Recent surveys show a notable increase in online fraud in the Middle East. Visa, Dubai Police, and Dubai Economy (DED) revealed cardinal findings in 2021, shedding light on a significant percentage of UAE consumers who experienced online fraud. Besides, the big four consultancies shared several surveys with a remarkable rise in fraud and financial crime in the Middle East. At the end of PwC’s recent Global Economic Crime and Fraud Survey, it states a gap between the good intentions of Middle East organizations to prevent fraud and their ability to improve their performance in this area. A couple of weeks ago, Dr. Scott Nowson -AI lead @pwc ME- dived into harnessing the top notch technology to reduce the false positives for anti-money laundering, payment fraud, or financial crime.
At Zero&One, we have raised the flag to compact online fraud. The ML team is sharing the end-to-end technical demo with different approaches to identify cases that represent financial and regulatory risks and show the power of machine learning models on AWS.
The dataset used to demonstrate the fraud detection solution is the dataset collected and analyzed during a research collaboration between Worldline and the Machine Learning Group (http://mlg.ulb.ac.be) of ULB (Université Libre de Bruxelles) on big-data mining and fraud detection.
Github: https://github.com/MohAlbakerKaw/Fraud-Detector.git
Technology advances and new advantages come to light, but it is not without any problems. New undiscovered issues arrive with everything new. One of the issues that have advanced with the advancement of technology is fraud. Fraud existed since the beginning of humankind, however after online transactions and payments became a thing (add more here on where fraud can occur) it proved a gift for online hackers, exploits and fraudsters as the main types of fraud experienced by consumers are phishing, credit card fraud and receiving counterfeit goods (Research more on types of fraud for consumers and businesses).
The survey conducted by Deloitte in 2021 entitled Middle East Fraud Survey, found that 48% witnessed an increase in fraudulent incidents compared to earlier years, with the leading cause for fraud over the last two years in the MENA region being Cyber-crime and technology frauds which stands at 24%.
According to PwC Middle East Economic Crime and Fraud Survey, in the region, traditional fraud types continue to feature prominently, compared with the global survey average. Procurement fraud, which may include the practice of favoring known associates with vendor and supplier contracts, remains a significant and growing problem. In 2018, 22% of Middle East respondents said their organization had suffered procurement fraud. In 2020, the proportion has risen to 42%, more than double the global survey average of 19%. Customer fraud is also a growing problem for Middle East organizations, with 47% of respondents reporting an incident during the past two years, up from 36% in 2018.
In addition, a 2020 UAE cybercrime survey by KPMG revealed that 73% of respondents anticipate their business to invest in changes to their cybercrime prevention initiatives. Compared to the rest of the world, the middle east is expected to have a high increase and commitment in fraud combat.
The increasing fraud due to technology has resulted in the development of counter measurements to reduce the impact and losses. Those counter measurements include an increase in implementation of anti-fraud policies and organizations increasing the spending on combatting fraud/economic crime. Organizations are learning more and ready to dedicate resources to fighting cybercrime. One of the most advanced systems used to win the fight is machine learning. It helps in recognizing and analyzing the patterns, which in turn helps in understanding and preventing threats with same or similar patterns. In addition, machine learning helps cybersecurity teams be more proactive in preventing threats and responding to active attacks in real time.
Moreover, after fraud became a grave issue, major companies started creating services devoted to countering the problem and reducing the impact. Amazon was one of those companies, their web services provide various machine learning services that aid in forming the most efficient applications. The most notable of these services for fraud detection are Amazon SageMaker and Amazon Fraud Detector.
Amazon SageMaker is a Platform as a Service (PaaS) that is used to build, train, and deploy machine learning models allowing users to focus on the development without having to worry about the infrastructure. It is the perfect service for organizations that prefer building their own models. It also provides built-in algorithms and pre-trained models through the AWS Marketplace to ease and speed up building fraud detection models. One of the major advantages it provides is the ability to scale up quickly and train models faster.
Fraud Detector is a fully managed machine learning service that enables customers to identify potentially fraudulent activities and catch more online fraud faster and in real time. This model has been developed after learning patterns from AWS for over 20 years while attempting to defraud Amazon.com, through evaluating the fraud data to generate model scores and model performance data. A decision logic can be configured to interpret the score and assign outcomes for each fraud evaluation. Amazon Fraud Detector is made specially for organizations with no machine learning experience as it can be set up and added to the solution application in a short amount of time. It has proven to be of a great addition to organizations that made use of it, such as Omnyex who has reduced fraudulent transactions by 6% and Icony has decreased the time dealing with fake accounts by 77%.
Our analysis has been put together on Amazon Fraud Detector and the usage of SageMaker platform to create machine learning models for fraud detection. Different approaches were used to demonstrate the usage of AWS for the use case of Fraud Detection. Besides the sample –provided by AWS- that demonstrate how to operationalize Amazon Fraud Detector, we are sharing three approaches that enable the deployment of fraud detector machine learning models on SageMaker.
Top comments (0)