32
Making machines work for you MACHINE LEARNING
The uses of machine learning in the banking sector are many and varied. But not all banks are using this potentially game changing technology to maximum advantage. Guy Matthews reports
T
he deployment of machine learning (ML) tools in the banking sector is nothing new. For years, banks have been trying to automate the process of learning from data and so improve a
range of business processes.
Recent research by predictive analytics specialist Squirro into the use of ML technology by 200 global banks revealed that most respondents have evaluated its possible benefits, with 67% having already deployed some kind of ML solution. But there is uncertainty too, with 83% admitting they were still figuring out how best to use such solutions to solve business challenges.
“There is some confusion in the banking sector about which types of ML to apply to which task,” claims Alex Vaystikh, CTO of SecBI, a specialist in AI and Big Data analytics.
He cites the example of banks that use ML to detect anomalies that might indicate money laundering or cybercrime. “If you’re trying to detect an anomaly with ML, then you need a baseline to measure against,” he says. “But baselines are changing all the time, almost every day. A static baseline will create many, many alerts, at least 95% of which are false positives.”
The danger, he says, is that poorly deployed ML will trip so many false alarms that security experts end up spending more time investigating than they did pre-ML, with the main outcome being the stretching of already tight resources.
“At the moment, cybercrime against banks is asymmetric,” he says. “It’s a lot harder for the defender than the attacker. The attacker can write five lines of code and they’re in; the bank needs huge amounts of code to defend. ML can close that gap.”
Jim Heinzman, executive vice president of financial services solutions at ThetaRay, a developer of ML solutions for the banking sector, explains that the ML technology that banks are using is often ‘rules-based’.
“With this kind of so-called supervised learning, you are taking
something you know about and teaching the system to find more things like that,” he says. “This doesn’t do anything about new threats you’ve never seen before.”
Heinzman claims that newer unsupervised ML techniques will speedily detect money laundering and other criminal activity, whether previously known about or not. “The aim is to find out something is wrong before the money has left the bank,” he says.
The threats that banks are facing are, claims Heinzman, unprecedented: “The bad actors are getting smarter and smarter, and using AI and ML to attack the banks. Using basic rules-based ML against this is like showing up to a gunfight with a knife.”
Banking regulators are part of the problem, claims Julian Dixon, CEO of anti-money laundering and Big Data specialist Fortytwo Data, as yet not permitting a fully automated approach to the problem. “The handbrake is on,” he says. “What banks can do at least is use ML to supplement
www.ibsintelligence.com | © IBS Intelligence 2018
exdez/iStock
Page 1 |
Page 2 |
Page 3 |
Page 4 |
Page 5 |
Page 6 |
Page 7 |
Page 8 |
Page 9 |
Page 10 |
Page 11 |
Page 12 |
Page 13 |
Page 14 |
Page 15 |
Page 16 |
Page 17 |
Page 18 |
Page 19 |
Page 20 |
Page 21 |
Page 22 |
Page 23 |
Page 24 |
Page 25 |
Page 26 |
Page 27 |
Page 28 |
Page 29 |
Page 30 |
Page 31 |
Page 32 |
Page 33 |
Page 34 |
Page 35 |
Page 36 |
Page 37 |
Page 38 |
Page 39 |
Page 40 |
Page 41 |
Page 42 |
Page 43 |
Page 44 |
Page 45 |
Page 46 |
Page 47 |
Page 48 |
Page 49 |
Page 50 |
Page 51 |
Page 52
