By Vidya Ranganathan
SINGAPORE (Reuters) - Stung by billion-dollar fines for malpractice on their trading floors, the world's big banks are using 'fuzzy logic' tools such as relationship mapping and behavioural analytics to read the minds of would-be cheats among their traders.
On Wednesday six global banks agreed to pay $4.3 billion to settle claims they failed to stop traders from trying to rig foreign exchange markets, which came hot on the heels of similar fines for manipulating benchmark interest rates.
Older systems to catch misconduct, still in use at some firms, pick through conversations for trigger words, but they can be easily circumvented.
"Traders have moved on. They know their communications are being monitored; they are using different channels and new words," said Richard Moore, head of financial crime and security services at DBS Bank in Singapore.
The latest technology hopes to overcome that via behavioural analytics, using fuzzy logic, which instead of hunting for a smoking gun, builds up an understanding of the relationships and probabilities that might predict transgressions.
Moore said DBS employs analytics experts in its compliance team and is working on a module to extend their behavioural profiling of ATM transactions and cardholders to chats and emails.
"What we need is the ability to profile behaviour, to know what is normal behaviour and realise when it begins to change," he said.
A host of surveillance firms have stepped forward with analytics that promise to do that.
Singapore-based TrustSphere specialises in uncovering all the relationships a person has from all digital interactions across an organisation.
Two people colluding in a crime, such as fixing prices or trading confidential information, would need to have a strong relationship, said Manish Goel, founder of TrustSphere, whose clients include financial services firms and governments.
"It's not what you know, it's who you know. Once our technology sees a relationship, we have it mapped ... even though the actual collusion conversation may happen offline or in a coffee shop," Goel said.
Catelas, which has patented algorithms in behavioural science, specialises in social network and relationship analytics and has tied up with Nasdaq OMX, which has a market surveillance product.
In a typical such system, a computer algorithm sifts through thousands of email and messaging conversations, bundling them into topics. Another algorithm updates the traders' relationships and interactions map. A third aligns these results to the risk score of the trader, which is derived by software that maps business and individual risk profiles.
The objective is to match suspect conversations and behavioural clues with parties that have a relationship, said Michael Karbouris, head of Asia business development at Nasdaq OMX.
Complex event processing platform Apama, a unit of Germany's Software AG, is another tool used by banks to identify and prevent fraud and comply with risk controls.
The company says more than 40 trading institutions use it, including UniCredit, Royal Bank of Canada and ANZ Bank.
The banks, however, are not talking.
"When one side reveals their hand, the other side reacts to it, and then often people will work to find a way around that. It's very difficult to be entirely secure," said Giles Nelson, vice-president for Apama product strategy.
Even traditional word-based tools are getting smarter, enabling monitors to add channels of formal and informal communication to the search and update vocabulary in real time.
U.S.-based NICE Actimize sells software that surveys communication content, with a learning capability that constantly adds new terms to its suspect word cloud.
"New language can develop in a matter of minutes between two individuals, especially if they have a strong relationship," said Catelas CEO Eddie Cogan. "Traditional lexicon-based tools and their derivatives are impotent in the face of this new and informal language."
This arms race is having an impact on the working environment at the larger banks, where traders can now be required to surrender their mobile phones and the right to use private email.
"There is hardly any exchange of information or views. Hardly any conversation," said one trader at a bank in Singapore. "Because no one knows when the bank will use this information against you."
(Editing by Will Waterman)