Abstract: Targeted messages are sent from a server to a client device through an app installed on the client device, but which operates without sending any personal data from the client device to any remote device. Bloom filters are created from data about the target audience and loaded in the app. User logins are resolved against the Bloom filter to targeted messages based on user identifiers. In this way, the processing to deliver a targeted message is performed on the client device, rather than on any remote device, and thus no personal information leaves the client device. Real-time view-through and click-through measurement is performed by firing conversion events to an integrated mobile network.

Abstract: A system and method for efficiently matching into an extremely large data set uses two data stores. A search data store contains identifiers that correspond to localized portions of the main data store, along with matching values for that portion of the main data store. To perform a matching operation, the search data store is first searched for a matching value. Then the identifier for the localized portion of the main data store is retrieved from the search data store. Using this identifier, a search may be performed against only the corresponding localized portion of the main data store rather than the entire main data store.

Abstract: A system provides for a client to receive enhancement data without having personally identifiable information leave its systems. The system receives access to a client configuration and a data graph to perform configuration defined filtering and aggregation steps to produce a set of client files. These files contain a hashed version of PII from the data graph. They are then used by the client to match the identity of its population of objects to keys, the keys also being included in the set of client files. The client associates corresponding keys with objects in its own data graph, then requests enhancement data using only the keys. The data is returned using the matched keys without the use of personally identifiable information.

Abstract: A machine analyzes an entity resolution data graph using a curated peer review set of data structures. The machine uses independent peer data structures that model similar or related yet different universes and entities. These data structures may include other entity resolution data graphs or file-based data structures. The machine first performs candidate screening to determine which of these data structures meet the requirements for use in the analysis. The machine then selects from among the candidate peer data structures with which to perform the analysis. Finally, the entity resolution data graph is analyzed using queries against the selected peer data structures to provide an analysis of the quality of its data graph.

Abstract: A computerized method for implementing a matching platform for entities includes a real-time processing layer using real-time linking on an input data stream, and a high-throughput layer linking entities from a batch data source stream. The real-time linking results are stored in a state change store and the high-throughput linking results are stored in a state store. On-demand linking uses the state change store and the state store. A lambda switch ensures that the entity states remain consistent when the on-demand linking switches from an older generation of the state store and the stage change store to a newer generation of the state store and the state change store. These stores are implemented in an entity database that parses queries into instructions and creates an instruction stream to control entity matching.

Abstract: A system and method for processing a multiple data controller workflow fully honors data subject opt-outs but nevertheless takes advantage of the synergy of the combined data from multiple controllers to improve the effectiveness of messaging. A data services provider performs data resolution processing against the data from each data controller. Using the results of the resolution processing, opt-out lists that are in different formats and that may track different sorts of data may all be used for the joint campaign. Metadata is added to the data so that the source data controller is tracked through the processing. The data is then combined, resulting in a master opt-out list. If a data subject has opted out with respect to any controller, then the opt-out is applied to the joint campaign.

Abstract: A cyber security system creates a behavioral framework for evaluating the cyber security of an organization's computer systems based on its employees. The system leverages offline and online individual identity information and then translates this data to anonymous identifiers to protect privacy. The identifiers are used to pull data from an identity graph, which includes behavioral data. A business-to-business identity graph correlates the name of an organization that maintains the targeted computer system with the anonymous identifiers of employees. Online activity is gathered by pixels fired from websites accessed by user browsers and gathered by one or more remote servers.

Abstract: An identity graph is partitioned into records pertaining to entities with similar data. A single authentication string or asserted relationship (AR) that represents a person identifies the person in the graph whose composite information is closest to the input by using all of the available context provided by the input string. By providing more contextual data to the match service than simply a name and single touchpoint, the service is able to match to a more focused and unique set of persons. In certain embodiments, the invention allows for the ability to take a small set of input data and more accurately identify its location in an identity graph much faster in time and lower in computational cost.

Abstract: A method for edge profiling within a data graph measures the impact that a particular type of edge has on a data graph. A list of all edges contained within a single connected component within the graph is generated in order to search for bridges across connected subcomponents. The process is implemented as two MapReduce jobs on separate compute clusters. The first job is the edge profile job, which is implemented as a Map-only job. The second job reads the output of the first, and builds multiple in-memory data structures representing each connected component within the data graph. After the graph is created, it is traversed to find bridges.

Abstract: An entity resolution data structure system compares two data graphs by creating a confusion matrix in a distributed processing environment. A benchmark file is created from one data graph for comparison to a reference data graph. Identifiers and metadata are appended to the benchmark file to allow the comparison to take place and the construction of a confusion matrix. The confusion matrix provides a high-level indication of the results of the comparison. When the data graphs contain personally identifiable information (Pll), the process does not require any Pll to be transmitted in either direction between the parties who maintain the data graphs to be compared.

Abstract: The invention is a system based on decentralized ledgers to enable compliance with privacy regulations. A consumer identifier (consumer ID) allows consent provenance to be saved in a shared decentralized ledger. The consent ID also empowers consumers to query as to how their data has been moved through the data supply chain. The consumer data itself is not stored in the blockchain, only consents and data transfer events. A consent API allows companies to enter the original consumer consent in a consent ledger; a data transfer API allows companies to record transfers from one to another; and a subscription API allows companies to be notified of changes to a given consent.

Abstract: A system and method for propagating an opt-in through an identity graph fully honors opt-in requests without otherwise modifying the graph data structure. From a node corresponding to the touchpoint of the opt-in, the graph is traversed to find the associated primary node, from which all connected edges are traversed. Nodes on paths that are not connected to other primary nodes are opted in, along with the primary node. Edges that lead to nodes which have edges to other primary nodes are not opted in, but only the edge from the primary node itself is opted in. If household nodes are used, an identifier for the person at the household node may be opted in without opting in other members of the household node by opting in only the edge between the primary opt-in node and the household node.

Abstract: In one aspect, a computerized method for implementing a matching platform for entities includes the step of, in a real-time data processing layer, implementing a real-time linking on an input event stream. The method includes storing an output of the real-time linking in a state change store. The method includes the step of, in a high-throughput layer, implementing a high-throughput linking of entities from a batch data source stream. The method includes storing an output of the high-throughput linking of entities in a state store to generate a unified and consistent view of the entities across a different representation of the entities. The method includes implementing an on-demand linking using the state change store and the state store.

Abstract: A system and method for extracting search terms for corresponding data elements from a natural language document identifies meaningful words within the context; identifies and structures the keywords; expounds on the keywords to optimize the search results; and captures the most relevant data elements from the corresponding database. Predetermined demographic characteristics and short (one- or two-word) search phrases that capture descriptors of behavioral characteristics are structured in the process. The result of the completed process yields a parameter set naming demographic and behavioral characteristics along with a structure that is optimized for search within a database comprising a large number of data elements.

Abstract: A system for “horizontal” salting of database tables, text files, and data feeds utilizes a key field and character position within that field (the “Key Character”) and a Salting Field, which contains content that can legitimately be in one of at least two states without impacting the usefulness of the data. A unique identifier, which is assigned to the recipient of the data, is hidden within the data by using the variations of the states in the Salting Field, with the value of the Key Character identifying the position within the unique identifier. This type of salting is invisible to the recipient of the data file, does not alter the accuracy of the data, and can be made unique for a particular party receiving data files or unique for each data file.

Abstract: A system for applying fingerprinting/watermarking of consumer data, and analyzing “wild files” of consumer data to assign a guilt score for a particular party who may have leaked the data, allows the owner of data sources (“Data Owners”) to identify and assert ownership of textual data that has been distributed outside of their firewall in the clear (i.e., without encryption), either intentionally or unintentionally, and assign guilt to parties misusing the data. The system can be used by Data Owners who transmit, lease, or sell data to individuals or organizations (“Trusted Third Parties” or “TTPs”) to recognize and assert ownership of their data in the case where one or more TTPs leaks the data (the leaked file is defined as a “Leaked Subset”) into the hands of others (“Bad Actors”) who either knowingly or unknowingly use the data illegally.

Abstract: A system for applying fingerprinting/watermarking of consumer data, and analyzing “wild files” of consumer data to assign a guilt score for a particular party who may have leaked the data, allows the owner of data sources (“Data Owners”) to identify and assert ownership of textual data that has been distributed outside of their firewall in the clear (i.e., without encryption), either intentionally or unintentionally, and assign guilt to parties misusing the data. The system can be used by Data Owners who transmit, lease, or sell data to individuals or organizations (“Trusted Third Parties” or “TTPs”) to recognize and assert ownership of their data in the case where one or more TTPs leaks the data (the leaked file is defined as a “Leaked Subset”) into the hands of others (“Bad Actors”) who either knowingly or unknowingly use the data illegally.

Abstract: A system and method for transforming input data in a data graph is structured in such a way that it does not destroy embedded contextual data yet also keeps the number of edges in the data graph sufficiently small in number that computation with respect to the data in the data graph is feasible with existing computational resources on extremely large graph sets. Incoming data is represented as a collection of “cliques” rather than placing each data object into its own node in the graph database. Maintaining the clique structure though the graph build pipeline dramatically reduces the exponential increase in the number of edges in the graph, while also maintaining all of the contextual data presented on the input record.

Abstract: Change fingerprinting is applied to a text file, database table, or data feed to determine the timeframe in which an identified “wild file” was generated, even when its file creation meta-data is missing. Each row in the data contains information on a single object. At least one column in the data contains an age for each object at the time the file was created. The age data can be used to determine the date the file was created, such as by using recognition processing or by looking at data that has been added or dropped from the file based on age. By identifying the timeframe in which the wild file was created, the data owner may greatly reduce the computational burden needed to determine if the wild file contains stolen data because it greatly reduces the universe of files that must be compared to the wild file.

Abstract: A system and method for identifying a leaked data file and assigning guilt to one or more suspected leakers proceeds through a plurality of levels. At a first level, primary watermark detection occurs. Data is inserted into a subset of data to determine correlation with data in the suspected leaked file. The guilt probability that results is then weighted based on the number of bits matched. In a second level, another search process is performed for detecting additional salt-related patterns. The guilt score is then computed for every detected recipient identifier for the suspected leaked data file, and the relative guilt of these recipients is weighted. In a third layer, the statistical distribution of data in the suspected leaked file is compared with that of corresponding data in the reference files. After this layer is complete, the average of guilt scores across each of the layers is calculated.