Preventing fraud in African elections

Voters queue patiently during Kenya's recent presidential and parliamentary elections, held in March. Biometric verification systems put in place for the elections failed to work and many polling stations were forced to abandon them altogether. Image courtesty ILRI/Susan MacMillan, Flickr.

Godfrey Magila, 21-year-old software developer and CEO of Magilatech, did not always have a good relationship with technology. "I was somewhat destructive: my relatives could not leave me alone in the sitting room because they would return and find the TV dismantled. Everything that was electronic at home was hiding somewhere," he says.  Things seem to have improved since then as the self-taught computer programmer is now developing Tanzania’s first biometric voting system. 

Work on the system started when Godfrey was just 18 years old, at Dar es Salaam's first hackathon in 2011. After winning their category with an online voting application, Godfrey and his team were invited to develop their idea at the business incubator hub of the Tanzania Commission for Science and Technology. The system is now finished and undergoing testing. If Godfrey and his colleagues are able to successfully complete their pilot study by the end of the year, the system will be ready to be taken before the National Electoral Commission for consideration. 

Tanzania’s National Electoral Commission has already announced that 2015 will be the country’s first general election to use biometric voter registration. In taking this step, the country will join a long list of African nations that have turned to technology to make their elections fairer and more transparent. Biometric systems can help prevent fraud by ensuring that the people who turn up to vote are who they say they are. A de-duplication process also deletes from the biometric database anyone that appears more than once, preventing people from voting multiple times. 

Biometric systems can also prevent another kind of fraud: zombie voters. People have been known to register under the names of people who have died, so that they can vote twice. But using fingerprinting and facial recognition technology, as Godfrey’s system will, ensures this type of fraud cannot happen. Electronic fingerprinting extracts a series of features called minutiae points from the ridges and valleys of the skin surface. There can be up to 60 minutiae points on any one fingerprint, creating a unique pattern. Faces are more difficult to quantify in a meaningful way and, since changes of expressions, light and ageing all affect the accuracy of recognition, most registration systems only use facial features to identify voters alongside other biometric data. 

Once the data is collected, Godfrey and his team want to make fraud even more unlikely by introducing an electronic system for voting on election day itself. Although Tanzania has so far only announced biometric registration for 2015, an electronic voting system would allow results to be counted instantly and sent to control stations in each region. By significantly reducing the waiting time between counting and announcing the results, this would further reduce the opportunity for tampering. There is also an embedded encryption security system with four different algorithms to prevent anyone altering the results. Each time a security layer is hacked, it self-destructs, renewing itself in a different form from the backup. 

Preventing fraud and speeding up results are two ways in which biometric systems can help build confidence in African voting systems often plagued with accusations of malpractice. But technological solutions are not always enough to ensure the smooth running of an election. After the post-election violence in Kenya in 2007 left over 1,000 people dead, this year’s election made full use of technology to try to ensure trustworthy results. Despite Kenya’s 2013 election being largely peaceful, the biometric verification system failed to work and many polling stations were forced to abandon it altogether.

"There are two issues with the adoption of biometric technology for voter registration or any kind of civil registration. One is the technical issue of the biometrics and the other is the administrative issue of how to go about implementing it," says Anil Jain from the Biometrics Research Group at Michigan State University, US. No matter how well designed the technology may be, there are logistical challenges that are outside of the control of software developers. Fingerprint scanners can run out of batteries, networks can crash, and without enough training people can make mistakes. 

Training is especially crucial during the registration phase: weak fingerprints, for example, can be caused by sweaty hands, not enough pressure on the scanner and bruises on the fingers. If data has not been collected correctly, it can make biometrics hard to match. Demographic information can also only be verified by people; no amount of biometrics will be able to tell whether a person is a resident of the country or of an eligible age to vote. “There is no magic bullet with any of this,” says Jonathan Bhalla, research manager for the African Research Institute in London, UK. “If you’re going to implement such a system, you really need this to be planned years in advance, to have full-time election staff who can be trained in using this technology properly and efficiently,” he says. 

But Godfrey is convinced that the main problem with biometric systems is that the technology is not built with Africa in mind. “You cannot be doing an election in Tanzania and use a system adapted from a country in Europe,” he says. By developing and testing the system locally, Magilatech hopes to create a system that is tailored to their environment. All the hardware, for example, has been streamlined so that it consumes minimum electricity and can work on solar power. Some of these local features might even prove appealing for other countries. “You never know,” says Godfrey. “Maybe one day it could be that the Western world will acquire a system from Africa.”

Your rating: None Average: 4.6 (7 votes)


Post new comment

By submitting this form, you accept the Mollom privacy policy.