Launching AI-driven Fraud Detection
If you’ve found this article, the chances are you’re looking for:
- A fraud detection tool that’s powerful, efficient, and scalable.
- A way to quickly tell what differentiates leading solutions.
- A straightforward list of features and use cases.
This article summarizes six top fraud detection software vendors, listing their key strengths and explaining the use cases they respond best to.
Top Fraud Detection Software Companies
Fraud Detection from ComplyAdvantage is an AI-based solution that enables firms to detect and prevent payment fraud efficiently. Our machine learning models have won hackathons organized by ACAMS and PwC and, in relation to fraud detection, have four key capabilities:
- Explainability: An “ensemble model” considers various fraud scenarios, explaining why each alert was created.
- Dynamic Thresholds: Calibrated to maximize precision and minimize false positives.
- Identity Clustering: Uses behavioral and personal characteristics to link accounts that a single individual or organization behind key fraud scenarios, such as synthetic identity fraud or money muling fraud, may control.
- Graph Network Detection: Uses graph network analyses to track fraudulent money within your system after the fraud has been committed.
Top ComplyAdvantage features:
Fraud Detection by ComplyAdvantage is well suited to digital and regional/mid-market banks, international money transfer and foreign exchange firms, PropTech, and many payment companies. To help these firms detect fraud effectively, our solution includes the following capabilities:
- Advanced machine learning (ML) detections – Capture unknown fraud risks rules alone can’t see, using ML models, identity clustering, and graph analytics.
- Anomaly detection – See the unseen with peer group and individual anomaly detection using unsupervised machine learning algorithms.
- Dynamic tuning – Fraud detection models continuously learn from analyst feedback.
- Fine-tuned, real-time risk approach – Apply different rules and thresholds for different customer tiers.
- Smart Alerts – Identify and prioritize alert risk factors, increasing team efficiency.
- Flexible rules engine – Leverage fraud and industry-specific rules and scenarios, or work with industry experts to set your own.
- 50+ fraud scenarios – Leverage our wide array of payment-agnostic fraud scenarios, including Account Take Over (ATO) Fraud, Synthetic Identity Fraud, Relationship Fraud, Email Compromise Fraud, Authorized Push Payment (APP) Fraud, and many more.
- Coverage for all payment rails – Rest easy with comprehensive payment coverage, including ACH, Swift MT, SEPA, Direct Debit, FedNow, Faster Payments, and SEPA ICT.
- Coverage for non-transaction events – Take advantage of holistic insights on customer behavior events like logins and profile changes.
- Integrated case management – Open an investigative case directly from an alert, automate repetitive tasks, and reduce human error.
- Custom workflows – Move transactions through the various states for reporting and escalation.
- Flexible data format – Use as much of the data you have as possible in your format and build your compliance program around it.
- Ongoing configuration – Independently perform ongoing rule and scenario configuration.
- Behavioral analytics – Take advantage of identity clustering fuelled by customer behavior and identity data to link accounts controlled by a single hidden entity.
- Neural network models – The wider ComplyAdvantage platform uses neural network models and algorithms to identify factors such as anomalies in transaction behavior. This includes, for example, discrepancies between reference text and transaction values to help identify fraud where the reference text may be used to mask the source or purpose of transferring funds.
According to Crunchbase, Featurespace “provides an adaptive behavioral analytics technology and automated deep behavioral networks for risk management.”
Top Featurespace features (1) (2):
- Advanced machine learning– ML-powered behavioral analytics lends powerful insights to investigations.
- Anomaly detection – Allows analysts to “spot suspicious activity and prioritize alerts.”
- Behavioral analytics – Use multiple data points such as device ID, customer behavior, and cross-channel transactions to detect fraud across payment types.
- An array of fraud scenarios – Including Deposit Fraud, ATO Fraud, Man in the Middle/Browser Fraud, Authorized Pushed Payment Fraud, Scams, and Money Muling.
- Coverage for non-transaction events – For a comprehensive view of customer behavior.
- Integrated case management – The 3.19 release includes streamlined case management so investigators can take advantage of AI insights for their cases in a single centralized view.
According to Featurespace, its customers include a number of major financial institutions, including HSBC, Worldpay, NatWest, and Danske Bank.
Crunchbase describes Unit21 as a “no-code platform for risk and compliance operations.”
Top Unit21 features (19) (35):
- An array of fraud scenarios – Including ATO Fraud, Elder Fraud, Compromised Card, Bust-out Fraud, and Money Muling.
- ML-driven alert prioritization – Fewer false positives and reduced backlogs thanks to AI-driven prioritization.
- Real-time customer monitoring – Take advantage of real-time alerts and allow analysts to action transactions before they go through.
- Centralized data ingestion – Use a single data view to gain a holistic perspective on customer activity.
- Flexible rule management – Independently create and test rules according to team needs without waiting for support.
According to Unit21, its customers include Intuit, Flywire, Permont Bank, and Bakkt.
Crunchbase describes Feedzai as a developer of “risk management tools to prevent fraud and money laundering in transactions.”
Top Feedzai features:
- Single omnichannel platform for detecting and preventing fraud across the customer journey.
- Behavioral models that are continuously learning to evolve with emerging fraud risks.
- ATO data integration, including biometric, network, and device information from each customer session to enrich and improve decision-making accuracy.
- Scam prevention using machine learning to detect behavioral changes related to scams and authorized push payment fraud.
- Out-of-the-box fraud and risk strategies to support instant decision-making on transactions.
According to Feedzai, its customers include Citi, Fiserv, Santander, and Lloyd’s Banking Group (36).
Crunchbase describes Sardine as a “fraud prevention and compliance software company for the digital economy.”
Top Sardine features:
- Single API through which behavioral patterns are monitored continuously to identify irregularities in real-time. This is designed to stop fraudsters before they create and fund their accounts.
- Integrated device intelligence and behavioral biometrics signals, feeding in data from phone, email, SSN, and documentary KYC to ensure global coverage, speed, and reliability.
- Real-time payment flow protection to safely fund wallets without holding times, prevent card fraud, and enable instant digital asset purchases.
- Single platform to manage, investigate and customize fraud detection operations.
- Automated case manager to quickly approve or decline suspicious users and transactions.
- No-code rule builder to customize risk scoring based on business requirements and to leverage pre-built rules.
According to Crunchbase, Hawk:AI is a “money-laundering detection & investigation platform.”
Top Hawk AI features:
- Transparent data in natural language – Augmented with visual representations of AI models and risk indicators, designed to help analysts investigate cases more efficiently.
- AI-based false positive reduction – Traditional rules are augmented with AI to reduce false positives.
- Pre-trained AI models for complex fraud typologies to support rule-independent detection of known risks.
- Behavioral analytics to detect deviations from expected behavior with models learning from customer transaction patterns to enable fraud detection without fixed thresholds.
- Self-service configuration enables customers to fine-tune and conduct hypothetical analyses based on real data in a sandbox environment.
Best Fraud Detection Software Companies: Side-by-Side Comparison
|Anomaly Detection||Yes||Yes (1)||Yes (9)||Yes (18)||Yes (26)||“Anomaly Detection:
Find anomalies and patterns that evade legacy systems that rely only on
|Identity Clustering||Yes: using behavioral and personal characteristics, links accounts that
may be controlled by a single individual or organization behind Synthetic Identity fraud, Money
Muling fraud, etc.
|Not listed||Yes (9)||Not listed||Not listed||“We look at features such as age, income, gender, type and duration
of business, and all other information we can collect. In combination with their behavior, customers
get grouped into different clusters which are behaving similarly.” (30)
|Number of fraud scenarios covered||50+ fraud scenarios, including Account Take Over (ATO) Fraud, Synthetic
Identity Fraud, Relationship Fraud, Email Compromise Fraud, Authorized Push Payment (APP) Fraud, and
|Various fraud scenarios include Deposit Fraud, ATO Fraud, Man in the
Middle/Browser Fraud, Authorized Pushed Payment Fraud, Scams, and Money Muling. (37)
|Various fraud scenarios mentioned, including (10)card not present fraudAPP fraud.||An array of fraud scenarios are mentioned, are mentioned, including ATO Fraud, Elder Fraud, Compromised Card, Bust-out Fraud, and Money Muling. (19)||“Sardine allows
companies to centralize their user workflow and risk management across their Fraud, Compliance, and
Trust & Safety teams through 100+ Transaction monitoring rules that cover any fraud, user
security, and money laundering typologies, all the while empowering teams to create new
approval/rejection rules on the fly and back-test them against historical data to benchmark
|“AI powers rule-independent detection of known
fraud typologies to deliver well-explained, comprehensive fraud
alerts, immediately stopping any type of suspicious transaction.Our Pattern Library contains a wide variety of pre-trained, off-the-shelf AI
models for complex fraud typologies. We make new patterns available to all customers immediately,
without additional costs.” (31)
|Dynamic Thresholding||Yes: calibrated to maximize precision and minimize false positives.||Yes (38)||Yes – “Master risk with detection controls that continuously improve. Rapidly adjust to
fraud countermeasures with straightforward interactions of rule, machine-learning and hybrid
|Yes (20)||Not listed.||“Behavioral analytics flag deviations from expected behavior. Our
models learn customer transaction patterns, allowing for detection of fraud without the need for
fixed thresholds.” (31)
|Detection Explainability||Yes: explains the reason each alert was created using an “ensemble
model” that takes into account various fraud scenarios.
|Yes (3) – “ARIC™ Risk Hub
uses advanced, explainable anomaly detection to enable financial institutions to automatically
identify risk, catch new fraud attacks and identify suspicious activity in real-time.“
|Feedzai’s risk engine ingests large volumes of data from any source and is pre-packaged with
rules and models optimized for various fraud scenarios(14)
|Not listed.||Not listed.||“Transparent data in natural language, augmented by graphical
explanations of AI models and risk indicators, helps fraud and AML analysts investigate cases
|Graph Network Detection||Yes: uses graph network analyses to track fraudulent money within your
system after the fraud has been committed.
|Yes (4) – “Holistic Graph Analytics models assess risk levels, detecting networks and
commonalities to identify relationships and anomalous behavior.”
|Not listed as an active feature. (15)||Yes – “Recently, the Unit21 team added ‘graphical link analysis’ to their
platform’s suite of tools. Powered by ReGraph and KronoGraph, these components provide link
and timeline analysis capabilities that make complex financial crime data simpler, clearer and more
intuitive than ever.”
|Not listed.||“Hawk AI uses Entity Resolution as one method to power its
AI-enhanced products…we apply it in real-time, combining internal and external data sources,
and augmenting it with an additional tech stack such as entity graphs and Elastic Search.
…Hawk AI uses Elastic Search to navigate within the Entity Graph.” (33)
|Behavioral Analysis||Yes: using behavioral and personal characteristics, Fraud Detection links
accounts that may be controlled by a single individual or organization behind Synthetic Identity
fraud, Money Muling fraud, etc.
|Automated Deep Behavioral Networks architecture which undermines scams and protects against fraud more forcefully and accurately than ever thought possible.”||Scoring is based on device, behavioral and transactional patterns that continuously learn over time. Our
hyper-targeted detection easily spots subtle changes in behavior.” (13)“FeedzaiSegment-of-One
constantly surveilles each
customer and builds intelligence that distinguishes normal behavior versus anomalous behavior for
each customer.” (9)
|Not listed.||“Sardine continuously monitors behavior patterns and detects device irregularities in
real-time to stop fraudsters before they create and fund their accounts.”“At the time of card or bank account linkage, our advanced fraud models analyze customer behavior,
and identity signals, with dozens of curated third-party data providers:
IP address location
Email account age
Social media presence
Credit bureaus.” (28)
|Behavioral analytics flag deviations from expected behavior. Our models learn customer transaction
patterns, allowing for detection of fraud without the need for fixed thresholds.” (31)
|User Interface?||User-friendly one-stop shop: a holistic view of
risk across fraud and AML for better decision-making.
|“One Risk Hub to view all customers involved in suspicious transactions.” (2)||Assess fraud from all channels in a single view with our comprehensive
architecture.” (13)Feedzai Genome provides an intuitive way to quickly identify emerging financial crime
patterns and perform more in-depth, more accurate investigations.” (9)
|“Unit21 provides a simple API and dashboard for digital ID
verification, customer onboarding, identity fraud and regulatory compliance.” (22)
Visualize, investigate potential threats, and monitor known bad actors from
one unified, fraud and compliance dashboard.
Quickly approve or decline suspicious users and transactions. Assign cases
for further review, or to return funds.” (28)
|Self-service no-code algorithm configuration manager
“FIGHT FRAUD AND IMPROVE AML COMPLIANCE WITH A HOLISTIC VIEW
|Customization||Yes: detect complex fraud risks tailored to unique
risk profiles.Flexible & self-service
rules engineCustomers can stay with their existing system and just use the AI overlay, or
they can use the full solution.
|A customizable user interface… configurable analyst dashboard, and tiered user access …”
(1)“Adaptive Rules Engine: Utilize Adaptive Behavioral Analytics
technology for adaptive rules which can be authored and configured by your risk analysts. Sandbox:
Author new rules, models, and thresholds – and test the impact of these changes against production
data.” (1)“All solutions are available with out-of-the-box rules and models as
well as customized models.” (6)
|Rules can be tailored and grouped to include specific scam clauses that can be adapted as fraudsters
change their patterns.” (10)
|“Our no-code solution is designed to be simple to work with, while still enabling anti-fraud teams to
create sophisticated rules that protect customers.”(23)“Risk and Compliance teams define the rules
and workflows for fraud and AML detection without the need for engineers.” (24)
|No-code Rule Builder:
Enable Fraud Ops to customize risk scoring based on your business needs, or
utilize hundreds of pre-built rules.” (28)
|Self-service configuration: Fine-tune the algorithm and conduct what-if analyses based on real data in a
sandbox, committing changes only when you are ready.
Configure user management, roles and workflows yourself at any time with our
no-code configuration manager.” (31)
|API Integrations||Yes: ComplyAdvantage’s REST API allows seamless sync with existing AML systems, such as case
management and customer
|Yes: “Integration via API: Connectivity via API Connectivity via API allows for easy integration with existing systems and workflows.
Retrieve real-time responses for active management of frauds.” (7)
|Yes: “Flexible deployment model Pre-integrated
with Feedzai for Transaction Fraud for Banking, standalone, or in conjunction with other Transaction
Monitoring solutions, SIEMs, or other fraud management systems via API.” (17)
|Yes: “Unit21’s fully customizable no-code platform provides a simple API and
dashboard for detecting, investigating and reporting on fraud, money laundering, and other
sophisticated risks across multiple industries.” (25)
|Not listed.||“We provide a RESTful JSON API, easy to integrate for many use
|API Data Ingestion||Flexible data formats – Use as much of the data
you have as possible in your format and build your compliance program around it.
|“Event API: Receives event data, validates the data against the data
schema and returns a HTTP response, with a business payload (real-time responses) or without as a
simple technical acknowledgment (asynchronous processing).” (7)
|“Our solution creates a clear and accurate picture of legitimate transactions versus fraudulent
ones by harnessing data from cross-channel, cross-product, and third party data to provide a
holistic view of how a customer interacts with the bank.” (9)
|Not listed.||Not listed.||“Dynamically score customer risk using internal and external data.
Add behavioral analytics for richer context.” (34)
|Payment Types Ingested (rails)||Comprehensive payment coverage including ACH, Swift
MT, SEPA, Direct Debit, FedNow, Faster Payments, and SEPA ICT.
|“Productized, out of the box Fraud Models available include; Instant
Payments, Bill Payments, ACH, Check, Wire, Person to Person (P2P), Debit Card, Commercial Card and
Credit Card payments (inclusive of Card Present and Card-notPresent), ATM transactions, and other
online banking initiated payment types.” (1)“Powered by the multi-award winning ARIC Risk Hub, ARIC White
Label protects your clients from suspect transactions across all payment types and channels.”
|Feedzai’s real-time fraud monitoring
solution works across multiple payment types including: cards, transfers,
instant transfers, digital wallets, deposits, and withdrawals. Our solution includes
production-ready APIs for a wide range of payments…” (9)
|Not listed.||Not listed.||Not listed.|
All information sourced from publicly available websites, and correct as of May 2023. If you’d like to request a correction, please e-mail [email protected] and we’d be happy to review this with you.
Next Steps: Explore Fraud Detection from ComplyAdvantage
Originally published 14 June 2023, updated 02 August 2023
Disclaimer: This is for general information only. The information presented does not constitute legal advice. ComplyAdvantage accepts no responsibility for any information contained herein and disclaims and excludes any liability in respect of the contents or for action taken based on this information.
Copyright © 2023 IVXS UK Limited (trading as ComplyAdvantage).