ABSTRACT
Abstract
A method involves receiving digital on blockchain information and digital off blockchain information. The digital data from the digital on blockchain information and the digital off blockchain information is extracted. The entity knowledge base engine contextualizes the relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine analyzes the digital data and transforms the digital data to an identified behavior category. The risk scoring regression engine analyzes the classified risk data and assigns a risk score to each classified risk data. The risk policy engine analyzes the classified risk data and determines if any deviations from rules or standards have or will occur. The security control system takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards.
Description
BACKGROUND
Cryptocurrency offers a decentralized system to exchanges funds in the form of digital assets. Due to the decentralized nature of the blockchain, tracing the origin of the funds and final destination of transactions can be difficult. However, as the value of cryptocurrency has risen, individuals and organizations have emerged that attempt to utilize this system to transfer funds to malicious entities or launder funds to hide the illicit origin of the funds. Due to the potential risks associated with cryptocurrency, regulations have been emplaced to prevent the money laundering and the funding of terrorism. Many legitimate cryptocurrency exchanges are required to comply with the regulations if they wish to continue to operate but struggle to identify transactions or accounts that may potentially violate these regulations placing them at risk of not complying with the regulations. Therefore, a need exists for a way to identify transactions and/or accounts that carry an increased risk of violating financial regulations.
BRIEF SUMMARY
A method of operating a risk management system for blockchain digital assets involves receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine. The method involves extracting digital data from the digital on blockchain information and the digital off blockchain information. In the method, an entity knowledge base contextualizes relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. A risk classification engine including a machine learning model, analyzes the digital data and transforms the digital data to an identified behavior category, thereby creating classified risk data. A risk scoring regression engine with machine learning analyzes the classified risk data and assigns a risk score to each classified risk data. A risk policy engine analyzes the classified risk data and determines if any deviations from rules or standards have or will occur, wherein the risk policy engine is a rules based engine. A security control system takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards. The action by the security control system includes, on condition the risk score is high, at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
A risk management system for blockchain digital assets comprises a digital asset intake engine, an entity knowledge base engine, a risk classification engine, a risk scoring regression engine, a risk policy engine, and a security control system. The digital asset intake engine is configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information. The entity knowledge base engine is configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine includes a machine learning model and is configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data. The risk scoring regression engine includes machine learning and is configured to analyze the classified risk data and assign a risk score to each classified risk data. The risk policy engine is a rules based engine. The security control system is configured to take an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score. The action by the security control system includes, on condition the risk score is high, at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS
To easily identify the discussion of any particular element or act, the most significant digit or digits in a reference number refer to the figure number in which that element is first introduced.
FIG. 1 illustrates a system 100 in accordance with one embodiment.
FIG. 2 illustrates a process 200 in accordance with one embodiment.
FIG. 3 illustrates a method 300 in accordance with one embodiment.
FIG. 4 illustrates method 400 in accordance with one embodiment.
FIG. 5 illustrates method 500 in accordance with one embodiment.
FIG. 6 illustrates a graph 600 in accordance with one embodiment.
FIG. 7 illustrates a graph 700 in accordance with one embodiment.
FIG. 8 illustrates an AutoML pipeline 800 in accordance with one embodiment.
FIG. 9 illustrates a basic deep neural network 900 in accordance with one embodiment.
FIG. 10 illustrates an artificial neuron 1000 in accordance with one embodiment.
FIG. 11 illustrates a decision tree 1100 in accordance with one embodiment.
FIG. 12 illustrates a graph 1200 showing a best fit line.
FIG. 13 displays a set graphs as examples of different learning rates.
FIG. 14 illustrates a blockchain transaction process 1400 in accordance with one embodiment.
FIG. 15 illustrates a blockchain formation 1500 in accordance with one embodiment.
FIG. 16 illustrates a blockchain 1600 in accordance with one embodiment.
FIG. 17 depicts an illustrative system architecture and data processing device 1700 that may be used in accordance with one or more illustrative aspects described herein.
DETAILED DESCRIPTION
A system and method for blockchain transaction risk management using machine learning provides security for individuals or entities that deal with blockchain related transactions. The system and method can make real time predictions and suspend a suspicious cryptocurrency transfer transaction from the user to an unknown wallet address that has observed strong linkage with a known terrorist. The system and method can detect suspicious behaviors based on historic transactions such as frequent attempts to exploit the blockchain systems, which will generate a high risk score to block this transaction.
Monitoring a decentralized system such as a blockchain can be difficult due to their nature. As the monetary value of the digital assets increases the need to identify suspicious transactions becomes more important to ensure the exchange is not being used to for money laundering or financing terrorism. The systems and methods in this disclosure allow for regulatory compliance on Anti-Money Laundering, and Counter Terrorism Financing (AML/CFT) by identifying risky transactions associated with a cryptocurrency exchange to block the transaction, as well as freeze or suspend the accounts.
Existing applications such as blacklisted blockchain addresses or sanctioned terrorists' personal information lack of adaptability to new emerging risks.
By leveraging machine learning to analyze the behaviors, combining on-blockchain and off-blockchain data sources, the methods and systems of this disclosure may be more accurate, have higher coverage, and may be more preventive.
A method of operating a risk management system for blockchain digital assets involves receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine. The method involves extracting digital data from the digital on blockchain information and the digital off blockchain information. In the method, an entity knowledge base contextualizes relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. A risk classification engine including a machine learning model, analyzes the digital data and transforms the digital data to an identified behavior category, thereby creating classified risk data. A risk scoring regression engine with machine learning analyzes the classified risk data and assigns a risk score to each classified risk data. A risk policy engine analyzes the classified risk data and determines if any deviations from rules or standards have or will occur, wherein the risk policy engine is a rules based engine. A security control system takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards. The action includes a security control system, where on condition the risk score is high, the action at least one of blocks a blockchain transaction, freezes user assets, or suspends user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
In some configurations, the digital on blockchain information and digital off blockchain information include blockchain address, transaction identification, user information, device information, device IP address, business type, and exchange or custodian determination. Custodians are third-parties that store digital assets for users.
In some configurations, the entity knowledge base includes a blacklist intelligence database, a device intelligence database, a computer network intelligence database, and a blockchain ledger.
In some configurations, the risk classification engine includes a decision tree classification model.
In some configurations the machine learning model includes at least one of a machine learning classification model and a risk score model to calculate the risk score.
In some configurations, the machine learning classification model is an AutoML model. The operation of the AutoML model involves preparing a labeled dataset, pre-processing the labeled dataset, extracting AutoML features, transforming AutoML features, training an AutoML model, evaluating metrics of the AutoML model, selecting a best machine learning model using an automated selection process, and serializing the best machine learning model.
In some configurations, the AutoML model may be utilized as an offline training pipeline or an online prediction pipeline in the cloud or a decentralized blockchain node. The offline training pipeline may involve feature extraction and transformation, parallel model training, model metric evaluation, and model selection. The online prediction pipeline may involve feature extraction and transformation, model prediction, and result formatting.
In some configurations, the machine learning classification model is a behavior based model and may involve recognizing behavioral characteristics in at least one feature category for an entity. These feature categories include a statistics feature category, a topology feature category, a temporal feature category, a temporal feature category, a linkage feature category, a derived feature category, a sequential feature category. The machine learning classification model may also identify exchange behavioral wallet addresses using exchange behavior characteristics, a money laundering behavioral address, a bot behavioral address, and a bad actor group cluster.
In some configurations, the machine learning classification model is a regression model for the risk score. The regression model may include input features comprising at least one of a blockchain transaction and an external information related to blockchain addresses. The regression model may also include parameters including a time decaying factor lambda. The regression model may also include an output including risk score of the address, reasons for the prediction, and a suspicious transaction summary.
A risk management system for blockchain digital assets comprises a digital asset intake engine, an entity knowledge base engine, a risk classification engine, a risk scoring regression engine, a risk policy engine, and a security control system. The digital asset intake engine is configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information. The entity knowledge base engine is configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine includes a machine learning model and is configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data. The risk scoring regression engine includes machine learning and is configured to analyze the classified risk data and assign a risk score to each classified risk data. The risk policy engine is a rules based engine. The security control system is configured to take an action on the digital on blockchain information and digital off blockchain information based on the assign
BACKGROUND
Cryptocurrency offers a decentralized system to exchanges funds in the form of digital assets. Due to the decentralized nature of the blockchain, tracing the origin of the funds and final destination of transactions can be difficult. However, as the value of cryptocurrency has risen, individuals and organizations have emerged that attempt to utilize this system to transfer funds to malicious entities or launder funds to hide the illicit origin of the funds. Due to the potential risks associated with cryptocurrency, regulations have been emplaced to prevent the money laundering and the funding of terrorism. Many legitimate cryptocurrency exchanges are required to comply with the regulations if they wish to continue to operate but struggle to identify transactions or accounts that may potentially violate these regulations placing them at risk of not complying with the regulations. Therefore, a need exists for a way to identify transactions and/or accounts that carry an increased risk of violating financial regulations.
BRIEF SUMMARY
A method of operating a risk management system for blockchain digital assets involves receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine. The method involves extracting digital data from the digital on blockchain information and the digital off blockchain information. In the method, an entity knowledge base contextualizes relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. A risk classification engine including a machine learning model, analyzes the digital data and transforms the digital data to an identified behavior category, thereby creating classified risk data. A risk scoring regression engine with machine learning analyzes the classified risk data and assigns a risk score to each classified risk data. A risk policy engine analyzes the classified risk data and determines if any deviations from rules or standards have or will occur, wherein the risk policy engine is a rules based engine. A security control system takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards. The action by the security control system includes, on condition the risk score is high, at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
A risk management system for blockchain digital assets comprises a digital asset intake engine, an entity knowledge base engine, a risk classification engine, a risk scoring regression engine, a risk policy engine, and a security control system. The digital asset intake engine is configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information. The entity knowledge base engine is configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine includes a machine learning model and is configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data. The risk scoring regression engine includes machine learning and is configured to analyze the classified risk data and assign a risk score to each classified risk data. The risk policy engine is a rules based engine. The security control system is configured to take an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score. The action by the security control system includes, on condition the risk score is high, at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
BRIEF DESCRIPTION OF THE SEVERAL VIEWS OF THE DRAWINGS
To easily identify the discussion of any particular element or act, the most significant digit or digits in a reference number refer to the figure number in which that element is first introduced.
FIG. 1 illustrates a system 100 in accordance with one embodiment.
FIG. 2 illustrates a process 200 in accordance with one embodiment.
FIG. 3 illustrates a method 300 in accordance with one embodiment.
FIG. 4 illustrates method 400 in accordance with one embodiment.
FIG. 5 illustrates method 500 in accordance with one embodiment.
FIG. 6 illustrates a graph 600 in accordance with one embodiment.
FIG. 7 illustrates a graph 700 in accordance with one embodiment.
FIG. 8 illustrates an AutoML pipeline 800 in accordance with one embodiment.
FIG. 9 illustrates a basic deep neural network 900 in accordance with one embodiment.
FIG. 10 illustrates an artificial neuron 1000 in accordance with one embodiment.
FIG. 11 illustrates a decision tree 1100 in accordance with one embodiment.
FIG. 12 illustrates a graph 1200 showing a best fit line.
FIG. 13 displays a set graphs as examples of different learning rates.
FIG. 14 illustrates a blockchain transaction process 1400 in accordance with one embodiment.
FIG. 15 illustrates a blockchain formation 1500 in accordance with one embodiment.
FIG. 16 illustrates a blockchain 1600 in accordance with one embodiment.
FIG. 17 depicts an illustrative system architecture and data processing device 1700 that may be used in accordance with one or more illustrative aspects described herein.
DETAILED DESCRIPTION
A system and method for blockchain transaction risk management using machine learning provides security for individuals or entities that deal with blockchain related transactions. The system and method can make real time predictions and suspend a suspicious cryptocurrency transfer transaction from the user to an unknown wallet address that has observed strong linkage with a known terrorist. The system and method can detect suspicious behaviors based on historic transactions such as frequent attempts to exploit the blockchain systems, which will generate a high risk score to block this transaction.
Monitoring a decentralized system such as a blockchain can be difficult due to their nature. As the monetary value of the digital assets increases the need to identify suspicious transactions becomes more important to ensure the exchange is not being used to for money laundering or financing terrorism. The systems and methods in this disclosure allow for regulatory compliance on Anti-Money Laundering, and Counter Terrorism Financing (AML/CFT) by identifying risky transactions associated with a cryptocurrency exchange to block the transaction, as well as freeze or suspend the accounts.
Existing applications such as blacklisted blockchain addresses or sanctioned terrorists' personal information lack of adaptability to new emerging risks.
By leveraging machine learning to analyze the behaviors, combining on-blockchain and off-blockchain data sources, the methods and systems of this disclosure may be more accurate, have higher coverage, and may be more preventive.
A method of operating a risk management system for blockchain digital assets involves receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine. The method involves extracting digital data from the digital on blockchain information and the digital off blockchain information. In the method, an entity knowledge base contextualizes relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. A risk classification engine including a machine learning model, analyzes the digital data and transforms the digital data to an identified behavior category, thereby creating classified risk data. A risk scoring regression engine with machine learning analyzes the classified risk data and assigns a risk score to each classified risk data. A risk policy engine analyzes the classified risk data and determines if any deviations from rules or standards have or will occur, wherein the risk policy engine is a rules based engine. A security control system takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards. The action includes a security control system, where on condition the risk score is high, the action at least one of blocks a blockchain transaction, freezes user assets, or suspends user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
In some configurations, the digital on blockchain information and digital off blockchain information include blockchain address, transaction identification, user information, device information, device IP address, business type, and exchange or custodian determination. Custodians are third-parties that store digital assets for users.
In some configurations, the entity knowledge base includes a blacklist intelligence database, a device intelligence database, a computer network intelligence database, and a blockchain ledger.
In some configurations, the risk classification engine includes a decision tree classification model.
In some configurations the machine learning model includes at least one of a machine learning classification model and a risk score model to calculate the risk score.
In some configurations, the machine learning classification model is an AutoML model. The operation of the AutoML model involves preparing a labeled dataset, pre-processing the labeled dataset, extracting AutoML features, transforming AutoML features, training an AutoML model, evaluating metrics of the AutoML model, selecting a best machine learning model using an automated selection process, and serializing the best machine learning model.
In some configurations, the AutoML model may be utilized as an offline training pipeline or an online prediction pipeline in the cloud or a decentralized blockchain node. The offline training pipeline may involve feature extraction and transformation, parallel model training, model metric evaluation, and model selection. The online prediction pipeline may involve feature extraction and transformation, model prediction, and result formatting.
In some configurations, the machine learning classification model is a behavior based model and may involve recognizing behavioral characteristics in at least one feature category for an entity. These feature categories include a statistics feature category, a topology feature category, a temporal feature category, a temporal feature category, a linkage feature category, a derived feature category, a sequential feature category. The machine learning classification model may also identify exchange behavioral wallet addresses using exchange behavior characteristics, a money laundering behavioral address, a bot behavioral address, and a bad actor group cluster.
In some configurations, the machine learning classification model is a regression model for the risk score. The regression model may include input features comprising at least one of a blockchain transaction and an external information related to blockchain addresses. The regression model may also include parameters including a time decaying factor lambda. The regression model may also include an output including risk score of the address, reasons for the prediction, and a suspicious transaction summary.
A risk management system for blockchain digital assets comprises a digital asset intake engine, an entity knowledge base engine, a risk classification engine, a risk scoring regression engine, a risk policy engine, and a security control system. The digital asset intake engine is configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information. The entity knowledge base engine is configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine includes a machine learning model and is configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data. The risk scoring regression engine includes machine learning and is configured to analyze the classified risk data and assign a risk score to each classified risk data. The risk policy engine is a rules based engine. The security control system is configured to take an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score. The action by the security control system includes, on condition the risk score is high, at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction. On condition the risk score is normal, the action approves the blockchain transaction.
In some configurations, the digital asset intake engine includes blockchain address, transaction identification, user information, device information, device IP address, business type, exchange or custodian.
In some configurations, the entity knowledge base includes a blacklist intelligence database, a device intelligence database, a computer network intelligence database, and a blockchain ledger.
In some configurations, the knowledge base engine updates the information upon receiving threat intelligence to include a reentrancy vulnerability pattern.
In some configurations, the risk classification engine includes a decision tree classification model.
In some configurations, the risk scoring regression engine includes a machine learning classification model and a risk score model to calculate the risk score.
In some configurations, the machine learning model includes at least one of a machine learning classification model and a risk score model to calculate the risk score.
In some configurations, the machine learning classification model is an AutoML model involves preparing a labeled dataset, pre-processing the labeled dataset, extracting AutoML features, transforming AutoML features, training an AutoML model, evaluating metrics of the AutoML model, selecting a best machine learning model using an automated selection process, and serializing the best machine learning model.
In some configurations, the AutoML model is at least one of an offline training pipeline or an online prediction pipeline in the cloud or a decentralized blockchain node. The offline training pipeline involves feature extraction and transformation, parallel model training, model metric evaluation, and model selection. The online prediction pipeline involves feature extraction and transformation, model prediction, and result formatting.
In some configurations, machine learning classification model is a behavior based model comprising recognizing behavioral characteristics in at least one feature category for an entity and identifying a set of behaviors. The feature categories include a statistics feature category, a topology feature category, a temporal feature category, a temporal feature category, a linkage feature category, a derived feature category, and a sequential feature category. The set of behaviors includes exchange behavioral wallet addresses using exchange behavior characteristics, a money laundering behavioral address, a bot behavioral address, and a bad actor group cluster.
In some configurations, the machine learning classification model is a regression model for the risk score. The regression model includes input features comprising at least one of blockchain transactions and external information related to blockchain addresses. The regression model includes parameters including a time decaying factor lambda. The regression model includes outputs such as risk score of the address, reasons for the prediction, and suspicious transaction summary.
FIG. 1 illustrates a system 100 for blockchain transaction risk management. The system 100 comprises a digital asset intake engine 108 , a risk classification engine 102 , a risk scoring regression engine 104 , a risk policy engine 148 , a security control system 106 , and an entity knowledge base engine 110 . The risk classification engine 102 comprises a machine learning model 150 . The risk scoring regression engine 104 comprises a machine learning algorithm 152 . The entity knowledge base engine 110 comprise a black list intelligence database 112 , a device intelligence database 114 , a computer network intelligence database 116 , and a blockchain ledger 118 .
The digital asset intake engine 108 is configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information. The entity knowledge base engine 110 is configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information. The risk classification engine 102 includes a machine learning model and is configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data. The risk scoring regression engine 104 includes machine learning and is configured to analyze the classified risk data and assign a risk score to each classified risk data.
The risk policy engine 148 is a rules based engine. The security control system 106 configured to take an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score.
In the system 100 , a user 120 performs a set of user actions 122 comprising a withdraw 124 , a deposit 126 , a swap 128 , and/or a transfer 130 , of funds or cryptocurrency. These actions are understood as digital on blockchain information and digital off blockchain information and are received by a digital asset intake engine 108 . The digital asset intake engine 108 extracts data (extracted data 132 ), which pulls out the digital data and the digital off blockchain information and the digital on blockchain information that includes a blockchain address 134 , a transaction identification 136 , a user information 138 , a device information 140 , a business type 146 , and a device IP address 142 , as well as the exchange or custodian information 144 , associated with the user 120 's actions. The data pulled by the digital asset intake engine 108 may be obtained from the cryptocurrency exchange where the user actions 122 were performed or other sources that track these actions. The extracted digital data and the digital off blockchain information and the digital on blockchain information is then contextualized by a risk classification engine 102 , which leverages information stored in entity knowledge bases that include black list intelligence database 112 , a device intelligence database 114 , a computer network intelligence database 116 , and a blockchain ledger 118 . The risk classification engine 102 analyzes the information stored in the entity knowledge bases to transform the digital data into an identified behavior category, creating classified risk data. The classified risk data is then communicated to a risk scoring regression engine 104 that analyzes the classified risk data and assigns a risk score to each classified risk data. The classified risk data is then communicated to the risk policy engine 148 , which is a rules based engine. The classified risk data is then analyzed by the risk policy engine 148 to determine if any deviations from rules or standards have or will occur. The security control system 106 takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards. For example, if the risk is high, the security control system 106 may block, freeze, or suspend the transaction and/or the account that is performing the actions. If the risk is viewed as normal, the security control system 106 may approve the transaction.
FIG. 2 illustrates a flow chart for process 200 describing machine learning (ML) Entity categorization and risk engine processes. In the process 200 , blockchain address and device information/ IP 218 is received following user action on the blockchain or associated with the blockchain on a cryptocurrency exchange. At decision block 202 , the blockchain address and device information/ IP 218 is analyzed against a blacklist database 210 and a determination is made whether the blockchain address and device information/ IP 218 have been blacklisted from operating on the exchange. If the blockchain address and device information/ IP 218 has been blacklisted according to the blacklist database 210 , the process 200 moves to reject 204 the user action on the exchange. If the blockchain address and device information/ IP 218 has not been blacklisted the process moves to block 222 where a machine learning prediction is performed. The machine learning prediction utilizes an entity knowledge base comprising an identity verification database 212 , a computer network intelligence database 214 , a device intelligence database 216 , and the blockchain ledger 220 to predict entity category 224 and predict risk score 226 . The process 200 then moves to decision block 206 that determines if the predicted risk score is high. If the risk score is high, the process 200 moves to reject 204 the user action. If the risk score is not high, the process 200 moves to accept 208 the user action.
FIG. 3 illustrates a method 300 for managing risk to a block utilizing machine learning. The method 300 involves receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine (block 302 ). In block 304 , the method 300 extracts digital data from the digital on blockchain information and the digital off blockchain information. In block 306 , the method 300 contextualizes relationships based on the digital data and the digital off blockchain information and the digital on blockchain information, the contextualizing includes an entity knowledge base. In block 308 , the method 300 analyzes the digital data and transforms the digital data to an identified behavior category, thereby creating classified risk data, wherein the analyzing and transforming includes a risk classification engine including a machine learning model. In block 310 , the method 300 analyzes the classified risk data and assigns a risk score to each classified risk data, wherein the analyzing and assigning includes a risk scoring regression engine and machine learning. In block 312 , the method 300 analyzes the classified risk data and determines if any deviations from rules or standards have or will occur, wherein the analyzing and determining includes a risk policy engine that is a rules based engine. In block 314 , the method 300 takes an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards, wherein taking the action includes a security control system.
FIG. 4 illustrates a method 400 for training the machine learning classification model in accordance with on embodiment. The method 400 involves preparing a labeled dataset (block 402 ). The in block 404 , the method 400 pre-process the labeled dataset. In block 406 , the method 400 configures the AutoML Parallel Training System 418 with parameters and success criteria settings. The method 400 then moves to block 408 where the AutoML features are extracted and transformed. In block 410 , the method 400 trains the AutoML model. This training may occur in parallel. In block 412 , the method 400 evaluates the metrics of the AutoML model. In block 414 , the method 400 selects a best machine learning model using an automated selection process. In the block 416 , the method 400 serializes the best machine learning model.
An example of the prepared labeled dataset may be a dataset that includes information such as: A is a hacker address with 100 transactions; B is an exchange wallet address with 2000 transactions.
Pre-processing the labeled data may involve cleaning the data, such removing null values, outliers and normalizing the data into the form for AutoML.
The AutoML parameters and criteria setting may include a computing resource setting, such as how many CPU/RAM/Disks allowed, time setting, such as how much time allowed for training the best ML model, and establishing a success criteria for the winning ML model, such as: AUC >0.99? Precision >95%.
An example of the AutoML feature extraction may be performing a PCA (probably approximately correct learning) for dimensionality reduction. An example of AutoML feature transformation may be a logarithm transformation.
During the AutoML Model training, the selected machine learning model will be run in parallel, and compete to meet the success criteria. The AutoML model training supports all supervised/semi-supervised machine learning models, both Classification and Regression. These models include Tree ensemble models such as Gradient Boosting Trees and Random Forest, GLM Generalized linear models, Support Vector Machines, and Deep learning models such as Convolutional Neural Nets, LSTM, RNN, etc.
During the AutoML Model metric evaluation, any machine learning quantitative evaluation metrics can be adopted. These metrics include Precision, Recall, Accuracy, AUC: Area Under ROC, and F1 scores.
During the AutoML Automated model selection may be determined by criteria of selecting the winning models. In some cases when a time out occurs, the best ML model will still be selected.
The serialization of the best ML model may be done for version control. The winning machine learning model may be selected, serialized into a file, and will thus be ready to be deployed into a prediction pipeline.
In some configurations, the AutoML model may be an offline training pipeline or an online prediction pipeline in the cloud or a decentralized blockchain node. The offline training pipeline includes feature extraction and transformation, parallel model training, model metric evaluation, and model selection. The online prediction pipeline includes feature extraction and transformation, model prediction, and result formatting.
Automated machine learning (AutoML), is a system and methodology that automates various stages of the machine learning process, such as model selection, hyperparameter optimization, etc. The AutoML system takes the labeled data as input, runs a parallel competition to select the best machine learning model that meets the success criteria, and eventually emits a serialized machine learning model that can be deployed in the prediction pipeline.
The advantage of using AutoML is mostly about efficiency. AutoML helps creates high quality ML models quickly, while using minimal data science labors.
AutoML model can be modified to operate similarly to a traditional machine learning model, by limiting the compute resources. For example, AutoML can be configured to use only 1 CPU and train 1 model (say Gradient Boosting Tree) at a time, which is actually a traditional machine learning model.
FIG. 5 illustrates a method 500 for operating the AutoML model in accordance with one embodiment. The method 500 involves inputting a data feed (block 502 ). In block 504 , the method 500 preprocess the inputs from the data feed. The preprocessing may involve cleaning the data, such removing null values, outliers and normalizing the data into the form for AutoML. The preprocessed data is then sent to the online prediction pipeline 514 . The online prediction pipeline 514 is in the cloud or a decentralized blockchain node. In block 506 , the method 500 performs feature extraction and transformation. An example of the AutoML feature extraction may be performing a PCA for dimensionality reduction. An example of AutoML feature transformation may be a logarithm transformation. In block 508 , the method 500 performs a model prediction based on one or more models. In block 510 , the method 500 formats the results. In block 512 , the method 500 exports the machine learning prediction results out of the online prediction pipeline 514 .
The online prediction pipeline may be deployed as a de-serialized model file, to the cloud, or through decentralized blockchain deployment. The online prediction pipeline may be utilized to perform data preprocessing, feature extraction and transformation, predictions, and generate results.
The machine learning classification model for blockchain behavioral entities may be utilized to classify a blockchain address into an entity category, based on blockchain transactions, intelligence databases, and other related info.
The machine learning classification model for blockchain behavioral entities may be applied to any blockchain behavioral entities that include crypto exchanges, crypto mixers, bots, hackers, terrorists, and smart contract based decentralized applications.
The advantage of the machine learning classification model is that it can learn from the transaction data mostly by recognizing the specific behavioral characteristics of each entity category. In contrast, the traditional methods maintain a database for addresses with known entities, which has low coverage.
The machine learning model may include feature categories for describing behavior characteristics. These categories include a statistics feature category, a topology feature category, a temporal feature category, a linkage feature category, a derived feature category, and a sequential feature category.
The statistics feature category may include items such as counts of inbound and outbound transactions, a sum of inbound and outbound amounts, maximum transaction amounts, etc.
The topology feature category may include items such as the number of input addresses, graph centrality, etc.
The temporal feature category may include items such as the number of active hours per day, the number of active Mondays in a year, the entropy of the time series range, etc.
The linkage feature category may include items such as the number of connections to known entities in the intelligence database, such as Binance exchange wallets, etc., and the amount sent to known entities, such as exchange wallets, etc.
The derived feature category may include items such as the ratio between outbound and inbound amounts, the ratio between the exchange amount and the hacker amount, etc.
The sequential feature category may include items such as the time interval between 2 consecutive transactions of one address, the percentage of transactions that have the same dollar amount within each rolling time window, etc.
The machine learning model may be utilized to identify exchange behavior addresses. A crypto exchange may have wallet addresses stored as part of their on-blockchain infrastructure. These wallets may also include designations categorizing them as cold or hot deposit wallets. Each cryptocurrency exchange may have different characteristics for making these classifications. Through the use of the machine learning model, predictions can be made to determine and identify exchange behaviors and characteristics associated with the addresses. These exchange behaviors may include a transaction funding amount that is usually substantially large, whether hot wallets tend to be more active, generating a significant amount of transactions, whether hot wallets and cold wallets tend to cluster together, and if there are consistent temporal traits that observe high autocorrelation factors.
The machine learning model may be utilized to identify addresses associated with behaviors such as mixer, tumbler, and money laundering. The mixer, tumbler, or money laundering behavior may be characterized by the fact that most addresses are âpass throughâ or have âtransientâ behaviors, such that the funding only stays in a short time interval until all the funds are wired out. Another characterization may be if there are thrown away addresses, or if they are recycled but with repetitive transient patterns. Another characterization may be identifying behavioral traits depending on the blockchain. For example, on Bitcoin there can be certain multisig patterns that can identify the mixers' encryption patterns.
The machine learning model may be utilized to identify bot behavioral addresses. Bot behavior may be characterized by their activity or their active percentage, where the more activity, the higher likelihood the that the actions are being performed by a bot. Another characterization may be done by looking at temporal regularity by auto-correlation such as if there are repeated patterns that hint at bot behavior. Another characterization may be transaction amount distribution entropy. For example, a bot is more likely to bet at a fixed amount, hence low entropy.
When the risk classification engine is informed with suspicious activities, from news or threat researchers, a seed anomaly address can be identified. For example, a group of bad actors may utilize a particular exploit, where if the machine learning model is updated with details regarding the exploit, the machine learning model may identify a cluster of addresses that benefit from the exploit and/or are associated with addresses associated with the group of bad actors.
FIG. 6 illustrates a graph 600 describing the risk score distribution. A machine learning regression model may be utilized to generate a risk score by the risk classification engine. The risk classification engine may utilize blockchain entity categorization/classification as part of the inputs to generate the risk score.</d
CLAIMS
Claims ( 20 )
What is claimed is:
1. A method comprising:
receiving digital on blockchain information and digital off blockchain information, wherein the receiving includes a digital asset intake engine;
extracting digital data from the digital on blockchain information and the digital off blockchain information;
contextualizing relationships based on the digital data and the digital off blockchain information and the digital on blockchain information, the contextualizing including an entity knowledge base;
analyzing the digital data and transforming the digital data to an identified behavior category, thereby creating classified risk data, wherein the analyzing and transforming includes a risk classification engine including a machine learning model;
analyzing the classified risk data and assigning a risk score to each classified risk data, wherein the analyzing and assigning includes a risk scoring regression engine and machine learning;
analyzing the classified risk data and determining if any deviations from rules or standards have or will occur, wherein the analyzing and determining includes a risk policy engine that is a rules based engine; and
taking an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score and any deviations from rules or standards, wherein taking the action includes a security control system, wherein the action includes:
on condition the risk score is high:
at least one of blocking a blockchain transaction, freezing user assets, or suspending user accounts related to the blockchain transaction; and
on condition the risk score is normal:
approving the blockchain transaction.
2. The method of claim 1 , wherein the digital on blockchain information and digital off blockchain information includes:
a blockchain address;
a transaction identification;
user information;
device information;
a device IP address;
a business type; and
an exchange or custodian distinction.
3. The method of claim 1 , wherein the entity knowledge base includes:
a blacklist intelligence database;
a device intelligence database;
a computer network intelligence database; and
a blockchain ledger.
4. The method of claim 1 , wherein the risk classification engine includes:
a decision tree classification model.
5. The method of claim 1 , wherein the machine learning model includes at least one of:
a machine learning classification model; and
a risk score model to calculate the risk score.
6. The method of claim 5 , wherein the machine learning classification model is an AutoML model comprising:
preparing a labeled dataset;
pre-processing the labeled dataset;
extracting AutoML features;
transforming AutoML features;
training an AutoML model;
evaluating metrics of the AutoML model;
selecting a best machine learning model using an automated selection process; and
serializing the best machine learning model.
7. The method of claim 6 , wherein the AutoML model is at least one of:
an offline training pipeline including:
feature extraction and transformation;
parallel model training;
model metric evaluation; and
model selection;
an online prediction pipeline in a cloud or a decentralized blockchain node, the online prediction pipeline including:
feature extraction and transformation;
model prediction based on one or more models; and
result formatting.
8. The method of claim 1 , wherein the machine learning classification model is a behavior based model comprising:
recognizing behavioral characteristics in at least one feature category for an entity, wherein the feature categories include:
a statistics feature category;
a topology feature category;
a temporal feature category;
a linkage feature category;
a derived feature category; and
a sequential feature category;
identifying at least one of:
exchange behavioral wallet addresses using exchange behavior characteristics;
a money laundering behavioral address;
a bot behavioral address; and
a bad actor group cluster.
9. The method of claim 5 , wherein the machine learning classification model is a regression model for the risk score, the regression model including:
input features comprising at least one of:
blockchain transactions including blockchain addresses; and
external information related to blockchain addresses;
parameters including a time decaying factor lambda;
output including:
the risk scores of the blockchain addresses;
reasons for the risk scores; and
a suspicious transaction summary.
10. A risk management system for blockchain digital assets, the system comprising:
a digital asset intake engine, configured to receive digital on blockchain information and digital off blockchain information, and extract digital data from the digital on blockchain information and the digital off blockchain information;
an entity knowledge base engine configured to contextualize relationships based on the digital data and the digital off blockchain information and the digital on blockchain information;
a risk classification engine including a machine learning model, the risk classification engine configured to analyze the digital data and transform the digital data to an identified behavior category, thereby creating classified risk data;
a risk scoring regression engine including machine learning, the risk scoring regression engine configured to analyze the classified risk data and assign a risk score to each classified risk data;
a risk policy engine, wherein the risk policy engine is a rules based engine; and
a security control system configured to take an action on the digital on blockchain information and digital off blockchain information based on the assigned risk score, wherein the security control system is configured to:
on condition the risk score is high:
at least one of block a blockchain transaction, freeze user assets, or suspend user accounts related to the blockchain transaction; and
on condition the risk score is normal:
approve the blockchain transaction.
11. The risk management system of claim 10 , wherein the digital asset intake engine includes:
a blockchain address;
a transaction identification;
user information;
device information;
a device IP address;
a business type; and
an exchange or custodian distinction.
12. The risk management system of claim 10 , wherein the entity knowledge base includes:
a blacklist intelligence database;
a device intelligence database;
a computer network intelligence database; and
a blockchain ledger.
13. The risk management system of claim 10 , wherein the entity knowledge base engine updates the digital on blockchain information and digital off blockchain information upon receiving threat intelligence to include a reentrancy vulnerability pattern.
14. The risk management system of claim 10 , wherein the risk classification engine includes:
a decision tree classification model.
15. The risk management system of claim 10 , wherein the risk scoring regression engine includes:
a machine learning classification model; and
a risk score model to calculate the risk score.
16. The risk management system of claim 10 , wherein the machine learning model includes at least one of:
a machine learning classification model; and
a risk score model to calculate the risk score.
17. The risk management system of claim 16 , wherein the machine learning classification model is an AutoML model comprising:
preparing a labeled dataset;
pre-processing the labeled dataset;
extracting AutoML features;
transforming AutoML features;
training an AutoML model;
evaluating metrics of the AutoML model;
selecting a best machine learning model using an automated selection process; and
serializing the best machine learning model.
18. The risk management system of claim 17 , wherein the AutoML model is at least one of:
an offline training pipeline including:
feature extraction and transformation;
parallel model training;
model metric evaluation; and
model selection;
an online prediction pipeline in a cloud or a decentralized blockchain node, the online prediction pipeline including:
feature extraction and transformation;
model prediction based on one or more models; and
result formatting.
19. The risk management system of claim 10 , wherein the machine learning classification model is a behavior based model comprising:
recognizing behavioral characteristics in at least one feature category for an entity, wherein the feature categories include:
a statistics feature category;
a topology feature category;
a temporal feature category;
a linkage feature category;
a derived feature category; and
a sequential feature category;
identifying at least one of:
exchange behavioral wallet addresses using exchange behavior characteristics;
a mixer behavioral address;
a tumbler behavioral address;
a money laundering behavioral address;
a bot behavioral address; and
a bad actor group cluster.
20. The risk management system of claim 10 , wherein the machine learning classification model is a regression model for the risk score, the regression model including:
input features comprising at least one of:
blockchain transactions including blockchain addresses; and
external information related to blockchain addresses;
parameters including a time decaying factor lambda;
output including:
the risk scores of the blockchain addresses;
reasons for the risk scores; and
a suspicious transaction summary.
US17/006,431
2020-08-28
2020-08-28
System and method for blockchain transaction risk management using machine learning
Active
2041-04-24
US11436615B2
( en )
Priority Applications (2)
Application Number
Priority Date
Filing Date
Title
US17/006,431
US11436615B2
( en )
2020-08-28
2020-08-28
System and method for blockchain transaction risk management using machine learning
US17/357,029
US11907955B2
( en )
2020-08-28
2021-06-24
System and method for blockchain automatic tracing of money flow using artificial intelligence
Applications Claiming Priority (1)
Application Number
Priority Date
Filing Date
Title
US17/006,431
US11436615B2
( en )
2020-08-28
2020-08-28
System and method for blockchain transaction risk management using machine learning
Related Child Applications (1)
Application Number
Title
Priority Date
Filing Date
US17/357,029
Continuation-In-Part
US11907955B2
( en )
2020-08-28
2021-06-24
System and method for blockchain automatic tracing of money flow using artificial intelligence
Publications (2)
Publication Number
Publication Date
US20220067752A1
US20220067752A1 ( en )
2022-03-03
US11436615B2
true
US11436615B2 ( en )
2022-09-06
Family
ID=80358708
Family Applications (1)
Application Number
Title
Priority Date
Filing Date
US17/006,431
Active
2041-04-24
US11436615B2
( en )
2020-08-28
2020-08-28
System and method for blockchain transaction risk management using machine learning
Country Status (1)
Country
Link
US
( 1 )
US11436615B2
( en )
Cited By (2)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20210097383A1
( en )
*
2019-09-30
2021-04-01
International Business Machines Corporation
Combined Data Pre-Process And Architecture Search For Deep Learning Models
US20250045361A1
( en )
*
2023-07-31
2025-02-06
Fireblocks Ltd.
Protection of digital assets within the blockchain by approval revocations
Families Citing this family (58)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US11507540B1
( en )
*
2018-02-15
2022-11-22
EMC IP Holding Company LLC
Decentralized file system and message bus architecture for processing training sets in multi-cloud computing environment
US11569981B1
( en )
*
2018-08-28
2023-01-31
Amazon Technologies, Inc.
Blockchain network based on machine learning-based proof of work
US11526614B2
( en )
*
2019-10-15
2022-12-13
Anchain.ai Inc.
Continuous vulnerability management system for blockchain smart contract based digital asset using sandbox and artificial intelligence
US11379844B2
( en )
*
2020-06-05
2022-07-05
Elementus Inc.
Systems and methods for quantifying and electronically displaying degrees of association between blockchain addresses
US12242980B2
( en )
*
2020-09-09
2025-03-04
International Business Machines Corporation
Machine learning with multiple constraints
US12131234B2
( en )
*
2020-10-28
2024-10-29
Capital One Services, Llc
Code generation for deployment of a machine learning model
US20220164798A1
( en )
*
2020-11-20
2022-05-26
Royal Bank Of Canada
System and method for detecting fraudulent electronic transactions
US12380358B2
( en )
*
2021-01-25
2025-08-05
Paypal, Inc.
Systems and methods for training and modifying a computer-based model to perform classification
CN113177791A
( en )
*
2021-04-23
2021-07-27
æå·å®æä¿¡æ¯ææ¯è¡ä»½æéå ¬å¸
Malicious mining behavior identification method, device, equipment and storage medium
US11853278B2
( en )
2021-06-10
2023-12-26
Jpmorgan Chase Bank , N.A.
Systems and methods for combining graph embedding and random forest classification for improving classification of distributed ledger activities
US11943244B2
( en )
*
2021-06-22
2024-03-26
International Business Machines Corporation
Anomaly detection over high-dimensional space
US20230114013A1
( en )
*
2021-10-12
2023-04-13
International Business Machines Corporation
Enhanced machine learning pipelines with multiple objectives and tradeoffs
US12259976B1
( en )
*
2021-11-23
2025-03-25
HackerStrike Corporation
Detecting ransomware attacks in a device based on device behavior
CN114581244A
( en )
*
2022-03-09
2022-06-03
æ·±å³ææ¯å¤§å¦
Dynamic intelligent delivery and review method based on block chain intelligent contract and related device
US12182807B2
( en )
2022-03-14
2024-12-31
Paypal, Inc.
Interface for constructing smart protocols for execution on blockchain platforms
US12051066B2
( en )
*
2022-03-15
2024-07-30
Capital One Services, Llc
Systems and methods for validating asset destinations in blockchain networks
JP7845074B2
( en )
*
2022-06-17
2026-04-14
æ ªå¼ä¼ç¤¾å³¶æ´¥è£½ä½æ
Data analysis method, data analysis system, and server for data analysis system
US12182800B2
( en )
*
2022-06-23
2024-12-31
The Toronto-Dominion Bank
API for incremental and periodic crypto asset transfer
US20230419326A1
( en )
*
2022-06-28
2023-12-28
Ancilia, Inc.
Security detection for blockchains
CN115130883A
( en )
*
2022-07-05
2022-09-30
å京åå¸æ³°æ ¼ä¿¡æ¯ç§ææéå ¬å¸
Address credit risk assessment method and system based on block chain
CN114943309B
( en )
*
2022-07-21
2022-10-21
人æ°æ³é¢ä¿¡æ¯ææ¯æå¡ä¸å¿
Method for constructing abnormity diagnosis model of block chain and abnormity diagnosis method
WO2024081977A1
( en )
*
2022-10-11
2024-04-18
Mnthali Bado Mtazama Jembemziro
A risk management system
CN115643267B
( en )
*
2022-10-12
2024-11-12
åè¥å·¥ä¸å¤§å¦
Method and blockchain system for dynamically improving blockchain performance based on network trust model
US12619693B2
( en )
*
2022-10-18
2026-05-05
Bank Of America Corporation
User credential authentication using blockchain and machine learning
CN115659286B
( en )
*
2022-10-27
2025-10-21
西å®çµåç§æ大å¦
A code hosting platform risk control system and method based on deep learning
WO2024161335A1
( en )
*
2023-02-02
2024-08-08
Pasha Azam
A computer-implemented system and method for blockchain-enabled documentation of trade transactions
AU2024229742A1
( en )
*
2023-03-02
2025-09-25
Fluence Energy, Llc
Systems and methods for using machine learning to predict critical constraints
CN116405187B
( en )
*
2023-04-21
2024-04-09
ç³å®¶åºéé大å¦
Distributed node intrusion situation sensing method based on block chain
CN116433333B
( en )
*
2023-04-27
2024-01-30
æ¦æ±åæ¦æ°åç§ææéå ¬å¸
Digital commodity transaction risk prevention and control method and device based on machine learning
CN116402512B
( en )
*
2023-05-31
2023-08-22
æ é¡é¡åé¶è¡è¡ä»½æéå ¬å¸
Account security check management method based on artificial intelligence
US12363150B2
( en )
*
2023-06-13
2025-07-15
Bank Of America Corporation
System and method for secured data analysis and synthetic identity detection in a distributed ledger network
CN117196800A
( en )
*
2023-06-20
2023-12-08
åå·ä»è¹è ¾é£ä¿¡æ¯ææ¯æéå ¬å¸
Digital management method for staff behaviors of banking outlets
CN116502277B
( en )
*
2023-06-30
2024-03-26
ä¸ç§ä¹çº¦å¥åº·ç§æ(æ·±å³)æéå ¬å¸
Medical data safety processing method, system and device based on blockchain
CN116980202B
( en )
*
2023-07-27
2023-12-26
广å·å°å ¨ä¿¡æ¯ææ¯æéå ¬å¸
Network security operation and maintenance monitoring method and system
US20250078081A1
( en )
*
2023-09-05
2025-03-06
Mastercard International Incorporated
Method and system for cryptocurrency fraud detection using unsupervised domain adaption
CN117238458B
( en )
*
2023-09-14
2024-04-05
广ä¸ç第äºäººæ°å»é¢ï¼å¹¿ä¸çå«çåºæ¥å»é¢ï¼
Critical care cross-mechanism collaboration platform system based on cloud computing
CN117094817B
( en )
*
2023-10-20
2024-02-13
å½ä»»è´¢äº§ä¿é©è¡ä»½æéå ¬å¸
Credit risk control intelligent prediction method and system
CN117151882B
( en )
*
2023-10-30
2024-05-07
å½ç½å¤©æ´¥å¸çµåå ¬å¸ç»æµææ¯ç 究é¢
A risk assessment method and system based on multi-variety electricity trading
CN117787792B
( en )
*
2023-12-27
2024-06-21
æ±èç§ä½³è½¯ä»¶å¼åæéå ¬å¸
A method and system based on medical device quality and safety risk supervision
US20250240220A1
( en )
*
2024-01-22
2025-07-24
Dropbox, Inc.
Dynamically selecting artificial intelligence models and hardware environments to execute tasks
CN118536609B
( en )
*
2024-06-05
2024-12-10
æ¦æ±ä¸å°å¤§æºæ §åå¸ç 究é¢æéå ¬å¸
A method and system for configuring an intelligent rule engine
CN118395505B
( en )
*
2024-06-25
2024-10-18
æ·±å³ç¹ç¹åä¸ä¾åºé¾æéå ¬å¸
A method and system for constructing agricultural product information database
CN118917945B
( en )
*
2024-07-17
2025-02-07
ä¸æµ·å¸å¤§æ°æ®ä¸å¿
Method for designing government chain supervision intelligent contract under AI drive and application
CN118898488B
( en )
*
2024-07-18
2025-06-13
æ·±å³çµå¨åæ°ä¿¡æ¯ææ¯æéå ¬å¸
Third-party payment security authentication method and system based on blockchain
EP4687089A1
( en )
*
2024-07-30
2026-02-04
Mastercard International Incorporated
Detecting batched transactions
CN119067767A
( en )
*
2024-07-31
2024-12-03
ä¸éåçåä¸ä¿çæéå ¬å¸
Factoring fund information security management method and system
CN118586046B
( en )
*
2024-08-05
2024-12-10
åçæå¾·è½æºé¡¾é®(å±±ä¸)æéå ¬å¸
Green energy transaction data monitoring system based on blockchain
WO2026091069A1
( en )
*
2024-11-01
2026-05-07
æ²³ååå·¥å»è¯èä¸ææ¯å¦é¢
Big data-based information collection system for software development
CN119444431B
( en )
*
2024-11-08
2025-12-05
æ·±å³é«æ¶¦ææºè½æéå ¬å¸
A method and system for auditing enterprise data assets based on big data
CN119648233B
( en )
*
2024-11-13
2025-10-24
æå·é«æ°åº(滨æ±)åºåé¾ä¸æ°æ®å®å ¨ç 究é¢
Blockchain address risk identification method, device, equipment and program product
CN119580155B
( en )
*
2024-11-19
2025-11-14
éåºé®çµå¤§å¦
A User Desktop Behavior Auditing Method and System Based on YOLO and LSTM
CN119648130B
( en )
*
2024-11-22
2025-07-08
西å®å¹¿ææ±æºè½æºç§ææéå ¬å¸
A software development full process supervision system
CN119671278A
( en )
*
2024-12-04
2025-03-21
åæ¹çµç½ç§å¦ç 究é¢æéè´£ä»»å ¬å¸
Method, device and computer equipment for assessing geological hazards along power grids
CN119921998B
( en )
*
2025-01-17
2025-11-18
ä¸ç§è¯æ°(æ·±å³)ç§ææéå ¬å¸
A method for security management and data integrity of switches based on blockchain technology
CN120162842B
( en )
*
2025-02-28
2025-12-26
æ±èå¤§å ¨ç®±åç§ææéå ¬å¸
Intelligent Integrated Design Method for Box-Type Substations Based on Cloud Data
CN120011791B
( en )
*
2025-04-18
2025-08-05
ç¦å·æºå 仪å¨è®¾å¤æéå ¬å¸
A safety assessment method and system for radioactive source equipment based on multi-factor analysis
CN121092208B
( en )
*
2025-07-18
2026-05-12
å½æ³°æµ·éè¯å¸è¡ä»½æéå ¬å¸
Method, device, processor and storage medium for realizing task calculation and adaptive parameter adaptation based on information creation big data platform
CN121563243B
( en )
*
2026-01-21
2026-05-08
èç°å¦é¢
Cross-border e-commerce risk monitoring system and methods based on data analysis
Citations (4)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20210042421A1
( en )
*
2019-08-05
2021-02-11
Yue Li
Blockchain security system
US20210110047A1
( en )
*
2019-10-15
2021-04-15
Anchain.ai Inc.
Continuous vulnerability management system for blockchain smart contract based digital asset using sandbox and artificial intelligence
US20210182859A1
( en )
*
2019-12-17
2021-06-17
Accenture Global Solutions Limited
System And Method For Modifying An Existing Anti-Money Laundering Rule By Reducing False Alerts
US20210390549A1
( en )
*
2020-06-15
2021-12-16
Capital One Services, Llc
Systems and methods for building blockchains for verifying assets for smart contracts
2020
2020-08-28
US
US17/006,431
patent/US11436615B2/en
active
Active
Patent Citations (4)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20210042421A1
( en )
*
2019-08-05
2021-02-11
Yue Li
Blockchain security system
US20210110047A1
( en )
*
2019-10-15
2021-04-15
Anchain.ai Inc.
Continuous vulnerability management system for blockchain smart contract based digital asset using sandbox and artificial intelligence
US20210182859A1
( en )
*
2019-12-17
2021-06-17
Accenture Global Solutions Limited
System And Method For Modifying An Existing Anti-Money Laundering Rule By Reducing False Alerts
US20210390549A1
( en )
*
2020-06-15
2021-12-16
Capital One Services, Llc
Systems and methods for building blockchains for verifying assets for smart contracts
Non-Patent Citations (1)
* Cited by examiner, â Cited by third party
Title
Machine Learning in/for Blockchain: Future and Challenges Chen.
Cited By (3)
* Cited by examiner, â Cited by third party
Publication number
Priority date
Publication date
Assignee
Title
US20210097383A1
( en )
*
2019-09-30
2021-04-01
International Business Machines Corporation
Combined Data Pre-Process And Architecture Search For Deep Learning Models
US11593642B2
( en )
*
2019-09-30
2023-02-28
International Business Machines Corporation
Combined data pre-process and architecture search for deep learning models
US20250045361A1
( en )
*
2023-07-31
2025-02-06
Fireblocks Ltd.
Protection of digital assets within the blockchain by approval revocations
Also Published As
Publication number
Publication date
US20220067752A1
( en )
2022-03-03
Similar Documents
Publication
Publication Date
Title
US11907955B2
( en )
2024-02-20
System and method for blockchain automatic tracing of money flow using artificial intelligence
US20220067752A1
( en )
2022-03-03
System and method for blockchain transaction risk management using machine learning
US20230342846A1
( en )
2023-10-26
Micro-loan system
US11928049B2
( en )
2024-03-12
Blockchain system for source code testing and script generation with artificial intelligence
US12411959B2
( en )
2025-09-09
Systems, methods, and apparatuses for detecting and implementing vulnerability fixes in an electronic environment
US20220318270A1
( en )
2022-10-06
Artificial intelligence (ai)-based blockchain management
US12388867B2
( en )
2025-08-12
Systems and methods for identifying electronic accounts based on near real time machine learning in an electronic network
US12423444B2
( en )
2025-09-23
System for software code cyber security based on machine learning vulnerability detection and generation and implementation of vulnerability test
US12499222B2
( en )
2025-12-16
Systems and methods for machine interpretation of security data via dynamic constraint specification matrix
US12282865B2
( en )
2025-04-22
Artificial intelligence (AI)-based blockchain management
US12495060B2
( en )
2025-12-09
System and method for assessing entity security using aggregated data sources
US20250159072A1
( en )
2025-05-15
Systems and methods for dynamically determining procedures for electronic communications using cellular automaton processing
US20240370554A1
( en )
2024-11-07
Systems, methods, and apparatuses for threat indexing and implementing ai to detect malfeasant user activity in an electronic environment
US20240338592A1
( en )
2024-10-10
Systems, methods, and apparatuses for generating a digital twin of a resource using partial sensor data and artificial intelligence
Malashin et al.
2024
Anomaly Detection in Financial Data Using GA-Optimized MLP Models
US12561455B2
( en )
2026-02-24
System and method of sequential cypher encryption based on contextual machine learning powered performance testing engine for production environment
US12632541B2
( en )
2026-05-19
Systems and methods to improve network and data security by automatically preventing rogue database queries
US20260142887A1
( en )
2026-05-21
Systems and methods for network event monitoring and network event remediation using artificial intelligence
US12353355B2
( en )
2025-07-08
Systems, methods, and apparatuses for tracking cross border data transmissions and generating a data passport entry in an electronic network
US12487801B2
( en )
2025-12-02
Systems and methods for determining software dependencies and software change impacts on computer processing
US20250036506A1
( en )
2025-01-30
System and method of using advanced computation models for data analysis and automated decision-making for root cause analysis within software
US20260052167A1
( en )
2026-02-19
Systems and methods for identifying data security threats and dynamically generating vulnerability solutions
US20250391410A1
( en )
2025-12-25
Systems and methods for enhancing security associated with networked devices via artificial intelligence enhanced processing
US12530255B2
( en )
2026-01-20
System, methods, and apparatuses for identifying and resolving anomalous data within a distributed network
US20260057104A1
( en )
2026-02-26
Systems and method for auto-redacting data in a shared-graphical-user-interface environment
Legal Events
Date
Code
Title
Description
2020-08-28
FEPP
Fee payment procedure
Free format text : ENTITY STATUS SET TO UNDISCOUNTED (ORIGINAL EVENT CODE: BIG.); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY
2020-09-09
FEPP
Fee payment procedure
Free format text : ENTITY STATUS SET TO SMALL (ORIGINAL EVENT CODE: SMAL); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY
2020-09-22
AS
Assignment
Owner name : ANCHAIN.AI, CALIFORNIA
Free format text : ASSIGNMENT OF ASSIGNORS INTEREST;ASSIGNORS:FANG, CHUNSHENG VICTOR;QUAN, WEI;WU, BEN BINYI;REEL/FRAME:053844/0218
Effective date : 20200902
2022-06-22
STPP
Information on status: patent application and granting procedure in general
Free format text : NOTICE OF ALLOWANCE MAILED -- APPLICATION RECEIVED IN OFFICE OF PUBLICATIONS
2022-07-27
STPP
Information on status: patent application and granting procedure in general
Free format text : PUBLICATIONS -- ISSUE FEE PAYMENT VERIFIED
2022-08-17
STCF
Information on status: patent grant
Free format text : PATENTED CASE
2026-04-27
FEPP
Fee payment procedure
Free format text : MAINTENANCE FEE REMINDER MAILED (ORIGINAL EVENT CODE: REM.); ENTITY STATUS OF PATENT OWNER: SMALL ENTITY
if (window.help && window.help.service) { helpApi = window.help.service.Lazy.create(0, {apiKey: 'AIzaSyDTEI_0tLX4varJ7bwK8aT-eOI5qr3BmyI', locale: 'en-US'}); window.requestedSurveys = new Set(); window.requestSurvey = function(triggerId) { if (window.requestedSurveys.has(triggerId)) { return; } window.requestedSurveys.add(triggerId); helpApi.requestSurvey({ triggerId: triggerId, enableTestingMode: false, callback: (requestSurveyCallbackParam) => { if (!requestSurveyCallbackParam.surveyData) { return; } helpApi.presentSurvey({ productData: { productVersion: window.version, customData: { "experiments": "", }, }, surveyData: requestSurvey