What did shock him, however, was the discovery that the thief had used his cheques and card to make ten £100 purchases from different shops over the next few weeks - and got away with feeble forgeries that looked nothing like Mr Hargreaves's signature. In one case, his name was not even spelt correctly.
Luckily, Mr Hargreaves's bank picked up the bill. But the incident shows how heavily banks and their customers depend on the dubious vigilance of shop assistants. Yet shops risk almost nothing by turning a blind eye to a doubtful signature; they risk losing a sale and an offended customer if they question it.
This could become a thing of the past, however, with the help of a new system developed by AEA Technology, formerly a part of the British Atomic Energy Authority. Countermatch, a software package designed by the company, is currently undergoing trials at Employment Service offices in Liverpool and Newcastle. Four thousand benefit claimants are being asked to sign on with an electronic pen connected to a digitising tablet and a computer. To reduce the culture shock, the pen contains ink - and the tablet has the appropriate form clipped on top of it.
The signature is then compared with the claimant's "signature template" on the database, formed from three sample signings at registration, and the system gives a view on whether it is acceptable. If not, they get a further two goes; if it still isn't, Employment Service staff do further identity checks on the claimant.
No claims are rejected solely on the results of the signature verification package - and they should not be, as the main bugbear of signature verification has historically been the false rejection rate - that is, the number of times the genuine customer is rejected.
AEA believes, however, that it has gone a long way towards solving the problem of false rejection.Its system monitors the signature dynamically - while it is being written - measuring the speed and timing of the strokes. From the initial three signatures, a model is built of the predicted variations that a signature can have.
In addition, data on the overall signature variation of the population is used to improve the model further. This "learning" ability, which is derived from neural network technology, is what makes AEA's system distinct from others.
So far, the results look promising: where other systems give a false reject rate of about 1 per cent, AEA's is 0.1 per cent. How many acceptances are false remains hard to tell, as people who have fooled the system in an Employment Service office are hardly likely to publicise the fact. But earlier in-house trials of the system suggested a figure of about 3 per cent. This performance is far superior to the humble clerk - Graham Hesketh of AEA puts human failure rates at around 20 per cent.
The Employment Service says client reactions are generally favourable, staff morale has improved, and they would like the equipment to remain in the trial offices. Some clients have been notable by their absence after the introduction of the system, so it acts as a deterrent, too.
And what about other uses for the technique? Credit-card checking would be the obvious choice, but concern about customer loss from electronic false rejections is still high, despite the £129.8m lost in credit-card fraud in 1993 - 75 per cent of which occurred at the point of sale.
If Countermatch succeeds in finding a solution to this problem, the technology could eventually make its way into the very core of our identities: soon even something so personal as one's signature may be demonstrably genuine only if it has been digitised.Reuse content