Rise of the machines: over 50% of fraud reports in UK are assessed solely by automated systems
During the financial year 2017/2018 there were a total 294,984 reports, of which 181,496 were assessed solely by an automated system.
The fact that novel technologies are invading various spheres of our personal and business life has been underlined by a statement by Rt Hon Ben Wallace MP, Minister of State for Security at the Home Office, who had to reply to a question about the proportion of reports to Action Fraud handled by computer systems.
Layla Moran MP has asked “how many and what proportion of reports made to Action Fraud have been reviewed by (a) a computer system and (b) a member of staff in each of the last three years for which figures are available”.
Rt Hon Ben Wallace MP stressed that no reports are dismissed by a computer system.
He explained that all reports submitted to Action Fraud are processed through an automated triage system that is designed to ensure that resources are targeted at those cases that have the most viable lines of enquiry. An automated process is required due to the large number of cases received (approximately 42,000 per month) to provide an efficient and effective response.
Crime reports to Action Fraud are assessed and triaged through the National Fraud Intelligence Bureau (NFIB) data analytics system which identifies links between seemingly unconnected fraud and cyber crime incidents from all over the country. The first phase of the crime assessment is automated using a computer based algorithm which scores the report’s suitability to be investigated against a number of criteria. The list of criteria includes the victim’s vulnerability, whether the identity of the offender is known, if there is evidence available to support this (e.g. a confirmed bank account money has been paid to), or the volume of reports made about a specific offender.
Reports with the highest viability scoring are then referred for further action by a crime reviewer. The crime reviewers will then undertake further work to identify whether there are sufficient lines of enquiry for the matter to be disseminated to law enforcement. Where opportunities for further action are identified, these crimes are referred to a local police force or other partner agency.
The data provided by Rt Hon Ben Wallace MP shows that over the past several years, at least 50% of such reports were assessed solely by automated systems.
- During the financial year 2017/2018 there were a total 294,984 reports. Of these reports 113,488 were both system and crime reviewer assessed, the remainder (181,496) were assessed solely by the automated system.
- During the financial year 2016/2017 there were a total 280,706 reports. Of these reports 128,564 were both system and crime reviewer assessed, the remainder (152,142) were assessed solely by the automated system.
- During the financial year 2015/2016 there were a total 234,201 reports. Of these reports 117,179 were both system and crime reviewer assessed, the remainder (117,022) were assessed solely by the automated system.
It is barely surprising that heavy volumes of data are managed by automated systems. Such solutions may also be very useful when tackling mundane tasks. Financial institutions have been using novel technologies to help them in complex processes concerning fraud detection. For instance, in March this year, Japan Exchange Regulation (JPX-R) and Tokyo Stock Exchange, Inc. (TSE) announced they would apply AI to market surveillance operations in order to detect misconduct such as market manipulation.