Paypal, Inc.

Chunmao Ran

Adam Cohen

Avishay Meron

Itzik Levi

Xing Wang

David Stein

Chuanyun Fang

Kun Fu

Doron Hai-Reuven

Matias Rotenberg

Amnon Jislin

Methods and systems are presented for classifying a particular user account as a fraudulent user account by analyzing links between the user account and two or more known fraudulent user accounts collectively. Attributes of the particular user account are compared against attributes of a plurality of known fraudulent accounts to determine that the particular user account has shared attributes with a first known fraudulent account and a second known fraudulent account. The shared attributes with the first known fraudulent account and the second known fraudulent account are analyzed collectively to determine a risk level for the particular user account. The risk level may indicate a likelihood that the particular user account corresponds to a fraudulent account.

Systems and methods for classifying accounts based on shared attributes with known fraudulent accounts

Placing an event into a particular cluster can allow various inferences about the event. A new payment transaction that looks similar to a previously identified cluster of mostly fraudulent payment transactions, for example, may be higher risk. The present disclosure includes structural data improvements to the way that online clustering of events (which may include web events and not just payment transactions) occurs. A new event can be classified into a particular segment very quickly using feature table searching, which can allow for better decision making when a short timeframe is required (e.g. transaction processing, online advertising, etc.).

Rapid online clustering

Machine learning techniques are used in combination with graph data structures to perform automated classification of accounts. Graphs may be constructed using a seed node and then expanded outward to second-degree nodes and third-degree nodes that are connected to a seed user account node via direct interaction between the accounts. Characterization information regarding the interaction between accounts can be stored in the graph (e.g., quantity of interactions, types of interactions) as well as other metrics and metadata. A classifier, using random forest or another technique, may be trained using a number of different graphs that can then be used to reach a determination as to whether a user account falls into one particular category or another. These techniques can identify accounts that may be violating terms of service, committing a security violation, and/or performing illegal actions in a way that is not ascertainable from human analysis.

Machine learning and security classification of user accounts

Rapidly handling large data sets can be a challenge, particularly in situations where there are millions or even hundreds of millions of database records. Sometimes, however, a service level agreement necessitates that a service return a response to a query in a small amount of time. Database organization techniques can be used that reduce potentially large datasets to smaller groups (neighbors) based on uncommon but shared attributes, in various instances. Using a limited set of related records, queries can be answered using a focused approximation based on characteristics of various identified clusters of records in the set of related records. A particular record may also be associated with an existing cluster of records based on that record's similarities to records in the cluster.

Database optimization concepts in fast response environments

Growing community of inventors

Shanghai, China

Chunmao Ran