What is Check Fraud?

Check fraud is a form of financial crime involving the illegal use of checks or check images to make unauthorized purchases or withdrawals. Methods include manipulation of existing checks or fabrication of counterfeit checks. The problem has evolved to include both individual and organized crime, with tactics such as mail theft and use of money mules.

Check Fraud Categories

Check fraud is primarily categorized as On-Us and Deposit fraud, with secondary categories of first party and third party fraud. This classification helps institutions tailor detection strategies.

• On-Us Check Fraud: Fraudulent checks processed by the paying bank.
• Deposit Fraud: Fraudulent checks deposited at a financial institution.
• First Party Fraud: Account holder exploits their own account.
• Third Party Fraud: External entity uses stolen/falsified info.

Check Fraud Types & Definitions

Common types include counterfeit checks, altered checks, forged checks, and check kiting. Each type involves unique tactics and risks.

• Counterfeit Checks: Fake checks mimicking real ones.
• Altered Checks: Legitimate checks modified to deceive.
• Forged Checks: Stolen checks with forged signatures.
• Check Kiting: Exploiting clearing delays to access non-existent funds.

Check Fraud Tactics & Schemes

Fraudsters use various tactics such as check washing, cooking/baking, account takeover, new account fraud, forged endorsements, mobile check fraud, and duplicate presentment.

• Check Washing: Chemically erasing ink to alter checks.
• Check Cooking/Baking: Using image editing to create fake checks.
• Account Takeover: Gaining access to victim's account.
• New Account Fraud: Opening accounts with false identities.
• Forged Endorsements: Illegally signing checks for others.
• Mobile Check Fraud: Exploiting mobile deposit channels.
• Duplicate Presentment: Depositing the same check multiple times.

Check Fraud Market Trends

Check fraud attempts are rising rapidly, with technological advancements making it easier for criminals. Organized crime, mail theft, and digital channels are key drivers.

• 600% increase in check fraud attempts since 2014
• Decline in check volume but increase in fraud attempts
• Shift from credit card to check fraud post-EMV chip introduction

OrboAnywhere Check Fraud Modules

Modular solutions for check fraud detection, including On-Us Fraud, Deposit Fraud, Validate, Positive Pay, Payee, and Compliance modules.

• Image forensic AI and transaction analysis
• Payee and account holder validation
• Payment negotiability and compliance checks

OrbNet AI Innovation Lab

Introducing the OrbNet AI Innovation Lab, a virtual innovation lab for evaluating and accelerating AI-based technologies for banking, check payments, and fraud detection. Led by OrboGraph's CTO and a team of data scientists, the lab formalizes the process of developing ANN-based products for faster time to market and optimal performance.

• Defines the technology foundation (AI and deep learning)
• Reinforces company charter to solve challenges in financial and healthcare payments
• Formalizes development process and aligns with industry leaders
• Demonstratable results with agile development
• Provides early access for POC scenarios (offline and on-premise testing)

A New Development Process

Describes the data-driven, AI-based development cycle replacing traditional algorithm-centric approaches. Models learn from large labeled datasets, enabling faster and more accurate predictions.

• Problem definition
• Dataset creation
• Model training infrastructure
• Deep learning model creation and optimization
• Model training and processing
• Integration into business modules
• Inference and ongoing model reinforcement

OrbNet AI Technology

OrbNet AI uses deep learning models not reliant on traditional recognition algorithms or OCR. The lab develops highly optimized models using advanced AI technologies and GPU acceleration.

• Uses CNN, RNN, Gradient Boosted Decision Trees, Gen-II text classification
• Built on TensorFlow RT and CUDA for GPU optimization

Enhancing Fraud Detection Capabilities

AI and deep learning enable banks and financial institutions to detect fraudulent activities before customer funds are accessed. Models analyze transactional history to spot suspicious activities in real time.

• Detects identity theft, COVID scams, stolen checks, healthcare fraud
• Learns from transactional history to spot suspicious activities

Graphical Processing Units

OrbNet AI leverages GPU acceleration for deep learning, enabling scalable and efficient processing of large volumes of data.

• Uses GPUs (e.g., NVIDIA Tesla V100) for high-performance deep learning
• Models can consume 1-10 billion FLOPS per CAR field

OrbNet AI Process Flow

Describes the AI-based process for recognizing fields on checks and documents, which is more efficient than traditional OCR/ICR/CAR/LAR techniques.

• Field Detection (object detection)
• Text Classification
• Interpretation (output values and scores)

Looking Forward

OrbNet AI and OrbNet Forensic AI represent the future of check recognition and fraud detection. For more application benefits, visit OrboAnywhere.

Modernization of Legacy Check Systems

Financial institutions are modernizing legacy check systems to improve check processing and fraud prevention. AI, machine learning, and deep learning-based systems provide stronger automation, payment negotiability testing, fraud detection, compliance, and behavioral analysis.

• Improves teller, branch, mobile, RDC, archive, lockbox, and ATM image capture
• Justifies continued investment in check processing

Check Processing Terminology and Components

Explains the basics of check processing: CAR, LAR, and MICR, and their importance in automated check recognition.

• CAR: Automated recognition of the numerical amount on checks
• LAR: Reading the legal line for written amounts
• MICR: Reading bank routing, account, and check serial numbers using magnetic ink

Legacy Check Recognition Limitations

Legacy systems rely on traditional algorithms and OCR/ICR, resulting in lower performance and higher error rates.

• CAR/LAR performance: 75%-85%
• Error rates: 2-4%
• Difficulties with field location and handwritten fields

Check Recognition: AI-based Deep Learning Models

Describes how OrboAnywhere uses deep learning for field detection, text classification, and interpretation to automate and improve check recognition.

• Field detection using object detection neural networks
• Text classification with NLP and deep learning
• Interpretation and normalization of recognition scores

OrboAnywhere Check Automation Modules

Overview of the main modules: Anywhere Recognition, Anywhere Validate, and Anywhere Compliance.

• Anywhere Recognition: High-accuracy check reading
• Anywhere Validate: Payment negotiability validation
• Anywhere Compliance: Risk mitigation for regulatory compliance