EP4268170A1 - Procédé et système de détection automatisée de risque de fraude dans un système surveillé - Google Patents
Procédé et système de détection automatisée de risque de fraude dans un système surveilléInfo
- Publication number
- EP4268170A1 EP4268170A1 EP21847699.2A EP21847699A EP4268170A1 EP 4268170 A1 EP4268170 A1 EP 4268170A1 EP 21847699 A EP21847699 A EP 21847699A EP 4268170 A1 EP4268170 A1 EP 4268170A1
- Authority
- EP
- European Patent Office
- Prior art keywords
- events
- operator
- monitored system
- risk
- subset
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1433—Vulnerability analysis
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/09—Supervised learning
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q20/00—Payment architectures, schemes or protocols
- G06Q20/38—Payment protocols; Details thereof
- G06Q20/40—Authorisation, e.g. identification of payer or payee, verification of customer or shop credentials; Review and approval of payers, e.g. check credit lines or negative lists
- G06Q20/401—Transaction verification
- G06Q20/4016—Transaction verification involving fraud or risk level assessment in transaction processing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04L—TRANSMISSION OF DIGITAL INFORMATION, e.g. TELEGRAPHIC COMMUNICATION
- H04L63/00—Network architectures or network communication protocols for network security
- H04L63/14—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic
- H04L63/1408—Network architectures or network communication protocols for network security for detecting or protecting against malicious traffic by monitoring network traffic
- H04L63/1425—Traffic logging, e.g. anomaly detection
Definitions
- TITLE Method and system for automated fraud risk detection in a monitored system
- the present invention relates to a method and a system for automated detection of the risk of fraud in a monitored system, from signals characterizing said monitored system.
- the invention is in the field of automated fraud prediction and detection, and finds various applications in systems in which fraud can occur, for example financial systems, compliance management and control systems and performance, banking systems, insurance systems, industrial systems.
- heterogeneous data flows structured (for example execution logs, formatted files) or unstructured (for example images, audio recordings) which characterize events (e.g. operations, actions, transactions, executions, provision of data).
- Events are assigned to operators, they are performed or generated/controlled by operators of the monitored system.
- the data characterizing the monitored system are recorded temporally.
- a constant and recurring problem, whether it concerns financial systems, banking systems, industrial systems, and information systems, is to automatically detect, at an early stage, any irregularity, anomaly or inconsistency that may be due to fraudulent use of the system or errors in the use of the system or compromises or even intrusions that may affect it.
- Fraud is understood here to mean an operation carried out by intentionally violating one or more rules and defense barriers defining the legitimate operation of a system or of one or more of its attributes, an attribute designating a founding characteristic of the system (or "feature " in English).
- the operator who ordered such an operation/transaction is presumed to be a fraudster and is suspected of being the source of the fraud.
- Various fraud detection methods are known in the state of the art, in particular financial fraud detection, which are based on artificial intelligence and machine learning techniques, based on supervised learning on databases relating to financial transactions.
- the known methods generally provide a rating component, characterizing the risk or the probability of fraud.
- risk probability of error in decision-making, expressed in terms of false positive rate and false negative rate.
- the false positive rate is the percentage of cases where the method applied detects fraud, while it is a proven legitimate operator.
- the false negative rate is the percentage of cases where the method applied detects a legitimate operator, whereas it is a case of a proven fraudster.
- Such probabilities of error in decisions are calculated by applying the detection method to data relating to cases of fraud or proven legitimate events/operations.
- One of the objectives of the invention is to remedy this drawback, by proposing a more reliable method.
- the invention proposes, according to one aspect, a method for automated detection of the risk of fraud in a monitored system, from data flows generated by said monitored system and characterizing events, carried out or generated by operators in said monitored system.
- This method comprises steps, implemented by a calculation processor, of:
- the method of automated fraud risk detection in a monitored system independently implements a first legitimacy calculation process, concomitant with a second illegitimacy calculation process.
- the first process and the second process are robust calculation methods deployed substantially in parallel.
- the method for automated fraud risk detection in a monitored system according to the invention may also have one or more of the characteristics below, taken independently or in all technically conceivable combinations.
- the pre-processing comprises a determination of events from said recorded data, a calculation of a signature of each event of at least a part of said events, said signature being representative of a risk of malfunction of the monitored system following said event .
- the pre-processing also includes a determination of a subset of critical events whose signature is greater than a predetermined risk threshold.
- the method comprises successive iterations of the application steps of said first and second processes, the first process being applied to distinct parts of said subset of critical events until validation of a first convergence criterion, the second process being applied to distinct parts of said subset of critical events until validation of a second convergence criterion.
- step D) implements a third convergence criterion, and in the case of absence of convergence according to said third convergence criterion, the method comprises an iteration of steps A) to D) on another set of data recorded over a second time period, situated in the past with respect to said first time period.
- the calculation of the signature of each event is carried out according to past events, by a logistic regression method.
- the parameters of said first parameterized process are obtained by supervised learning.
- the parameters of said second parameterized process are obtained by unsupervised learning.
- the method includes a prior step of classifying said events into a plurality of classes, and the application of steps A) to D) for at least one class of events.
- the invention relates to an automated fraud risk detection system in a monitored system, based on data flows generated by said monitored system and characterizing events, carried out or generated by operators in said monitored system.
- This system comprises at least one calculation processor configured to implement:
- the invention relates to one or more computer program(s) comprising software instructions which, when implemented by a programmable electronic device, implement an automated fraud risk detection method in a monitored system as briefly described above.
- Figure 1 is an example of a fraud detection system according to one embodiment
- FIG 2 is a block diagram of the main steps of a pre-processing phase of an automated fraud risk detection method according to one embodiment
- FIG 3 figure 3 is a synoptic of the main processing steps by two distinct processes for the automated detection of fraud risk
- Figure 4 schematically illustrates a neural network architecture.
- the invention finds applications for the automated detection of the risk of fraud in any system to be monitored, and in particular in the case where the system to be monitored is a compliance and performance control management system in the industrial field.
- Figure 1 schematically illustrates an automated fraud detection system according to one embodiment in a monitored system 2.
- the monitored system generates data streams 4i to 4k , which are recorded in one or more electronic memory units 6, for example in the form of a database or any other data storage structure.
- unit 6 is formed by a plurality of interconnected memories.
- Each data stream 4i corresponds to a time period T, the time periods Ti to T k succeeding each other in time.
- each time period has the same predetermined duration, for example between one month and one year, for example equal to six months.
- the time periods follow one another, for example in a given chronological order, for example the period T i+i being prior to the period T, and so on.
- the data streams 4i are stored as they are generated by the monitored system 2.
- Each data stream includes structured or unstructured data and characterizes events performed or generated by the monitored system 2.
- the data When recorded by the monitored system, the data, structured or unstructured, is preferably categorized and labeled according to the specifics of the monitored system.
- the structured data includes for example execution logs, formatted files, alarm feedback by monitoring mechanisms, and the unstructured data includes for example images/videos, audio recordings.
- the recorded data characterizes events (e.g. operations, actions, transactions, executions, provision of data).
- Events are assigned to operators, they are performed or generated/controlled by operators of the monitored system.
- Operator information is also recorded, in association with the data streams.
- each operator denoted Op
- each operator is identified by a chosen ld_Op identifier, and optionally, each operator also has an associated Profil_Op profile, which is also stored by the monitored system 2.
- An operator profile includes characteristic attributes of the operator, for example age, sex, place of birth, in professional activity or not...
- the data streams 4i stored over associated time periods T are accessible to an automated fraud detection system 10 according to the invention.
- system 10 is adapted to also receive contextual information 9 from an external database 8, the contextual information 9 comprising for example statistical knowledge on profiles of suspected fraudsters.
- the system 10 implements the stored data streams 4, and optionally the contextual information 9 to determine events relating to feared situations or threats, hereinafter called critical events.
- a critical event means any operation, action, transaction, execution or provision of data which, according to an analysis carried out, may result from a potential malfunction of the monitored system, i.e. from an anomaly or inconsistency, or from a risk of anomaly or inconsistency in the event of detection of precursor events, relating to the operating rules of the monitored system.
- a potential malfunction of the monitored system i.e. from an anomaly or inconsistency, or from a risk of anomaly or inconsistency in the event of detection of precursor events, relating to the operating rules of the monitored system.
- Each event, and a fortiori each critical event is attributed to an operator.
- the invention makes it possible to detect whether the or each operator associated with critical events is a legitimate operator or a fraudulent operator (or presumed fraudster).
- the critical events detected corresponding to errors or malfunctions of the monitored system the detection carried out then constitutes an opportunity to improve the monitored system and/or the way of managing or carrying out operations, transactions, etc. in this system.
- the invention makes it possible to take measures to prevent the fraudulent operator from continuing to act, to take corrective measures, to avoid the occurrence of a major risk or threat .
- the automated fraud detection system 10 comprises one or more calculation units 11 (e.g. one or more calculation processors) for example distributed over several computers or on a single computer, or more generally a programmable electronic device, configured to put implemented:
- calculation units 11 e.g. one or more calculation processors
- a module 16 for applying, substantially in parallel, two parameterized processes for estimating the risk of fraud which are respectively: a first parameterized process for estimating the risk of fraud, the parameters of the first parameterized process being obtained by learning ( module 18) on a first database BDD1 of data representative of events carried out or generated by legitimate operators; a second parameterized process for estimating the risk of fraud, the parameters of the second process being obtained by learning (module 20) on a second database representative of events carried out or generated by fraudulent operators; a module 17 for comparing the results of said first process and second process to determine whether said operator is a legitimate operator or a fraudulent operator.
- the first process and the second process are applied to 4i data streams associated with periods Ti, until validation of a convergence criterion as explained in detail below.
- the calculation unit 11 is adapted to communicate with a memory unit 22, making it possible to store useful data and parameters for the implementation of the modules 12, 14, 16 and 17.
- the module 12, if any, as well as the modules 14, 16 and 17, are implemented in the form of software code, and form a computer program, comprising software instructions which, when they are implemented by a programmable electronic device, implement an automated fraud risk detection method in a monitored system.
- the modules 12, 14, 16, 17 are each made in the form of a programmable logic component, such as an FPGA (from the English Field Programmable Gate Array), or a GPGPII (from the English General-purpose processing on graphics processing), or else in the form of a dedicated integrated circuit, such as an ASIC (Application Specific Integrated Circuit).
- a programmable logic component such as an FPGA (from the English Field Programmable Gate Array), or a GPGPII (from the English General-purpose processing on graphics processing), or else in the form of a dedicated integrated circuit, such as an ASIC (Application Specific Integrated Circuit).
- the computer program for automated fraud risk detection in a monitored system is also capable of being recorded on a computer-readable medium, not shown.
- the computer-readable medium is, for example, a medium capable of storing electronic instructions and of being coupled to a bus of a computer system.
- the readable medium is an optical disc, a magneto-optical disc, a ROM memory, a RAM memory, any type of non-volatile memory (for example EPROM, EEPROM, FLASH, NVRAM), a magnetic card or an optical card.
- each of the learning modules 18, 20 is produced in the form of software code, and forms a computer program implementing a learning phase for each of the first and second risk estimation processes of fraud.
- Figure 2 is a block diagram of the main steps of a pre-processing phase of a fraud risk detection method according to one embodiment.
- This preprocessing phase comprises a step 30 for obtaining one or more 4i to 4k data streams generated by the monitored system.
- the data stream 4i corresponding to a chosen time period T is obtained in step 30.
- Step 30 is followed by a step 32 of data analysis to determine events, e.g. operations, actions, transactions, executions, provision of data to the monitored system, incident alerts, and an optional step 34 of classification of events, described below.
- events e.g. operations, actions, transactions, executions, provision of data to the monitored system, incident alerts, and an optional step 34 of classification of events, described below.
- Step 32 or step 34 is followed by a step 36 of calculating a signature representative of a risk of malfunction of the monitored system for each event, then by a step 38 of determining and extracting a subset of critical events.
- the steps 36 of calculating a signature representative of a risk of malfunctioning of the monitored system and 38 of determining a subset of critical events are carried out by implementing a method of capture and characterization of weak signals, as described in patent FR 3 009 615 B1.
- a risk signature is calculated, for an event under consideration, by using a dependency between the event under consideration and events from the past on which the event under consideration depends, for example by using a tree representation of the dependencies between events.
- a risk signature is calculated by a logistic regression algorithm.
- a risk signature S is calculated by the formula:
- ⁇ p(z) is an activation function with formula: ⁇ p(z) w is a vector of weights.
- a risk signature of an event is calculated from risk signatures of N past events, each having a signature 0 , j e ⁇ 1, ... , N ⁇ .
- the value 8 brings a bias.
- this signature is compared with a predetermined risk threshold, and the subset of critical events is determined as being the subset of events whose signature is greater than said predetermined risk threshold.
- step 38 a subset of critical events is determined, each critical event being associated with an operator.
- the following steps of the fraud risk detection method are carried out for one or each of the operators associated with one or more critical event(s) of the subset of critical events.
- the pre-processing steps 32 to 38 are performed on events associated with one and the same operator.
- the application of the pre-processing described above makes it possible to determine one or more subsets of critical events, each subset of critical events being associated with an operator identified in the monitored system.
- the method further comprises an optional step 34 of classification, for example implemented before or after the step 32 of analyzing the data to determine events, eg operations, actions, transactions, executions, provision of data to the monitored system, incident alerts, classification of events into several classes of events.
- the principal component analysis (PCA) classification method can be used, for quantitative data or for qualitative data, using an association table.
- PCA principal component analysis
- MCA multiple correspondence factor analysis
- FIG. 3 is a block diagram of the main stages of processing by two distinct parameterized processes for estimating the risk of fraud, this processing being carried out substantially in parallel, for the same operator, and, where applicable, for the same class of 'events.
- the method comprises an initialization step 50, comprising a supply of data representative of critical events, associated with, and more precisely attributed to the same operator Op_M with identifier I d_M , these data corresponding to at least part of the sub- set of critical events determined at the end of the preliminary preprocessing phase.
- Data representative of critical events include, for example, measurements, evaluations, deviations, thresholds, etc.
- Data representative of events is provided over one or more time periods.
- An index k indicates a time period T k for which the processing is applied, the time periods being processed in a chosen order, for example an inverse chronological order, ie starting with a time period close to a present time instant and progressing l reverse chronological order, ie moving away from the present moment towards the past.
- the index k is initialized to 1, Ti designating a last time period before the present time (i.e. the time of application of the processing).
- Ti designating a last time period before the present time (i.e. the time of application of the processing).
- the initialization step 50 is followed by steps 52 and 54, substantially concomitant, which are respectively a step 52 of application of a first parameterized process for estimating the risk of fraud, called “process 1” below, and an step 54 of applying a second parameterized fraud risk estimation process, called “process 2” below.
- the “process 1” and the “process 2” are applied to the same part of the data representative of critical events associated with the operator identified by ld_M, of the time period T k .
- the “process 1” implements a machine learning algorithm, for example an artificial neural network comprising a plurality of layers, respectively an input layer, at the minus one hidden layer, one output layer.
- learning 46 of the parameters of “process 1” is learning of the supervised type.
- the “process 2” implements a machine learning algorithm, for example an artificial neural network comprising one or more layers.
- learning 48 of the parameters of “process 2” is learning of the unsupervised type.
- the two processes apply different neural networks, parameterized by training on different learning bases, respectively BDD1 and BDD2, the first base BDD1 containing data representative of events carried out or generated by legitimate operators; and the second database BDD2 containing data representative of events carried out or generated by fraudulent operators.
- Data Xi to XN supplied as input to such a neural network are represented in the form of a vector or a matrix. These data are for example representative of previously determined critical events, associated with the operator identified by ld_M.
- Figure 4 schematically illustrates a deep neural network with two hidden layers.
- the neural network 100 of Figure 4 has an input layer 102, two hidden layers 104, 106 and an output layer 108.
- Each layer Ci has a number Ni of neurons 1 10.
- the numbers Ni vary from one layer to another.
- each neuron of each layer is connected to each neuron of the preceding layer and to each neuron of the following layer, such a network of neurons being a fully connected network.
- a neuron 110 calculates a weighted sum of its inputs, with an optional addition of bias b, then applies a function called the activation function c to the calculated weighted sum.
- each neuron performs a nonlinear combination of inputs.
- the activation function c is for example the sigmoid function:
- the parameters defining a neural network, for one or more chosen activation function(s) are the weights Wi and the bias values b for each neuron of each layer. These parameters are calculated in a learning phase, on supervised or unsupervised learning data, to achieve an objective, that is to say to obtain output values from the output layer 108 which satisfy a predetermined criterion, for example a cost function denoted J(W,B), where W denotes the set of weights defining the neural network and B the set of biases.
- the cost function J(W,B) is defined so as to minimize the probability of error in the desired result.
- the neural network applied by “process 1” provides at the output of the output layer a first legitimacy score Score 1(M,X,k) characterizing the operator identified by ld_M and a first probability of occurrence Prob1(M,X,k) associated with the critical events tested, over the time period T k .
- the first legitimacy score is for example a real value between 0 and 1, corresponding respectively to 0% legitimacy up to 100% legitimacy.
- the value of the first legitimacy score indicates an estimate of distance between the identifier operator ld_M and a legitimate reference individual or an estimate of distance between the identifier operator ld_M and a fraudulent individual of reference.
- a legitimate reference individual is either a statistical individual representing the statistical distribution of the population that he represents, or else one or more individuals, whose legitimacy has been concretely proven, either after a suspicion, ultimately not proven, or else depending an optimal temporal history justifying his qualification as a legitimate reference individual.
- the cost function to be minimized for the neural network applied by “process 1” is for example a function associated with the detection error of a fraudulent operator.
- process 1 implements a convolutional neural network CNN (for “Convolutional Neural Network”), comprising for example an input layer, an output layer and two hidden layers.
- CNN for “Convolutional Neural Network”
- the neural network applied by “process 2” provides at the output of the output layer a second legitimacy score Score2(M,X,k) characterizing the operator ld_M and a second probability of occurrence Prob2(M,X,k) associated with the critical events tested, over the time period T k .
- process 2 implements a convolutional neural network CNN (for “Convolutional Neural Network”).
- CNN for “Convolutional Neural Network”.
- simpler architectures are preferred, for example a single-layer neural network, in order to gain in interpretability and explainability.
- a first criterion of convergence of the results of “process 1” is verified at step 56, and in the case where this first convergence criterion is not verified, “process 1” is iterated on the data representative of other critical events of the subset of critical events of the time period T k .
- a second criterion of convergence of the results of “process 2” is verified at step 58, and in the case where this second convergence criterion is not verified, “process 2” is iterated on the data representative of other critical events of the subset of critical events of the time period T k . Then, when the first convergence criterion and the second convergence criterion are verified, the respective results of “process 1” and of “process 2” are compared during a comparison step 60 implementing a third convergence criterion. For example, a calculated absolute difference between the first legitimacy score and the second legitimacy score is compared to a predetermined legitimacy score threshold.
- the index k indicating the time period processed is incremented (step 62), and the steps 52, 56 and 54, 58 are iterated over a new time period, preferably a period time of the past.
- step 60 is followed by a step 64 for verifying the final result. Verification of the third convergence criterion indicates that the first legitimacy score and the second legitimacy score are close to the same final legitimacy score. It is checked in step 64 whether the final legitimacy score corresponds to a reference fraudster individual, i.e. indicates that the operator identified by ld_M is a presumed fraudster operator.
- step 64 is followed by a step 66 of enrichment of the first learning database BDD1.
- step 66 is followed by the implementation of step 46 for learning the parameters of “process 1”.
- step 64 is followed by a step 68 for raising the alert, for example by sending a notification, or any other means, to authorities managing the system monitored.
- the alert is processed by applying a check 70 by a human supervisor, who validates or invalidates the automated fraud detection.
- this step is followed by a step 72 of enriching the second database BDD2, and optionally, an implementation of the step 48 of learning the parameters of "process 2".
- the solution will delimit the scope of investigation associated with the process in question, by identifying on the one hand, all the human actors operating and/or interacting in the perimeter concerned, by listing and evaluating on the other hand, all the organizational factors associated processes, and finally, the operational and industrial history of the industrial installation housing the process.
- each individual present or intervening in the process, or even upstream or downstream of the process is assessed using this solution.
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Security & Cryptography (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Business, Economics & Management (AREA)
- Computer Hardware Design (AREA)
- Computer Networks & Wireless Communication (AREA)
- Signal Processing (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Debugging And Monitoring (AREA)
- User Interface Of Digital Computer (AREA)
- Financial Or Insurance-Related Operations Such As Payment And Settlement (AREA)
Abstract
Description
Claims
Applications Claiming Priority (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| FR2014184A FR3118518A1 (fr) | 2020-12-28 | 2020-12-28 | Procédé et système de détection automatisée de risque de fraude dans un système surveillé |
| PCT/EP2021/087710 WO2022144347A1 (fr) | 2020-12-28 | 2021-12-28 | Procédé et système de détection automatisée de risque de fraude dans un système surveillé |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| EP4268170A1 true EP4268170A1 (fr) | 2023-11-01 |
Family
ID=75746765
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| EP21847699.2A Pending EP4268170A1 (fr) | 2020-12-28 | 2021-12-28 | Procédé et système de détection automatisée de risque de fraude dans un système surveillé |
Country Status (7)
| Country | Link |
|---|---|
| US (1) | US12556569B2 (fr) |
| EP (1) | EP4268170A1 (fr) |
| JP (1) | JP2024501035A (fr) |
| CN (1) | CN116710944A (fr) |
| CA (1) | CA3203085A1 (fr) |
| FR (1) | FR3118518A1 (fr) |
| WO (1) | WO2022144347A1 (fr) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150026027A1 (en) * | 2009-06-12 | 2015-01-22 | Guardian Analytics, Inc. | Fraud detection and analysis |
Family Cites Families (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP4494570B2 (ja) * | 2000-01-20 | 2010-06-30 | 株式会社小森コーポレーション | 振りローラの振り装置 |
| EP2288987A4 (fr) * | 2008-06-12 | 2015-04-01 | Guardian Analytics Inc | Modélisation d'utilisateur pour la détection et l'analyse de fraude |
| CA2819643C (fr) * | 2010-12-01 | 2019-05-21 | Headwater Partners I Llc | Dispositif pour utilisateur final qui garantit l'association d'une application a une politique de service au moyen de la verification d'un certificat d'application |
| FR3009615B1 (fr) | 2013-08-06 | 2016-10-14 | Commissariat Energie Atomique | Procede et systeme de captage, discrimination et caracterisation de signaux faibles au moyen de leurs signatures respectives |
| US10009358B1 (en) * | 2014-02-11 | 2018-06-26 | DataVisor Inc. | Graph based framework for detecting malicious or compromised accounts |
| US9787723B2 (en) * | 2014-07-18 | 2017-10-10 | Ping Identify Corporation | Devices and methods for threat-based authentication for access to computing resources |
| US20160196615A1 (en) * | 2015-01-06 | 2016-07-07 | Wells Fargo Bank, N.A. | Cross-channel fraud detection |
| US20190311367A1 (en) * | 2015-06-20 | 2019-10-10 | Quantiply Corporation | System and method for using a data genome to identify suspicious financial transactions |
| JP7467033B2 (ja) * | 2018-07-04 | 2024-04-15 | キヤノンメディカルシステムズ株式会社 | 診療情報処理装置 |
| CN111047428B (zh) * | 2019-12-05 | 2023-08-08 | 深圳索信达数据技术有限公司 | 基于少量欺诈样本的银行高风险欺诈客户识别方法 |
| CN111552680A (zh) * | 2020-04-27 | 2020-08-18 | 深圳壹账通智能科技有限公司 | 业务欺诈识别数据库的构建方法、装置和计算机设备 |
| WO2023128865A2 (fr) * | 2021-12-29 | 2023-07-06 | Gp Network Asia Pte. Ltd. | Serveur de communication, procédé, dispositif utilisateur et système |
| US20240152926A1 (en) * | 2022-11-03 | 2024-05-09 | Chime Financial, Inc. | Preventing digital fraud utilizing a fraud risk tiering system for initial and ongoing assessment of risk |
| US20250029101A1 (en) * | 2023-07-20 | 2025-01-23 | Mastercard International Incorporated | Tokenized device identifier generator and tracking system |
| US20250148469A1 (en) * | 2023-11-03 | 2025-05-08 | Fmr Llc | Deposit fraud detection |
-
2020
- 2020-12-28 FR FR2014184A patent/FR3118518A1/fr active Pending
-
2021
- 2021-12-28 CA CA3203085A patent/CA3203085A1/fr active Pending
- 2021-12-28 CN CN202180091378.5A patent/CN116710944A/zh active Pending
- 2021-12-28 EP EP21847699.2A patent/EP4268170A1/fr active Pending
- 2021-12-28 WO PCT/EP2021/087710 patent/WO2022144347A1/fr not_active Ceased
- 2021-12-28 US US18/259,335 patent/US12556569B2/en active Active
- 2021-12-28 JP JP2023539348A patent/JP2024501035A/ja active Pending
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150026027A1 (en) * | 2009-06-12 | 2015-01-22 | Guardian Analytics, Inc. | Fraud detection and analysis |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116710944A (zh) | 2023-09-05 |
| CA3203085A1 (fr) | 2022-07-07 |
| US20240056471A1 (en) | 2024-02-15 |
| WO2022144347A1 (fr) | 2022-07-07 |
| FR3118518A1 (fr) | 2022-07-01 |
| US12556569B2 (en) | 2026-02-17 |
| JP2024501035A (ja) | 2024-01-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Rahul | Strengthening Fraud Prevention with AI in P&C Insurance: Enhancing Cyber Resilience | |
| US12066918B2 (en) | System to track and measure machine learning model efficacy | |
| US11321164B2 (en) | Anomaly recognition in information technology environments | |
| US20210406112A1 (en) | Anomaly classification in information technology environments | |
| US12411959B2 (en) | Systems, methods, and apparatuses for detecting and implementing vulnerability fixes in an electronic environment | |
| US11573882B2 (en) | Systems and methods for optimizing a machine learning-informed automated decisioning workflow in a machine learning task-oriented digital threat mitigation platform | |
| US20250124124A1 (en) | Utilizing machine-learning models to determine take-over scores and intelligently secure digital account features | |
| US12326925B2 (en) | Systems and methods for password spraying identification and prevention using hash signature segmentation and behavior clustering analysis | |
| US12423444B2 (en) | System for software code cyber security based on machine learning vulnerability detection and generation and implementation of vulnerability test | |
| FR3065558A1 (fr) | Systeme et procede pour gerer la detection de fraudes dans un systeme de transactions financieres | |
| CN119180713A (zh) | 金融风险预测方法、预测装置及计算机设备 | |
| US12373277B2 (en) | System and method for determining causes of network anomalies across a distributed network | |
| US20250016158A1 (en) | System and method for secure network access management using a dynamic constraint specification matrix | |
| Moriah et al. | Measuring and mitigating biases in motor insurance pricing | |
| EP4268170A1 (fr) | Procédé et système de détection automatisée de risque de fraude dans un système surveillé | |
| US20250029106A1 (en) | Automated rule creation for fraud detection using a machine learning-driven computing framework | |
| US20170154279A1 (en) | Characterizing subpopulations by exposure response | |
| US20230229492A1 (en) | Automated context based data subset processing prioritization | |
| WO2022135972A1 (fr) | Procede et dispositif de diagnostic d'anomalies | |
| EP3736743B1 (fr) | Procédé de conception d'un générateur de signature d'une entité réalisant des actions dans une architecture informatique, procédé de détection de comportement anormal, programme d'ordinateur et système correspondants | |
| US12513165B2 (en) | System and method to detect data leaks via a distributed data orchestration apparatus and transmitted fabricated data responses | |
| US20260080304A1 (en) | Data quality model for drift-resistant inferences | |
| US20260010641A1 (en) | Systems and methods for automatically generating and updating data security parameters using generative artificial intelligence | |
| CN121260221A (zh) | 一种风险芯片修复方法及装置 | |
| EP4488893A1 (fr) | Génération de données d apprentissage pour l apprentissage machine d'un modèle de prédiction d'un risque de survenue d'un évènement rare |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: UNKNOWN |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: THE INTERNATIONAL PUBLICATION HAS BEEN MADE |
|
| PUAI | Public reference made under article 153(3) epc to a published international application that has entered the european phase |
Free format text: ORIGINAL CODE: 0009012 |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: REQUEST FOR EXAMINATION WAS MADE |
|
| 17P | Request for examination filed |
Effective date: 20230627 |
|
| AK | Designated contracting states |
Kind code of ref document: A1 Designated state(s): AL AT BE BG CH CY CZ DE DK EE ES FI FR GB GR HR HU IE IS IT LI LT LU LV MC MK MT NL NO PL PT RO RS SE SI SK SM TR |
|
| DAV | Request for validation of the european patent (deleted) | ||
| DAX | Request for extension of the european patent (deleted) | ||
| RAP3 | Party data changed (applicant data changed or rights of an application transferred) |
Owner name: COMMISSARIAT A L'ENERGIE ATOMIQUE ET AUX ENERGIESALTERNATIVES |
|
| STAA | Information on the status of an ep patent application or granted ep patent |
Free format text: STATUS: EXAMINATION IS IN PROGRESS |
|
| 17Q | First examination report despatched |
Effective date: 20260317 |