Electronic payment network security
Systems and techniques are provided for electronic payment network security. Payment data including an origin and a destination for a payment in an electronic payment network may be received. A route of the payment in the electronic payment network may be estimated based on the origin and the destination. The estimated route of the payment in the electronic payment network may be input to a neural network. Fraud probabilities may be determined using the neural network. A fraud probability may include a value indicating a probability of fraud in the payment in the electronic payment network.
1 . A computer-implemented method comprising:
receiving, at a computing device, payment data comprising an origin and a destination for a payment in an electronic payment network;
estimating, by the computing device, a route of the payment in the electronic payment network based on the origin and the destination;
generating, by the computing device, separate vector representations of each of two or more segments of the estimated route;
inputting, by the computing device, each of the generated separate vector representations sequentially to an input layer of an encoder neural network of a neural network on the computing device;
determining, by the computing device, using the neural network, one or more fraud probabilities based on inputting a hidden layer of the encoder neural network to a decoder neural network of a neural network after the separate vector representations of each of the two or more segments of the estimated route have been input to the input layer of the encoder neural network; and
sending an electronic communication to at least one ledger system in the estimated route to cause the at least one ledger system to close or suspend an account of a connector system or suspend services usable through an account of a connector system based on the one or more fraud probabilities for the connector system, reject the payment, or initiate a review of the payment.
2 . The computer-implemented method of claim 1 , further comprising:
receiving a label for the payment data comprising one or more fraud probabilities comprising values indicating a probability of fraud in the payment in the electronic payment network;
determining a level of error of the neural network based on a comparison of the fraud probabilities from the neural network and the fraud probabilities from the received label; and
updating weights of the neural network based on the level of error of the neural network.
3 . The computer-implemented method of claim 1 , wherein each of the two or more segments comprises an indication of two ledger systems and one connector system.
4 . The computer-implemented method of claim 1 , wherein estimating a route of the payment in the electronic payment network based on the origin and the destination is further based on a known topology of the electronic payment network.
5 . The computer-implemented method of claim 1 , further comprising updating a risk score for at least one user, connector system, or ledger system in the estimated route based on the one or more fraud probabilities for that user, connector system, or ledger system.
6 . The computer-implemented method of claim 1 , wherein the origin comprises an identification of a first user and wherein the destination comprises and identification of a second user.
7 . The computer-implemented method of claim 6 , wherein the identification of the first user is a fingerprint comprising data associated with a computing device used by the first user to initiate the payment in the electronic payment network.
8 . The computer-implemented method of claim 1 , wherein the neural network further comprises a decoder neural network that receives output from the encoder neural network.
9 . A computer-implemented system comprising:
one or more storage devices;
a communications device that receives payment data comprising an origin and a destination for a payment in an electronic payment network; and
a processor that estimates a route of the payment in the electronic payment network based on the origin and the destination, generates separate vector representations of each of two or more segments of the estimated route, inputs each of the generated separate vector representations sequentially to an input layer of an encoder neural network of a neural network, and determines, using the neural network, one or more fraud probabilities based on inputting a hidden layer of the encoder neural network to a decoder neural network of the neural network after separate vector representations of each of the two or more segments of the estimated route have been input to the input layer of the encoder neural network, and wherein the communications device further sends an electronic communication to at least one ledger system in the estimated route to cause the at least one ledger system to close or suspend an account of a connector system or suspend services usable through an account of a connector system based on the one or more fraud probabilities for the connector system, reject the payment, or initiate a review of the payment.
10 . The computer-implemented system of claim 9 , wherein the processor further receives a label for the payment data comprising one or more fraud probabilities comprising values indicating a probability of fraud in the payment in the electronic payment network, determines a level of error of the neural network based on a comparison of the fraud probabilities from the neural network and the fraud probabilities from the received label, and updates weights of the neural network based on the level of error of the neural network.
11 . The computer-implemented system of claim 9 , wherein each of the two or more segments comprises an indication of two ledger systems and one connector system.
12 . The computer-implemented system of claim 9 , wherein the processor estimates the route of the payment in the electronic payment network further based on a known topology of the electronic payment network.
13 . The computer-implemented system of claim 9 , wherein the processor further updates a risk score for at least one user, connector system, or ledger system in the estimated route based on the one or more fraud probabilities for that user, connector system, or ledger system, wherein the risk score is stored in the one or more storage devices.
14 . The computer-implemented system of claim 9 , wherein the origin comprises an identification of a first user and wherein the destination comprises and identification of a second user.
15 . The computer-implemented system of claim 14 , wherein the identification of the first user is a fingerprint comprising data associated with a computing device used by the first user to initiate the payment in the electronic payment network.
16 . A system comprising: one or more computers and one or more storage devices storing instructions which are operable, when executed by the one or more computers, to cause the one or more computers to perform operations comprising:
receiving payment data comprising an origin and a destination for a payment in an electronic payment network;
estimating a route of the payment in the electronic payment network based on the origin and the destination;
generating separate vector representations of each of two or more segments of the estimated route;
inputting each of the generated separate vector representations sequentially to an input layer of an encoder neural network of a neural network;
determining, using the neural network, one or more fraud probabilities based on inputting a hidden layer of the encoder neural network to a decoder neural network of the neural network after the separate vector representations of each of the two or more segments of the estimated route have been input to the input layer of the encoder neural network; and
sending an electronic communication to at least one ledger system in the estimated route to cause the at least one ledger system to close or suspend an account of a connector system or suspend services usable through an account of a connector system based on the one or more fraud probabilities for the connector system, reject the payment, or initiate a review of the payment.
17 . The system of claim 16 , wherein the instructions further cause the one or more computers to perform operations further comprising:
receiving a label for the payment data comprising one or more fraud probabilities comprising values indicating a probability of fraud in the payment in the electronic payment network;
determining a level of error of the neural network based on a comparison of the fraud probabilities from the neural network and the fraud probabilities from the received label; and
updating weights of the neural network based on the level of error of the neural network.
18 . The system of claim 16 , wherein each of the two or more segments comprises an indication of two ledger systems and one connector system.
19 . The system of claim 16 , wherein estimating a route of the payment in the electronic payment network based on the origin and the destination is further based on a known topology of the electronic payment network.
20 . The system of claim 16 , wherein the instructions further cause the one or more computers to perform operations further comprising:
further comprising updating a risk score for at least one user, connector system, or ledger system in the estimated route based on the one or more fraud probabilities for that user, connector system, or ledger system.