Framework and system for identifying partners in nefarious activities
AssigneeInfernotions Technologies Ltd.
InventorsVarun Madhok, Leszek Reimus
Life of this patent
Estimating the term…
Abstract
The invention is a method and a system for discovering collusion among a network of interacting entities. Collusion is defined as a conspiracy among two or more interacting entities to use deception or an unfair advantage for private gain above their peers or other stakeholders in their network. The invention comprises a mathematical framework for measuring the collusion propensity of two or more entities in a network.
Claims20
A method using a computer to identify collusion partners for use in assessing collusion risks involving two or more entities in an ecosystem network, wherein said ecosystem network having a plurality of entities, entity feature attributes, and transactions between entities, said method comprising the steps of:
- calculating an entity suspicion score for each of said plurality of entities based on each entity's feature attributes and on its relationships with other entities in its neighborhood in said ecosystem network;
- finding all possible entity pairs from said plurality of entities, therein each pair having a first entity and a second entity;
- searching all non-intersecting paths between said first entity and said second entity of each found pair of entities;
- calculating a collusion risk score for each of said non-intersecting paths;
- sorting collusion risk scores of all non-intersecting paths for said all possible entity pairs;
- and
- identifying collusion partners from entity pairs having top collusion risk scores.
A system of group collusion risk assessment for use in assessing collusion risks involving two or more entities in an ecosystem network, wherein said ecosystem network having a plurality of entities, entity feature attributes, and transactions between entities, comprising:
- a data layer capturing data on said plurality of entities, transactions between entities and entity features;
- and
- an analytical layer comprising:
- a fraud suspicion score module computing entity fraud suspicion scores for each of plurality of entities;
- a path finding module finding non-intersecting paths among all entities in said ecosystem;
- a collusion risk score module computing collusion risk scores for each of found path;
- and
- a collusion identification module for identifying collusion partners based on sorting said collusion risk scores among all said found non-intersecting paths.
A non-transitory storage media containing instructions for identifying collusion partners for use in assessing collusion risks involving two or more entities in an ecosystem network, wherein said ecosystem network having a plurality of entities, entity feature attributes, and transactions between entities, said non-transitory storage media comprising:
- instructions for calculating an entity suspicion score for each of said plurality of entities based on each entity's feature attributes and on its relationships with other entities in its neighborhood in said ecosystem network;
- instructions for finding all possible entity pairs from said plurality of entities, therein each pair having a first entity and a second entity;
- instructions for searching all non-intersecting paths between said first entity and said second entity of each found pair of entities;
- instructions for calculating a collusion risk score for each of said non-intersecting paths, wherein said collusion risk score is directly proportional to the sum of the entity suspicion scores of every entity in said each path and to compactness of said each path;
- instructions for sorting collusion risk scores of all non-intersecting paths for said all possible entity pairs;
- and
- instructions for identifying collusion partners from entity pairs having top collusion risk scores.
Dependencies are inferred from the claim text.