March of the machines

Analytic Product Manager at BAE Systems Applied Intelligence Read time: 4 mins
No longer the stuff of science fiction, artificial intelligence is now blazing a trail around the world. Dr David Nicholson examines its burgeoning impact in the field of financial crime.
March of the Machines Just a few years ago the idea of artificial intelligence (AI) becoming mainstream so quickly was seen by many as a pipedream. A possibility, sure, but hardly guaranteed.

Today, smart home devices like Alexa and Google Home increasingly adorn domestic settings. Predictive algorithms shape our online shopping habits, our social media feeds and media streaming. Autonomous vehicles roam on the ground, in the air and at sea. Travel plans are made and shaped by real-time traffic data delivered our phones.

AI and machine learning are also being put to use in cracking down on criminal activity – even before it happens in some cases. While it’s no silver bullet, the new techniques it fuels make a particularly valuable weapon in the fight against fraud and financial crime.

Making money talk

Fraudsters and financial criminals often adapt their tactics to avoid detection by hiding the ‘signal’ of their nefarious actions within the ‘noise’ of regular transactions: this is where machine learning comes in. 
Historically, machine-based defences have relied on rule-based logic with low threshold settings to detect weak criminal patterns. Invariably, this generates a crippling number of false alarms that burden investigation teams.

Machine learning, on the other hand, excels in the recognition and discovery of complex shifting patterns immersed in a sea of noise and can generally detect them far more accurately and more effectively than static detection rules.

Fundamentally a data driven approach, machine learning figures out how to discriminate between criminal and non-criminal patterns without the need for an explicitly programmed rule. It learns a more powerful detection model from the data itself, digging deep into the data to mine for nuggets of unusual behaviour or anomalous patterns, even in the absence of any historical precedents.

Connecting the dots

Machine learning views data through the lens of statistical pattern recognition, but this is not the only lens at an analyst’s disposal. Take our own NetReveal system for example, which can also expose patterns of criminal behaviour by viewing data through the lens of a network representation.

Nearly all serious financial crime exploits the connected nature of the global financial system. Its fingerprints are an unusual network of interactions, typically involving multiple entities (people, accounts, companies) interacting through a range of mechanisms (transactions, account changes, real word events). The right tools connect these dots and automatically assign risk scores to networks, helping an analyst to conduct more effective and efficient investigations.

Completing a triad of lenses through which we can view data to detect financial crime is the temporal lens. Often it is the sequential order and frequency of events that helps us to separate potential criminal signal from noise, for example a peculiar pattern of physical cash withdrawals or deposits that are large when aggregated over a period of time.

Using machine learning and AI tools and technology to blend our view of the data through all three of these complementary lenses – statistical, network/graphical, temporal – is certainly key to a robust and effective solution for catching financial criminals who seek to evade detection.

Yet the human factor is also critical. The real future of AI in the fight against financial crime relies on skilfully blending multiple analytical perspectives with deep domain expertise and an investigator’s intuition; a true team of humans and machines.

Can this happen? It already is – and we’re just getting started.

About the author
Dr David Nicholson is Financial Crime Data Science Leader at BAE Systems Applied Intelligence
David Nicholson Analytic Product Manager at BAE Systems Applied Intelligence 9 May 2019