This article first appeared in Digital Edge, The Edge Malaysia Weekly on April 24, 2023 - April 30, 2023

Scandal, corruption and fraud have long plagued government institutions across Asean, with the infamous 1MDB scandal in Malaysia and other high-profile scandals leaving a bitter taste in the mouths of citizens.

These issues caused a range of consequences for the economy, society and governance. Unfortunately, many cases of fraud or scandal are not discovered until significant financial losses have occurred — normally costing millions or even billions of taxpayers’ money, leading to public outrage and loss of trust in government institutions.

Governments have doled out strict punishments to offenders in standard cases across the region. However, fraudsters and scammers are never fearful of the consequences and will never stop looking for potential gaps that could help them “earn” quick money.

Many fraud prevention strategies are applied by governments such as fraud risk assessments, whistle-blower policies and conventional rule-based systems, but fraud still can’t be prevented before it happens.

Late detection is a significant challenge faced by government agencies, leaving them incapable of taking immediate preventive actions before fraud or scandals happen. This is an indication that the traditional methods of fraud detection and prevention may not be sufficient to keep up with the evolving tactics and strategies of fraudsters, making it increasingly important to leverage on advanced technologies to enhance their fraud prevention efforts and minimise the devastating consequences that late detection can cause.

This is the reason government officials must be vigilant in identifying potential signs of fraud and have the necessary tools and resources to investigate and address any suspicions on time.

A recent government fraud report by KewMann, an Asia-based artificial intelligence (AI) and big data analytics company, highlights why governments and the public sector should leverage on advanced technologies to enhance fraud detection and prevention to enable the automatic finding of patterns across huge quantities of streaming transactions and take immediate action before loss and damage occur.

Here are some advanced technologies that could come in useful:

Supervised and unsupervised AI models together

Organised crime groups are very smart and can change their tactics quickly to avoid being caught. This is why using one-size-fits-all analytic methodology to detect fraud will not work. Instead, different methods such as supervised and unsupervised models in fraud detection should be used.

By doing this, governments can create advanced fraud detection strategies that are effective against the ever-changing tactics of fraudsters for different government departments.

Behavioural analytics and targeting

It works by analysing and predicting behaviour in detail, including financial and non-financial transactions, such as change of address, request for duplicate cards and password reset.

Profiles are created for each user, merchant, account and device, and updated in real time with each transaction, enabling accurate forecasts of future behaviour. By analysing patterns such as average expenditure velocity and transaction timing, and using various analytic models and profiles, fraudulent transactions can be identified and prevented in real time.

Knowledge graph

A knowledge graph is a database of facts and relations that can be used for fraud detection by creating a graph of relationships between different pieces of information available about individual users, such as account identification, user names, IP addresses and identification numbers.

By defining suspicious patterns and running graph algorithms, knowledge graphs can help identify known fraudulent behaviour. Data visualisation through graph visualisation can then be used to analyse large amounts of data and quickly identify patterns indicating fraudulent activity, investigate specific transaction patterns, find anomalies in real time and visualise complex relationships between entities.

Self-learning AI and adaptive analytics

Machine learning is effective in combating fraud as fraudsters constantly evolve their tactics to compromise consumer accounts. Fraud detection specialists should consider adaptive solutions to improve performance by sharpening reactions, particularly on transactions close to the investigative triggers.

Accuracy is crucial in fraud detection, as there is a fine line between false positive and false negative events. The adaptive modelling approach automatically adjusts the predictive characteristics of fraud models, allowing analysts to accurately reflect the fraud environment and prevent new forms of fraud assaults.

Right combination of technologies to stop the ‘bleeding’

An income tax department in a developing Asean country with approximately US$25 billion annual tax revenue successfully improved its operational efficiency and fraud detection efficiency with KewMann. The department processed five times more cases and increased tax receipts 2.8 times more per investigation officer.

Another government agency in a different Asean country found approximately 30,000 potential fraudulent cases hidden in the agency after going through only the initial phase of the project with KewMann. It helped the agency to consider some policy and procedural changes to prevent government funds from being abused and found some common symptoms to act as guidelines for claims fraud.

These success stories demonstrate the enormous potential of advanced technologies in government operations. By leveraging the right technologies, governments and public sectors can prevent future frauds or scandals, restore public trust, and ensure that taxpayers’ money is being used effectively and responsibly.

Kew Yoke Ling is executive director of KewMann, an artificial intelligence and big data analytics company that leverages behavioural science to predict and influence human behaviour

Save by subscribing to us for your print and/or digital copy.

P/S: The Edge is also available on Apple's AppStore and Androids' Google Play.