We’re often asked: why do we even use blockchain, when so many other current use-cases have ended up back on more traditional solutions like databases?
On the face of it, blockchain sounds like a great fit for the insurance industry, where you have many different participants exchanging large amounts of data with differing levels of trust between the parties: having one immutable source of truth that is constantly being verified by all with a clear audit trail would solve so many inefficiencies and disputes. There is a catch, however, because insurers are quite rightly subject to strict controls over the personal data they handle, such as:
- regulations such as data protection, including the Right to be Forgotten; and
- compliance, which often demands an even higher standard of control, for example over which jurisdictions or entities a company chooses engage with or what sensitive competitive information they are comfortable to share.
Having an immutable record of personal information on blockchain would be an automatic breach of many data protection requirements, to note just one challenge, so does that mean blockchain can never be used? Clearly not, as there are already some very beneficial uses whose approaches include:
- only sharing the non-confidential and/or non-personal data on the ledger, with a link to locally-held sensitive information such as via a claim number; or
- having a very limited number of participants involved and limited scope such as just one contract, so it acts more as a distributed ledger than strictly blockchain. Here all participants are entitled to and trusted with the information and the whole ledger can be terminated if required, such as under Right to be Forgotten.
However these limitations do still restrict the ability of insurers to share enough intelligence to know if a known fraudster is back in play or a new one is making multiple claims, or they have developed a new bot or other trick such as deep fake video submissions. Yes, databases are increasingly large and utilising increasingly sophisticated AI to spot issues and trends, but participants still face that loss of control and limitation if they contribute their data.
BlockFrauds gets around these challenges in several easy steps:
- we use a private blockchain, where only trusted participants are permitted to join. This way we can all trust the intelligence that is being contributed, and trust the expert participants who are permitted to train the AI. Yes, many databases also restrict access to only trusted participants, so read on…;
- local processing of speech, text and image data anonymises that data irreversibly, so it can be shared compliantly. It has a claim reference tag to link it back to the locally held personal information, but if that information has been deleted under Right to be Forgotten – by the data controller - then it links to nothing: it has become “orphaned” and the person to whom it related can never be known;
- this anonymised intelligence is then held on a private node on this private blockchain, where the insurer / data controller has total control and can withdraw it at any time if they choose; and
- we use an AI called federated learning, which was specially designed to work across local data sources – such as the private nodes – without the data or anonymised intelligence itself having to be shared. The AI learns from that wider intelligence so all participants can be aware of relevant multiple frauds, known fraudsters and novel techniques essentially in real time, and prioritise their claim reviews accordingly.
This way the private blockchain and the AI mean that incredibly sensitive personal information, including someone’s actual voice and speech, their identity documents and other images, can effectively be shared, seamlessly and compliantly, whilst the data controller always retains control of the protected data.
The AI will always remember what it learned from the node, such as that “someone” with a particular voice or speech pattern was identified as a fraudster, but if the private node is withdrawn or the base data is deleted under Right to be Forgotten then that “someone” will never be specifically identifiable in future: the algorithm will just flag identical voice or speech contributions as a higher potential risk to be investigated.
So that’s why we use blockchain! It’s the enabling technology for insurers to collaborate in the fight against fraud even when this is rooted in customers’ personal data. And in this increasingly environmentally aware time, private blockchain also benefits from running on the more energy-efficient method of verification, being consensus algorithms rather than Proof of Work or Proof of State.
We’ve made it very easy to use and given even more control to participants so they can choose for themselves what level of risk they want to take, in a balance between customer service and fraud reduction. If you’d like to learn more, and see a demonstration, please do get in touch at info@fardoesoftware.co.uk!
Detect fraud better together Share the intelligence Keep control of your data
コメント