Sunday 3 June 2018

The biometric voter registration fallacy

If you listen to the discourse coming out of Zimbabwe BVR (biometric voter registration) is a magic bullet that is going to prevent all rigging. My countrymates, not so fast. If anything the use of biometrics is going to lead to more doubt and disputes than usual.

What is biometrics.

Biometrics is the science of analysing human traits using digital techniques. The most commonly analyzed traits are fingerprints, although facial and iris recognition can also be performed.

Biometric involves capturing and digitising an image of the desired feature and than using computer procesing to store the feature or compare it to an existing database.

Compter techniques used include pattern matching and neural networks. Neural networks belong to the area of computing called artificial intelligence.

Besides biometrics neural networks are also used in other areas of image processing such as automatic number plate recognition.

In my career I have worked with both number plate recognition and biometric technology. I developed an average speed calculation system based on automatic number plate recognition. I did not say average speed enforcement, I said calculation so if you got a ticket please don't blame me.

I have worked with biometrics in identity verification and access control. That is systems that open doors and gates if you are recognised.

The first point I would like to stress is that pattern matching whether neural networrk based or otherwise, is not exact science. It is a statistical endeavour where you come up with a probable match within a certain degree of confidence. The confidence level is never ever 100%. Top of the range systems usually have a confidence level of about 95%. That is there is a 95% probability that whatever result it comes up with is correct.

If you have ever been enroled on a fingerprint system you will notice that they will always ask you to go and test your finger immediately. Every 20 or so people the enrolment will fail and they will take another sample.

The 5% probability of being wrong is not a small number. Out of 5 million voters that is 250 thousand people. Or 6 constiuencies.

Mind you that is based on ideal conditions. That is you have high quality equipment, operating in conditions that are not harsh with consistent samples.

In real life biometrics is even less accurate. In rural African conditions biometrics is even worse.

First there is usually no mains electricity. Batteries discharge and as the voltage get's lower sensors loose their accuracy. Often the roads are rough dust roads wich means equipment is shaken and contaminated with dust increasing the probability of failure.

Human features also change for various reasons. A snall cut on your finger changes your fingerprint profile. If you touch hot surfaces it may also deform your fingerprints. I have heard of people rubbing their fingers on hot irons or horplate stoves before going to the department of home affairs to apply for documents. Apparently it is a reliable way of making sure the system doesn't recognise your fingerprints if you have an unsavoury previous record.

What this boils down is that a high number of people are likely to experience problems with fingerprint matching even if they successfully registered.

If those people are allowed to vote, somebody will cry foul. If they are barred, somebody will cry foul as well. In the Zimbabwe context this means the stage is already set for the elections to be hotly disputed. Mark my words.

By far the best way of ensuring all elligible people vote and preventing double-voting is having an accurate citizen registry and using indellible ink. Of which Zimbabwe has one of the highest citizen registration coverages in Africa. Almost everyone in Zimbabwe has got an ID.

Lastly let us talk about who uses BVR in in the world. None of the technologically advanced countries who manufacture the equipment. The European countries, Russia, Russia, China - all these do not use BVR.

Only about 50 countries in the world have adopted BVR and nearly all of them are in Africa and a few other poor Asian and Latin American countries.

The reason for that is that most poor countries are run by politicians who know little about technology and thus can be easily sold duds. An example is buying snow-ploughs while being told they are graders.

No comments:

Post a Comment