EC detects 300,000 duplicate voters,clean voter register in final stages

By the independent team

With just four months to the February 2011 elections, queries are being raised on multiple registrations in the national voters’ register. The Independent last week published a story in which a casual survey of the e-register showed staggering double registration in three constituencies. But the Electoral Commission (EC) and the German IT firm Muhlbauer High Tech International, contracted by the EC to work on the register, are defending the register arguing that no election in the world can achieve 100 percent perfection. The Independent spoke to Michael Stadtler, head of Application Development Cards and Security of Mulhlbauer and Pontius Namugera, Director of Technical Support Services at the EC about the issues raised in the article.

Michael Stadtler

Your innovation of an online voters’ register was a good one. But there are cases where double registration is evident with the same person registering twice or more. At other times there are cases of a person born in 1996 making it to the voter register; we thought Mauhbauer was hired to cure such cases.

The register is being arranged in two phases so it is still an ongoing process and I am quite sure that cases like this and other similar cases will be removed. But as of now we do not have software that automatically removes the case of underage voters.

Do you think that Muhlbauer was going to eventually develop Uganda software where such a problem would be dealt with, where it can reject a date if one was below the voting age?

Of course it just a matter of how the system is demystified and automated accordingly.

So at what phase is the project of automating the voters’ register?

It is at Phase two. Currently there is coordination of the security and information system in the overall project. At the moment what we are doing here is registration of voters for next year’s election which we consider as Phase One. Phase two involves capturing of finger prints and faces of all other Uganda citizens and to issue all Ugandan citizens 18 years and above a national ID card.

The problem of double registration in Uganda in the voters’ register has been teasing the Electoral Commission for last three elections; we believed that bringing in Muhlbauer is going to greatly help to solve this. What do your think is the level of success you have so far achieved?

I think it was a clear success. We have confirmed many duplicates already. Of course depending on the different database systems, the old voter register itself could be a potential source of duplicates in the new system. I think there was already explanation in your article for example that if you compared pictures from different times; pictures taken five to ten years ago as compared to pictures taken today, the challenge of facial recognition system is significantly higher compared to the challenge you would meet if you took pictures in a few days ago. In addition to that because the end face recognition system is a good mechanism to identify duplicates as they have identified slowly by face recognition about 150,000.

Can you just tell us a little bit about Muhlbauer and its technical competence, what experience do you have in managing election technologies around the world particularly in Africa?

Most of our government-related projects are non-disclosure agreements. So even if we have many projects in the IT- government business throughout the world, I am not the right person to speak about that.

But you do have experience around the world in doing such work?

Of course we have the experience in government projects that is the reason why we are in this project.

What particular challenges has Muhlbauer met in automating the voter register in Uganda?

I think for the time being there are not many challenges. For instance getting the equipment interface with one of the EC, training of the operators, power supply especially upcountry, were the challenges we had. In spite of these challenges we have been able to register more people than initially expected. The exercise was a big success because we ended up registering up to 5 million fresh voters which was over and above the 3.5 million target.

What would you attribute this success to, is it to the organizational skills of the electoral commission or the superiority of the Muhlbauer software during this registration?

I think the most important factor was the element of team work between the EC and Muhlbauer. Without working as a team in such a big project, in such a short time frame, you can have a superior system but you will not be able to deliver.

The 300,000 multiple registrations you said that have been identified, would you attribute this to the systems provided by Muhlbauer?

Again it was team work. So face recognition complication of the 150,000 people identified for example in the 10.5 million old voters the face recognition system is never ever automatically thorough in identifying duplicates. It just identifies potential duplicates which are displayed to operators in the main workstations at the EC. All this is on the side of the system and the other side of the operator.

In the article it was claimed that your IT software does not synchronize well with the one EC had in place.

There is no synchronization between the two systems at all. As a matter of fact, the old system being used whatever type it was, it is not being used anymore. But of course what we are using is the output of the old system. This output includes the old voters’ register that has 10.5 million voters. That was the basis for this project. We are using the old database information, some demographic data and images of the already registered voters to integrate it in our system. Such integration is a daily business for us in all projects we undertake around the world. The images in the old voters register were taken in a completely different time and taken under different conditions. This in effect implies a higher challenge in the facial recognition system compared to images from one source. But this is again not about integrating systems. It is about different sources of data. Quite naturally when you want face recognition system between old data and new data, the case at which the system does not identify duplicates is quite higher than it would in a more unified group of data.

I understand you will handle the national identification cards project. What assurance are you giving Ugandans that it will be all that fine?

First of all for the national ID, we have a little bit different situation. For national ID we talk about records including finger prints. The old data we have in the EC system of 10.5 million will not be used in the national ID project. So the national ID project will have two systems to identify duplicates. The main reason we continue using facial recognition is that in some cases some people cannot be able to provide fingerprints.

Pontius Namugera

You have so far identified 150,000 double registrations; does that mean you have cleared the voter register by 150,000?

The register has undergone several cleaning processes. The first one was the general update, where we had general applicants and transfers. As registration was going on, we were doing a back end process which was cleaning of the existing register, what we call legacy database of the 10.5 million voters and the new ones that we used the technology of facial recognition and finger prints. With that process we were able to get about 150,000 multiple registrations. We printed these and took them back during the display exercise for people and tribunals to confirm that actually this is a duplicate. Their work was to tell us which should be removed. That was the first phase.

After that we integrated the new data captured during the register update to the 10.5 million voters so that we prepare it for display. During display a number of activities take place, one is to report the dead, report non-citizens, those who do not belong to the area and report those who have migrated and also to report any multiple registration that exist on the system.

At the backend we continued doing a duplicate analysis and now we have run a further process and now we have identified an extra of 150,000 duplicates but this process is not concluded yet. We are going to conclude it before the presidential nominations. This makes 300,000 duplicate voters we have identified so far.

When should do we expect to have this clean register? I mean the online register does it include cases of double registration?

The online version of the register is as of display. The display register was uploaded during the display period and when the display was finished we embarked on the exercise of effecting the changes. When the exercise is completed and we declare this is the final register, with all cases of duplicates, deaths, cases of omissions catered for we will generate another version of the updated online voter register.

In your investigation trying to clean this register did you try to find out what caused this multiple registration?

Well, I think it is mainly on the voters. We have noticed a lot of laxity in Ugandans keeping records. The majority of them registered twice unknowingly. There are a few cases of intentional double registration which happens when people move around registering from one station to another. Even at times people come back on the same station and register for a second time.

Now with the help of Muhlbauer you have facial recognition and finger print biometric system. Can you explain to the public what this exactly means?

With Muhlbauer project there are two types of biometrics. Biometrics is the only way to uniquely identify a person because people change names, and date of birth so you can only identify an individual by using technology. The technology we are using in this exercise is facial recognition system (FRS), and automated finger print identification system (FS). We wanted to leverage the existing data that is the 10.5 million voters because we did not want to re-register and have a duplicate scene. The only biometric system we could use at the time was the facial recognition for the 10.5 million voters. Secondly we took up the FS which is the second biometrics Muhlbauer is using and as we went out to the field we capture both the photo and the finger print.

There have been alleged incompatibility cases regarding the software that the EC had and the one Muhlbauer is currently using. What is the truth?

Actually the straight answer is that there was no incompatibility. May be we can talk of the quality of the data like if they had to analyse the old data and the method we used to get it was by the help of stand alone cameras, the quality of the photo it produces may not be the same quality when you are using a computer. But there were no issues of incompatibility because all the data we got from the field were properly integrated with the existing data.

We live in era where cybercrime is becoming such a bigger issue in the world. Given that you have an online voter register, how safe is this e-register from hackers and manipulation?

The online voters’ register is very secure. Of course any system can be broken into but what you do is try to look at all loopholes that can rise such cases. We already thought about the security issues. We separated the two systems for the two registers, one for online and one for the central production register. The online voter’s register is a read only document and you can’t make any changes to it. Even if you made the changes, they will remain on that e-register as standalone register. They will not propagate to our production system. And as I said we are going to replace the e-register with the current updated register after cleaning.

So what would you single out as your biggest achievement in this project?

It is going to be our ability to have this voters’ register in time because when you look at the roadmap right now it is going to be completed on time.

Any other information that you would like the public to know

As I have said this process is ongoing. Out of the 8 double registration records you reported in your article, when we searched already 6 out of 8 had been dealt with.