How the bank is rolling out post-quantum cryptography and improving algorithmic trading with cutting-edge tech
The financial services sector is at the forefront of quantum computing, not only due to the threat to encryption but also for its potential in other areas. Augmenting algorithmic trading with quantum computers is already yielding promising results, with realistic scaled-up prototypes at HSBC showing a 34% improvement in accuracy over classical approaches.
Speaking during a press briefing at IBM Research’s Zurich labs, Philip Intallura, HSBC’s head of quantum technologies, said the bank’s primary motivations for adopting quantum computing are competitive advantage and security.
The financial services sector is predicted to be one of the top beneficiaries of quantum-generated value by 2035, with $600 billion of potential gains at stake over the next decade, according to McKinsey.
But the most pressing incentive for banks to familiarise themselves with quantum computing is the security threat. Banks, after all, are a prime target for cybercrime, and according to the Global Risk Institute, there’s a 19%-34% probability that a quantum computer will be able to crack RSA 2048 public key encryption in a matter of hours by 2035. Even by 2030 the projected risk is 5%-14%. These are not the kind of odds that banks (or regulators) relish, which explains why HSBC is in the process of introducing quantum-safe algorithms (post-quantum cryptography, PQC) across its entire estate.
“Given the consequence of not being able to keep customers data and payments safe, for me [5% in five years] is a very, very large probability,” said Intallura. The window for replacing RSA is narrowing, as the US National Institute of Standards and Technology (NIST) has declared its end-of-life by 2035, and regulators are mandating its removal by providers of essential services.
Crypto agility
Over the past 18 months, HSBC has extensively tested PQC algorithms, including a practical demonstration of protecting transfers of tokenised gold – a service by which investors can trade gold via tokens with transactions stored in a distributed ledger (blockchain). The bank has been working with NIST to modernise its cryptographic ecosystem, with a discovery process to find where RSA and other vulnerable cryptosystems exist across its vast global estate followed by an implementation of post-quantum alternatives alongside them, so that when the time comes they can be switched over as seamlessly as possible, and ensuring crypto agility – the ability to quickly switch again should an algorithm prove insecure.
During these 18 months the conversation has changed, Intallura said. “There’s less of a debate now around when we think this will actually happen, and it’s becoming more of an expectation from our regulators.”
Quantum key distribution
As well as working to replace RSA, Intallura’s team has also been demonstrating quantum key distribution (QKD) – key exchange using quantum mechanics to ensure that eavesdroppers cannot intercept the keys undetected.
In one experiment, keys were generated in the bank’s London HQ, transmitted to a nearby datacentre and used to secure a 30 million euro-to-dollar trade. More recently, HSBC worked with the Monetary Authority of Singapore to send QKD-secured data between different banks in the country. This type of cooperation between competing banks and the authorities to test a new system that could strengthen their mutual security sets an “important example”, said Intallura. (The UK NCSC, by contrast, is sniffy, recommending that “QKD should not be solely relied upon for generating and distributing cryptographic keys.”)
Intallura said he and his specialist team are not interested, at least in the short term, in the academic race to achieve “quantum advantage,” performing tasks with quantum computers that are impossible with classical computing.
“We’re very much looking at how we can apply quantum technologies to strategically important parts of the business where there’s a clear commercial value opportunity associated with it,” he said.
“Our philosophy is how can we use the quantum computing technologies of today – including quantum-inspired technologies, quantum random numbers – anything that will give us a commercial edge. If I can beat what I’ve got, even just by small amount, then it’s going to give me the runway to get to what we’re all really after, which is quantum advantage.”
One area where there’s an opportunity to add value is algorithmic bond trading. As a market maker, HSBC needs to respond to a quote from a client with a price, in competition with other dealers. Price too high and they will likely lose the deal; price too low and profits will be reduced.
“This is quite an intricate problem,” Intallura remarked, explaining that the bank’s analysts run a machine learning model to find the profitability sweet spot.
In an experiment to see if the ML model could be improved, his team worked with IBM to introduce a quantum computer into the workflow, transforming “features” (trading characteristics of the bond) into quantum computational space (Hilbert space). “That allowed us to identify more complex patterns that we weren’t able to see classically,” Intallura explained.
“We identified those patterns and transformed them back, then we used those transformed features to train our classical model.”
The result? In the best case, a 34% improvement in accurately identifying the sweet spot for quote pricing compared to purely classical approaches, with reduced – but always positive – enhancements for other runs.
Intallura was careful not to claim quantum advantage (a highly contested term) at this stage, instead promoting it as “a really good example of how using quantum computing in quite an innovative way allowed us to get this enhancement.”
Importantly, he went on, his team is “using a real quantum computer on a real set of data at a real scale. So this is as ‘real-world’ as you can get now.”
It’s about adopting new technology to enhance existing processes, he added. “It does show that if you use quantum computing as part of your development toolkit your quants are able to develop more sophisticated models. This is the approach we’re taking.”
