Let’s talk
thumbnail image

Financial Crime (Fraud)

hero image

The Challenge

Fraud is an enormous and growing economic drain, attacking many types of critical transactions and identity assets, including loan requests, credit card applications, payment authorizations, and account access. It’s been estimated that fraud costs the financial industry US$80 billion annually, while mid-sized businesses lose 5% of their revenue on average to fraud. The operational costs of investigating potential cases and the drag on customer satisfaction are also significant.

For each defensive measure that an organization takes, aggressive fraudsters find a new path to circumvent detection and prevention measures. This is especially true in the online world.

“TRADITIONAL ANTI-FRAUD TECHNOLOGIES HAVE TROUBLE KEEPING PACE”

Traditional fraud prevention systems have had trouble keeping up with the volume of transactions and innovations by conductors of fraud. Traditionally, most fraud identification systems have been data- and rule-based. For example, purchases over a certain amount or conducted in distant part of the world would trigger an alert. Most of the data considered has been from internal systems. In addition, the rules were developed at times independently through methodical analysis and construction by specialists.

While threshold-based and other types of rules have their place in fraud detection, one problem is that they only catch known behavioral patterns that people are able to articulate. To keep the logic manageable, the rules are generalized, so that it’s hard to hone in on reducing false positives through very granular risk scoring. In an era of growing transaction volumes from e-commerce, online lending, immediate funds transfer and digital wallets, older anti-fraud systems simply cannot process fraud-related logic precisely or fast enough.

 

BlackSwan Technologies Approach

BlackSwan Technologies offers a cutting-edge technological approach to discovering ever-evolving forms of fraud, automatically identifying many more fraudulent activities, and making your organisation’s fraud investigation team more efficient. This approach is embodied in our application software, ELEMENT™ of Financial Crime, which is based on the latest technologies for big data and artificial intelligence.

The application is an ideal solution for anti-fraud measures where:

  • Information about a business entity or individual, their situational profiles and relationships to others must be gathered from all accessible data sources and combined with transactional data to form a fraud risk assessment.
  • The myriad data sources include unstructured text such as news, relationship connections inferred from social media, and geospatial information including an individual’s location.
  • Behavioural patterns likely to be fraudulent require automatic detection by the software and evolution with new found instances.
  • Your organisation requires customisation flexibility in order to represent one or more proprietary risk-assessment models.
  • There is an balance to be struck between decisioning automation and fraud investigation team support – through escalations, case management tracking, visualisation of findings and data/workflow-based collaboration.

ELEMENT™ of Financial Crime’s cognitive computing approach enables fraud detection strategies that are not possible with many existing systems. Using network analysis and unstructured data interpretation, the software uncovers under-the-radar relationships among business parties, including executive/large shareholder relationships, mutual managers across organizations, and common counterparties. It also highlights unusual changes in the corporate structure of an entity, such as new affiliations or the appointment of executives and directors residing in locations other than that of the business itself.

Its detection engine accounts for time- and location- based dynamics,

so that acceleration in specific types of activity can trigger risk level elevation – a technique that is not possible with rule-based analysis alone. In addition, the application can combine time-stamped transactional information with network analysis to follow the flow of funds. It even detects a lack of economic interest by the customer in a location where they intend to open an account.

Unstructured data is interpreted using natural language processing. Concepts and relationships are extracted automatically from text, contributing to the web of connections between entities within the graph. Even discussions that may allude to fraud may be identified through sentiment evaluation and word clouds (word frequency/association visualization). With more than 80% of the world’s data being unstructured, it is essential to the scope and efficiency of fraud management to automate the incorporation of the information within the overall strategy.

BlackSwan Technologies’ approach results in far more comprehensive information being used to evaluate an entity or transaction, as well as up-to-the-minute recognition of new fraud pathways. It’s particularly suited to fraud in the domains of lending, payments and credit cards, identity, and certain forms of benefits/reimbursement claims.

ELEMENT™ of Economic Crime

ELEMENT™ of Economic Crime is built on BlackSwan’s ELEMENT™, our Enterprise A.I. Operating System. ELEMENT™ features four “hubs” that form an integrated Cognitive Computing framework. They are Sensory Hub for data processing, Cognitive Hub for a suite of advanced analytic engines, Discovery Hub for user interaction, and Management Hub for configuration and control. Here, select features of each hub relevant to fraud management are described:

Sensory Hub

Sensory Hub is a highly-scalable data integration and acquisition engine. It can ingest any structured or unstructured data, regardless of format, volume, velocity or method of acquisition. This includes CRM/KYC and transactional data as well as digitised notes, national fraud database records, SARs, open source intelligence, all kinds of supplemental information available through web crawling, even “dark web” communications. Sensory Hub also applies basic business and system policies and generates alerts, wherever immediate action is required. This immediate triage helps safeguard the integrity of instant money transfer/payment decisions.

Cognitive Hub

Cognitive Hub is the core of the application’s automated analysis and decisioning capabilities. It contains a robust set of cognitive computing capabilities to truly comprehend the business environment as an expert would, but with far greater data processing capacity and decisioning speed. Contextual analytics construct a depiction of every known party, attribute and transaction that may relate to potential case of fraud. Custom-definable ontologies capture the concepts and types of relationships in a business domain, and a knowledge graph represents the actual relationships between entities and their characteristics. ELEMENT™ of Economic Crime builds this graph in real-time, so it’s always fully up-to-date.

Cognitive Hub supports the three primary artificial intelligence methods of fraud detection: rule-based, structured and unstructured machine learning. Rules can be defined explicitly in the application to represent your domain knowledge and business policies while evaluating transaction/identity data parameters during transaction approval.

Rules do not automatically adapt to new types of fraud attempt. However, in Cognitive Hub they complement more cutting-edge methods of adaptive machine learning, that is, structured and unstructured learning algorithms. Structured learning relies on providing the algorithms with cases that have been investigated and reached a known resolution (i.e., fraudulent vs. not fraudulent). Unstructured, or deep, learning uncovers emerging or unusual patterns that may represent fraud, without requiring training data.

Discovery Hub

Discovery Hub focuses on the interaction by business analysts and investigators with the understanding and insights formed by the software. Foremost, it provides a visual overview of an entity’s characteristics and status. This makes it efficient for a user to assess the situational context of each case.

Users’ work queues display new and updated cases that the Discovery Hub escalated for investigation. The hub allows the analysts to explore the insights presented by the system via big data visualization methods and a variety of analysis tools (including link analysis, geographic and temporal analysis). Case information can be shared across departments and appropriate staff through collaborative workspaces or workflow routing. Discovery Hub, together with Cognitive Hub, ensure that fraud analysts and investigators staff make strategic use of their time, interpreting the most complex instances of possible fraud, and refining the organisation’s policies and practices, rather than poring over raw data.

Management Hub

This hub provides the control framework that enables you to configure your unique business logic, systems integrations, security and access privileges. There are distinct console areas for managing ontologies, taxonomies, workflows, users, access, integrations, data sources, rules, auditability and more. Use a drag & drop interface to define data pipelines, then build business logic and policy enforcement rules. Control user access according to configurable domains, security clearance or job level.

Examples of Capabilities

The power of ELEMENT™ of Fraud is well illustrated via examples, which connect fraud-related situations to features supported in the application.

 

Fraudulent behavior to detect ELEMENT™ of Economic Crime  capability
Takeover of a financial account is followed by test transactions then an acceleration of money movement internationally. Pattern discovery incorporates time and rate of change, as well as geo-location of activity, as standard dimensions.
After initial success with a fraud attempt, the offenders replicate that strategy across targets. Adaptive deep learning engine autonomously develops added sensitivity (weighting) to repeated behavior.
Similar but not exactly alike names and identification information are used when applying for an account/access. Entity-resolution engine resolves ambiguities across similarly-               identified parties. It automatically searches for new data sources to validation information as needed. Actions can be initiated based on findings; eg raise alert for an applicant using a new address.
Multiple actors from different entities act in concert. Knowledge Graph with ontologies and domain-tuned information search identify the nature of relationships across business entities, their employees, investors and shareholders, third party vendors, other individuals, using all internal and externally-accessible data sources. Relationships also can be inferred based on time sequencing and location of activities.
A relation (of another relation) of a customer appears in adverse news or bad actor list. Data integration hub sources from bad actor lists and open source intel and also scrapes adverse news. It then correlates high-risk parties with internal customer and compiled relationship data.
Suspicious behavior occurs well after a transaction has been processed. Fraud monitoring for customers and other tracked parties is continually updated during data processing and analysis. The system compares recent behavior to historic activity, including transaction amounts, counter parties and flow of funds.

 

Fraud may be related to data breach. ELEMENT™ offers an add-on application module that actively monitors cyber-threats and alerts appropriate staff and optionally customers.
Financial access attempts involve far flung locations. The analysis of transaction history tracks IP addresses, locations and timing of login attempts.

 

Forensic accounting is enlisted to look for evidence of occupational fraud by employees. ELEMENT™ performs automated discovery on the universe of accessible emails, documents and online interactions, highlighting allusions to improper behavior through natural language processing, contextual analytics, tone detection, and relationship graphs. Predictive coding identifies similar and related documents without resorting to keyword search.

 

“ELEMENT™ OF ECONOMIC CRIME SUPPORTS ONGOING MONITORING FOR CUSTOMERS AND OTHER TRACKED PARTIES”

Benefits of More Advanced, Accurate Fraud Prevention

Strengthening your fraud defenses using the latest technology can generate significant ROI, foremost by reducing fraud’s economic impact. Enterprises are finding a multiplier effect in terms of the number of investigations their team can pursue when many false positives are eliminated.
ELEMENT™ of Fraud delivers a broad set of capabilities in Cognitive Computing for fraud prevention. It simultaneously offer your enterprise the deployment speed and impact of an advanced, out of the box solution, with the strategic flexibility to customise, integrate and extend the solution to encompass your unique, fraud-related policies and systems. Total cost of deployment and operation is exceptionally low, compared to alternative solutions. This is a result of its unintrusive means of accessing data, autonomous learning capabilities, ease of configuration, and visual interfaces that make business users immediately more productive.

Augmented Intelligence for the Whole Organisation

BlackSwan expands the potential for A.I. to serve your enterprise in a number of ways.
Whereas five years ago, transaction fraud via fake account creation was common, today it’s more likely to be the result of the illegal acquisition of actual account data. For situations where fraud may be related to data breaches, BlackSwan has developed a Cyber-Threat Intelligence Hub. This monitoring application first constructs a knowledge graph for understanding potential business exposure in the operational, technological, and market/environmental realms. It then monitors active threat databases for information updates and financial transactional data for patterns that relate to cyber-attacks including data theft and mis-use.

BlackSwan offers additional A.I.-based business applications that may benefit your enterprise. One is ELEMENT™ of Trace for forensic accounting. It allows the auditor to trace the movement of digitally-represented assets and facilitate their recovery. Other applications include ELEMENT™ of Risk, ELEMENT™ of Compliance and ELEMENT™ of Audit.

We also offer our complete software platform, ELEMENT™, for big data, A.I./Cognitive Computing, contextual analytics, data visualization and decision support. It’s the world’s first Enterprise A.I. Operating System, enabling your organization to extend or build your own powerful, A.I.-based business applications.

Other applications

apps image

E-Discovery

A complete framework to support the identification, preservation, collection, processing, review, analysis and production of digital data in order to sustain mission critical business processes.

apps image

Risk

Detecting and quantifying different aspects of risk related to an enterprise operation, recommending policies for risk mitigation, and promoting offerings to address them.

apps image

Trade Intelligence

Generate deep and accurate profiles on targeted individuals and organizations, to monitor their reputation, mitigate risks and leverage business opportunities.

apps image

Transaction Intelligence

Interpreting events of temporal nature and monetary value, such as financial and commercial transactions, based on the event profile, related entities, and other related events.