Managed service for detection of anomalous transactions
A managed service for the detection of fraud and/or anomalous transactions is disclosed. In one implementation, the managed service receives data from customers over a wide area network, the data for each customer relating to information that the customer would like monitored for anomalous transactions. The service may analyze data from the customers based on Benford's law to detect potential anomalous transactions in the data from the customers. The managed service may further report occurrences of detected potential anomalous transactions to the customer.
1. A method comprising:
determining, by a network device, a type of data to be analyzed;
determining, by the network device and based on the type of the data, a threshold value associated with a particular probability distribution associated with the data,
the threshold value corresponding to an acceptable frequency of a particular digit occurring in the data;
analyzing, by the network device and based on the threshold value, the particular probability distribution to detect a potential fraudulent transaction; and
reporting, by the network device, an occurrence of the potential fraudulent transaction.
2. The method of claim 1 , further comprising:
applying Benford's law to the data to determine the particular probability distribution.
3. The method of claim 1 , where the particular digit comprises a leading digit.
4. The method of claim 1 , where analyzing the particular probability distribution includes:
generating, based on the data, a model or a set of rules to detect an anomalous transaction; and
analyzing the particular probability distribution based on the threshold value and based on the model or the set of rules.
5. The method of claim 1 , where determining the threshold value includes:
determining a quantity of samples included in the data; and
determining the threshold value based on the type of the data and the quantity of samples.
6. The method of claim 1 , where determining the threshold value includes:
receiving, from a customer device associated with the data being analyzed, an initial parameter associated with a historical analysis of the type of the data; and
determining the threshold value based on the initial parameter.
7. The method of claim 1 , further comprising:
receiving, from a customer associated with the data, a parameter relating to changing the threshold value; and
changing the threshold value based on the parameter.
8. A network device comprising:
a processor to:
determine a type of data to be analyzed;
determine, based on the type of the data, a threshold value associated with a particular probability distribution associated with the data,
the threshold value corresponding to an acceptable frequency of a particular digit occurring in the data;
analyze, based on the threshold value, the particular probability distribution to detect a potential fraudulent transaction; and
report an occurrence of the potential fraudulent transaction.
9. The network device of claim 8 , where the processor is further to:
apply Benford's law to the data to determine the particular probability distribution.
10. The network device of claim 8 , where the particular digit comprises a leading digit.
11. The network device of claim 8 , where, when analyzing the particular probability distribution, the processor is to:
generate, based on the data, a model or a set of rules to detect an anomalous transaction in the data, and
analyze the particular probability distribution based on the threshold value and based on the model or the set of rules.
12. The network device of claim 8 , where, when determining the threshold value, the processor is to:
determine a quantity of samples included in the data, and
determine the threshold value further based on the type of the data and the quantity of samples.
13. The network device of claim 8 , where, when determining the threshold value, the processor is to:
receive, from a customer associated with the data, an initial set of parameters associated with a historical analysis of the type of the data, and
determine the threshold value based on the initial set of parameters.
14. The network device of claim 8 , where the data is accounts payable data or corporate expenses data.
15. A non-transitory computer-readable medium storing instructions, the instructions comprising:
one or more instructions that, when executed by at least one processor, cause the at least one processor to:
identify a particular digit of data received from a network device;
determine a threshold value based on a type associated with the data, the threshold value corresponding to
an acceptable frequency of the particular digit occurring in the data;
analyze, based on the threshold value, the data to detect a potential fraudulent transaction in the data; and
report an occurrence of the potential fraudulent transaction.
16. The non-transitory computer-readable medium of claim 15 , where the particular digit is a leading digit.
17. The non-transitory computer-readable medium of claim 15 , where the one or more instructions to determine the threshold value comprise:
one or more instructions that, when executed by the at least one processor, cause the at least one processor to:
receive an initial parameter associated with a historical analysis of the type of the data; and
determine the threshold value based on the initial parameter.
18. The non-transitory computer-readable medium of claim 15 , where the one or more instructions to determine the threshold value comprise:
one or more instructions that, when executed by the at least one processor, cause the at least one processor to:
determine a quantity of samples included in the data; and
determine the threshold value based on the type of the data and the determined quantity of samples.
19. The non-transitory computer-readable medium of claim 15 , where the one or more instructions to analyze the data comprise:
one or more instructions that, when executed by the at least one processor, cause the at least one processor to:
generate, based on the data, a model or a set of rules to detect an anomalous transaction; and
analyze the data further based on the model or the set of rules.
20. The non-transitory computer-readable medium of claim 15 , where the one or more instructions to determine the threshold value comprise:
one or more instructions for that, when executed by the at least one processor, cause the at least one processor to:
classify the data based on the type of the data, and
determine the threshold value based on classifying the data.