Abstract: Methods and systems for a device identification system may be provided. The device identification system may determine an identity of a user device associated with a transaction. The identity may be determined by network address information, hard link information, soft link information, and/or other such information. The network address information may include IPv4 information, IPv6 information, a device ID, and/or other such information. The identity of the user device may be determined and a transaction conducted from the user device may be assigned a fraudulent transaction risk score according to the information. Transactions that are determined to be at a high risk of fraud may be reviewed or otherwise flagged and/or canceled.

Abstract: Techniques are disclosed relating to transitioning between data stores using a data migration server. In some embodiments, the data migration server may be used to access data stored on a preexisting data store to service requests from a plurality of services. A dual-write operation mode for the data migration server may then be enabled such that, in response to a given write request, the data migration server writes a given data entry to both the preexisting data store and a replacement data store. Further, a dual-read operation mode may be enabled such that, in response to a given read request, the data migration server reads a corresponding data entry from both the preexisting and replacement data stores. Configuration settings for the data migration server may then be adjusted to designate the replacement data store as the primary data store to service requests from the services.

Abstract: Methods and systems for generating a universal computer model for assessing a risk in an electronic transaction based on one or more risk assessment models are presented. The one or more risk assessment models may be incompatible with each other. Different portions of a risk assessment models may be extracted from the risk assessment models. A node structure is generated for each risk assessment model based on the portions extracted from a corresponding risk assessment model. The node structures generated based on the risk assessment models are merged to produce a merged node structure. The universal computer model is generated based on the merged node structure.

Abstract: An integration platform provides a third party channel products from a merchant, so that the third party channel may cause an opportunity to be presented to a user to purchase one of the products. In response to an indication of a user purchase, the integration platform transmits a purchase request to the electronic commerce platform to create an order for the purchase. The integration platform may create a payment for the purchase with the merchant's payment processor and transmit the payment to the payment processor, update the order with the merchant's electronic commerce platform with the payment information, update the status of the order to shipped, and capture payment information by communicating with the payment provider. The merchant's electronic commerce platform may then update the order with the integration platform, which may notify the third party channel of the update so that update may be communicated to the user.

Abstract: A method includes training a first machine learning model based on a set of training data and based on the training, determining a first performance metric corresponding to the first machine learning model. The method also includes determining one or more past performance metrics corresponding to one or more machine learning models that were previously trained based on the set of training data. Based on the first performance metric and the one or more past performance metrics, the method includes automatically selecting a second machine learning model to train based on the set of training data.

Abstract: Techniques are described relating to automatically classifying telephone calls into a particular category using machine learning and artificial intelligence technology. As one example, calls to a customer service phone number can be classified as related to prohibited activity, or as legitimate. In particular, a number of different telephony variables as well as additional variables can be used to make such a classification, after training an appropriate machine learning model. The training process may use an externally provided call classification score that is provide by an outside entity as an input, and can be calibrated so that the output score of the trained classifier provides a score that corresponds to a real-world percentage chance of an unclassified call falling into a particular category. Thus, a classifier score of “95” can indicate that a call is in fact believed to be 95% likely to correspond to prohibited activity, for example.

Abstract: Systems and methods for data augmentation in a neural network system includes performing a first training process, using a first training dataset on a neural network system including an autoencoder including an encoder and a decoder to generate a trained autoencoder. A trained encoder is configured to receive a first plurality of input data in an N-dimensional data space and generate a first plurality of latent variables in an M-dimensional latent space, wherein M is an integer less than N. A sampling process is performed on the first plurality of latent variables to generate a first plurality of latent variable samples. A trained decoder is used to generate a second training dataset using the first plurality of latent variable samples. The second training dataset is used to train a first classifier including a first classifier neural network model to generate a trained classifier for providing transaction classification.

Abstract: Systems and/or techniques for facilitating online-monitoring of machine learning models are provided. In various embodiments, a system can receive monitoring settings associated with a machine learning model to be monitored. In various cases, the monitoring settings can identify a first set of data features that are generated as output by the machine learning model. In various cases, the monitoring settings can identify a second set of data features that are received as input by the machine learning model. In various aspects, the system can compute a first set of statistical metrics based on the first set of data features. In various cases, the first set of statistical metrics can characterize a performance quality of the machine learning model. In various instances, the system can compute a second set of statistical metrics based on the second set of data features.

Abstract: Techniques are disclosed relating to a delayed presentation of authentication challenge for users, such as in the context of a chat session. In various embodiments, a server system receives an indication of a request for service initiated by a user in a chat session within an application executed by a client device. The request for service involves an authentication of the user that is dependent on the authentication being successfully completed within a particular time period after the authentication is initiated. The server system delays the initiation of authentication for the request for service until a readiness condition is satisfied. The readiness condition includes the server system being available to process the request for service, as well as subsequently detecting engagement with the user relating to the request for service. In response to the readiness condition being satisfied, the server system initiates the authentication of the user.

Abstract: Methods and systems are presented for translating informal utterances into formal texts. Informal utterances may include words in abbreviation forms or typographical errors. The informal utterances may be processed by mapping each word in an utterance into a well-defined token. The mapping from the words to the tokens may be based on a context associated with the utterance derived by analyzing the utterance in a character-by-character basis. The token that is mapped for each word can be one of a vocabulary token that corresponds to a formal word in a pre-defined word corpus, an unknown token that corresponds to an unknown word, or a masked token. Formal text may then be generated based on the mapped tokens. Through the processing of informal utterances using the techniques disclosed herein, the informal utterances are both normalized and sanitized.

Abstract: A method is provided that includes accessing, by a server provider server of a service provider, a database storing associations between network addresses and locations. Additionally, the method includes determining a subset of the database corresponding to a first network address, each association included in the subset corresponding to an association between the first network address and a respective location. The method also includes in response to determining that the subset of the database satisfies one or more clustering criteria, calculating a representative location corresponding to the first network address, and storing an association between the first network address and the representative location in a second database.

Abstract: Systems and methods for predicting language dialects for a user to improve a user interface of an application are disclosed. In one embodiment, a system receives a request to determine a default dialect for an application executing on a user device. The system acquires user information corresponding to a user of the user device. Based on the user information and using a clustering algorithm, the system assigns the user to a cluster associated with a dialect. The system provides the dialect to the user device for display in a user interface as part of a sliding scale of selectable dialects of the application such that the dialect is provided as the default dialect for text and other message communications in the application. The system learns from the user's choice of dialect to provide more accurate dialect predictions in the future for other users.

Abstract: Systems and methods for integrated multi-factor multi-label analysis include using one or more deep learning systems, such as neural networks, to analyze how well one or more entities are likely to benefit from a targeted action. Data associated with each of the entities is analyzed to determine a score for each of the proposed targeted actions using multiple analysis factors. The scores for each analysis factor are determined using a different multi-layer analysis network for each analysis factor. The scores for each analysis factor are then combined to determine an overall score for each of the proposed targeted actions. The entities and the proposed targeted actions with the highest scores are then identified and then used to determine which entities are to be the subject of which targeted actions.

Abstract: Techniques are disclosed relating to computer network security. In some embodiments, a computing system generates a plurality of executable binaries that include alerting beacons for a computer network associated with a transaction service. The computing system then deploys, within the computer network, the plurality of executable binaries as traps to detect privilege escalation attempts within the computer network. In some embodiments, the computing system detects that one or more alerting beacons included in the plurality of executable binaries have been triggered. In response to the detecting, the computing system may transmit, to a security management system, a notification indicating the one or more triggered alerting beacons. The disclosed detection techniques may advantageously reduce breaches in network security, which in turn may reduce or prevent the loss of private data.

Abstract: There are provided systems and methods for actionable insight into user interaction data. A service provider server can access user interaction data associated with an interaction between a first communication device and the service provider server, and generates feature representations of the user interaction data, in which the feature representations respectively correspond to extracted features that include textual data features or audio data features. The service provider server can determine an intent of the interaction from the feature representations using a machine learning-trained classifier, in which the intent corresponds to a first actionable insight category. The interaction is mapped to a first cluster based on the intent, and the service provider server issues a remedial action for the interaction based on the mapping of the interaction to the first cluster, in which the remedial action is associated with a particular type of activity in the first actionable insight category.

Abstract: Aspects of the present disclosure involve systems, methods, devices, and the like for obtaining real-time estimates. In one embodiment, a novel customizable, transferable system architecture is presented that enables the characterization of a task, designated for time estimate, into steps or processes. The characterization may occur using the novel architecture which introduces an estimates service module that can provide an initial timing estimate by obtaining a composite time estimate of all steps. Timing of each of the steps can be predicted using a job scheduling enterprise designed to model the job or step. In another embodiment, delays are captured by time estimate monitors which can provide the alerts and delta time differences. Such alerts and time differences can be updated and presented to the user.

Abstract: Systems and methods for detecting network anomalies are described. These may include determining burst scores for external network resources, determining burst scores for internal network resources, and using the burst scores to construct a burst graph where the edges are weighted by the number of connections between each resource. The graph is then analyzed by a graph convolutional neural network to identify patterns from which anomalous network traffic can be detected and from which corrective action can be taken. These techniques can allow for better detection and mitigation of abusive network traffic, improve computer network security, and provide more robust access to networked computer resources.

Abstract: Methods and systems are presented for providing concurrent data retrieval and risk processing while evaluating a risk source of an online service provider. Upon receiving a request to evaluate the risk source, a risk analysis module may initiate one or more risk evaluation sub-processes to evaluate the risk source. Each risk evaluation sub-process may require different data related to the risk source to perform the evaluation. The risk analysis module may simultaneously retrieve the data related to the risk source and perform the one or more risk evaluation sub-processes such that the risk analysis module may complete a risk evaluation sub-process whenever the data required by the risk evaluation sub-process is made available.

Abstract: There are provided systems and methods for initiating a device security setting on detection of conditions indicating a fraudulent capture of a machine-readable code. A service provider, such as an electronic transaction processor for digital transactions, may provide in-person or device-to-device data transfers through machine-readable codes, such as to effectuate a payment from one mobile device to another. However, unauthorized devices may be in close enough proximity to also capture the code and impermissibly use the code. Thus, device security settings are used to detect whether fraud may occur in certain settings and implement an operation to hide a valid code. This may include operations to obtain information for the unauthorized device, make the code dynamic, or mask the code in a display. Once a nearby valid scanner is detected, such as through emitted light or sound, the valid code may be displayed.

Abstract: There are provided systems and methods for intelligent analysis of digital symbols for message content determination. A user may engage in messaging with another user, such as using a messaging platform that may include electronic transaction processing through a service provider. The service provider may have compliance rules, such as to avoid illegal, fraudulent, or otherwise noncompliant electronic transaction processing. In this regard, messaging between users may be analyzed to detect noncompliant messaging and activities by users on the service provider's platforms. The messages may include graphical images or icons, which may be converted to text so that plain text of the message may be analyzed. This may be done through a mapping of images-to-text, as well as an artificial intelligence model train to convert images into text. Thereafter, the plain text may be analyzed to determine if the message violates any compliance rules.