7. The multi-party collaborative signature management method based on key splitting technology according to claim 1 is characterized in that: The specific steps of performing collaborative signing in S600 are: S601: The client initiates a signature request with a time window identifier t, removes the node shard private key with abnormal risk, and extracts the private key greater than the new threshold k. new The shard private keys of each node are aggregated and signed using the Schnorr aggregate signature protocol.
8. A multi-party collaborative signature management system based on key splitting technology, characterized by: The multi-party collaborative signature management system includes a data collection module, an initialization module, a training module, a sharding private key module, a sharding rotation module, an anomaly detection module, a dynamic adjustment module and a signature module; The data collection module is used to collect node behavior data when anomalies occur in the encryption system history; The initialization module is used to initialize and define the encrypted data. The training module is used to train the anomaly detection model using the collected historical node behavior data when abnormal issuance occurs; The shard private key module is used for the system to generate a master private key and calculate the shard private key using a hash algorithm; The shard rotation module is used to use the number of anomalies in history to construct an anomaly frequency function based on the time series and calculate the rotation period of the corresponding time series; When a node is judged to have abnormal risks, emergency rotation is immediately carried out; The anomaly detection module is used to calculate the anomaly score during real-time encryption using the anomaly detection module, and to determine whether a node has an abnormal risk using an anomaly threshold; The dynamic adjustment module is used to dynamically adjust the signature threshold using the anomaly score; The signature module is used to perform signature using the adjusted new threshold.
9. The multi-party collaborative signature management system based on key splitting technology according to claim 8 is characterized in that: The slicing rotation module includes a periodic rotation unit and an emergency rotation unit; The periodic rotation unit is used to construct an abnormality occurrence frequency function based on a time series by using the number of abnormalities in history, calculate the rotation period of the corresponding time series, and perform shard rotation by using the rotation period; The emergency rotation unit is used to immediately perform emergency rotation when it is determined that there is an abnormal risk in the node.
10. The multi-party collaborative signature management system based on key splitting technology according to claim 8 is characterized in that: The training module includes a model training unit and an abnormal threshold calculation unit; The model training unit is used to train the anomaly detection model using the collected historical node behavior data at the time of abnormal issuance; The abnormality threshold calculation unit is used to extract the abnormality score during model training and calculate the abnormality threshold of each node.
Citation Information
Cited By
Collaborative signature method and device, equipment and medium
CN120639313A
Multi-user online collaborative distributed signing system
CN121077681A
Multi-user online collaborative distributed signing system
CN121077681B
Knowledge vector library-oriented secure storage and efficient query method
CN121412361A
A knowledge vector library-oriented secure storage and efficient query method
CN121412361B
