" ITREALMS: Evaluating ICT systems and operations in INEC

pages

Sunday, March 01, 2015

Evaluating ICT systems and operations in INEC

Commentary@ITRealms:
Dr. Chris Nwannenna

I take more than a passing interest in biometrics, biometrics systems, and biometric records and, in one way or the other, try to appreciate the goings-on in INEC in this area. I have been doing this for some time now having been involved, as an external resource person, in the processing of voters’ records in INEC in the 2003 and 2007 elections.

Voter Registration Software

When the current INEC Chairman, Prof Jega, came on board he pointed out that the Voters’ register he inherited “lacked credibility both in real content and public perception and had to be discarded”. He therefore decided that an entirely new voters’ register “was the irreducible minimum for free, fair and credible elections”. The Commission did not experience any problems getting approvals as well as funding from government for the new voters’ registration exercise. “In its desire to reduce costs and to avoid dependence on Vendor–driven software, the Commission decided to undertake the registration exercise based on Open Source platform and technology.
Implementation of this process was coordinated by an ICT Consultant and a team of software developers who were able to undertake the challenging task of developing the software within two months” (INEC’s Report on the 2011 Elections).

Most people in the area of application software development would agree that delivering the registration software in two months was a feat. However the consequences of the time constraint manifested very clearly when the actual registration of voters commenced in 2011. The voter registration system was failing embarrassingly. The fingerprint scanner could hardly capture fingerprints. The processing systems got heated up easily and had to be shut down every now and then. At times the printers refused to print. . The system could not process one voter in any reasonable length of time. Queues built up and as people got frustrated the crowd naturally became unruly.

For the fingerprint scanner, several armchair experts emerged to proffer solutions: use methylated spirit to wipe the scanner or the fingers or both. In the ensuing confusion and embarrassment INEC passed the buck to the poor Youth Corpers who were doing the registration: “that they are not experienced”. How very unkind! The fact is that this scenario is characteristic of any software system that is not well tested before deployment. Sometimes it takes longer to test a solution than it takes to design and develop it.

However if the experience in 2011 could be attributed to inadequate time to properly test the system what about in 2014? In 2014, three clear years after 2011, the registration system was brought back hook, line and sinker for the Continuous Voters’ Registration (CVR) exercise. The whole symptoms that were prevalent in 2011 manifested themselves: fingerprint scanner failures, printer failures, excessive heat, etc. As usual queues built up. Rowdiness ensued. Prospective enrollees were given tally numbers with which they visited the enrollment center for days on end before they could be registered, if ever. Because of problems with the fingerprint scanners some people were registered without their fingerprints being captured. I witnessed this myself and INEC of course admitted this too when they announced that enrollees who could not see their Permanent Voter’s Cards (PVCs) must have been those registered without their fingerprints being captured. What an irony, people being disenfranchised for no fault of theirs but that of the inefficiency of the registration system.

So why was it not possible to fine tune the registration system before redeploying it for CVR in 2014? Why was it not possible to find out the problem with the fingerprint scanner? Was it faulty from the manufacturer or was it as a result of incompatibility between the scanner device driver and the operating system in use? The reason for this dereliction is simple: INEC takes Nigerians for granted. We do not deserve anything better so why should they strive for excellence.

But come to think about it. Why should an enrollee be registered without his fingerprint being captured? Fingerprint is the biometrics characteristic that INEC has adopted for identifying voters uniquely. Because of its importance therefore “fingerprint” should have been made a mandatory field during the design of the table schema of the biometrics registration software. The business rules for the application should specify very clearly that the fingerprint image of an enrollee must be captured first before his registration could succeed. The INEC registration system does not appear to have implemented this elementary requirement even though it is key to the success of the application. If this had been done the registration software would have had adequate user control to prevent voters being registered twice in the same registration station or Polling Unit. One could therefore not avoid the conclusion that INEC’s Voter registration software is not efficient enough to deliver as expected and therefore unsuitable for a project as important as voter registration.

SPOOFING

Consider another issue that should be of interest to many. In reference to their biometric registration for the 2011 elections, the Director of ICT in INEC said that “many Nigerians, in order to beat the system, used palm kernels to thumbprint”. The first question that one needs to ask immediately is why it was possible for people to successfully use palm kernels to thumbprint. The answer is very simple: INEC uses a very substandard fingerprint capture device that could easily be “spoofed” (to use the terminology in the biometrics industry). The capture device, in this case the fingerprint scanner in use by INEC, does not possess the electronic functionality that will ensure that the system will respond only if it detects that a “live” enrollee is indeed present at the point of data acquisition. It is this type of fingerprint scanner that INEC should use because scanners of this quality are the ones recommended by organisations, like FBI, for sensitive and important projects like voters’ registration. It is completely unacceptable that INEC uses fingerprint scanners that do not have ‘liveness detection’ mechanism. Why is INEC not using fingerprint scanners of this quality and specification? Is it because of lack of funds to purchase such scanners or because of lack of knowledge of the technical requirements of the registration system?

It cannot be as a result of lack of funds because INEC received very generous budget from the Federal government for the 2011 voters’ registration exercise (N72billion or thereabout). It cannot also be as a result of lack of technical knowledge because INEC has so many experts and consultants from UNDP, EU, AU, etc, as well as its own in-house staff. The use of non-standard fingerprint scanners by INEC is unacceptable and the reason for this practice lies somewhere inside INEC itself.

INEC also has to explain to us how people were able to use palm kernels to thumbprint. Who were responsible for the registration stations where these infractions were committed? Did INEC staff collude with people to do this or did they (INEC) staff commit the crime themselves? One would have wanted to know this.

Hard Disk Crash and Loss of Voters’ Records
I registered in 2011 in a Polling Unit (PU) in my neighborhood in Lagos. At the time of collection of Permanent Voters’ Cards in 2014 we were told that the records of all the registered voters in our PU were lost as a result of the crash of the computer’s hard disk. We were as a result advised to register again during the period of CVR exercise for our polling unit. The CVR exercise for our PU was conducted for some few days in November 2014. Subsequently INEC continued to announce one date after another for the collection of PVCs none of which materialized. At some stage INEC had announced that all PVCs would be ready by end of January 2015. However by 31st of January 2015 ours were not available. By the time the PVCs were available by the 5th of February I had left Lagos and INEC staff refused my wife from collecting it on my behalf. I was supposed to abandon my assignment and return to Lagos to collect it, in person, before the deadline of 8th February. Meanwhile the distribution started only on the 5th of February and elections were to have commenced on the 14th of February. In fact if the elections had not been postponed I would have been one of the over 23 million Nigerians that would have been callously disenfranchised. I cannot reconcile this with Prof Jega’s pronouncements in 2011. In his statement justifying the need for a new voters’ register he said that “The register of voters also excluded a sizeable number of eligible voters. If the Commission had gone ahead to use that register for the 2011 General Elections initially fixed for January, it would not only have disenfranchised a large chunk of voters, but also the credibility of the elections would also have been severely undermined”. What has really changed between 2011 and 14th February 2015? In 2011 it was not justifiable to disenfranchise “a large chunk of voters” but by 14 February 2015 it was justifiable to disenfranchise more than 23 million voters. Maybe 23 million voters, in his estimation, did not constitute “a large chunk of voters” and he was ready to hold elections without them and still beat his chest that he had conducted a free, fair and credible election. Not so, Prof, there is more to this than meets the eyes.

What about the issue of loss of voter records through crashing of computer hard disks? To my mind, and to that of many other people, INEC is supposed to be one of the most sensitive and important agencies in this country. How can an organization like that not back up its data in case of eventuality? Even the most elementary organisations back up their data. This is the standard practice everywhere in the world.

Another question that comes to mind is this: at what point did INEC discover that it has crashed hard disks? Was it just in November 2014- a few months to the rescheduled 14th February 2015 elections- given that the data were collected as far back as 2011? My wife who registered with me in the same unfortunate polling unit was out of the country during the Continuous Voter Registration period and therefore could not re-register. For her, and for many others in similar situations, the doors of this election are closed. Even though she still has her 2011 Temporary Voters’ Card she will not be able to vote not because she would not have wanted to but because of loss of records occasioned by lack of database maintenance culture in INEC.

“When we apply AFIS”

This expression is used very regularly by INEC staff. AFIS means Automatic Fingerprint Identification System and I believe that when they use the above expression they are referring to the process of eliminating duplicate records from their biometrics voters’ database. In biometrics this process is referred to technically as de-duplication.

De-duplication of biometrics records is a more advanced process than registration. When applied to the totality of the overall registered records in a project de-duplication ideally utilises more advanced hardware and software modules because the process involves comparing several hundreds, thousands or millions of records. The end result of de-duplication of a database of biometrics records is a reduction in the number of records when there are duplicates that have been discovered and have been removed. When there are no duplicates the total number of records before and after de-duplication remains the same. It stands to reason that under no circumstances will de-duplication give rise to a resulting figure that is higher than the original.

In 2011 INEC released total preliminary voter records of 67,764,334 which they said increased to 73,528,040 after de-duplication. Not a few people were shocked by this embarrassing inconsistency and the public outcry that followed was remarkable enough to attract attention. In the circumstances INEC had to offer a face-saving explanation: that their staff were under pressure to release figures and as a result had under-reported. Granted that this explanation could be true there must also have been a period of time when staff had relaxed sufficiently enough to come out with an accurate preliminary figure. But till today no such accurate preliminary figure has been given to sustain that explanation. In other words till today nobody knows what preliminary total voter records was de-duplicated to give the final figure of 73,528,040. In fact in the final report of the 2011general elections the figure 67,764,334 was completely and deliberately omitted to avoid being put to task on reporting inconsistent, questionable and inaccurate figures.

Card Readers and PVCs: The new Election System.

Many people believe that the issue of “security” was given so much weight by INEC in order to cover up its unpreparedness for the rescheduled 14th February Presidential elections. Several newspaper reports confirm this belief. During his interaction with Senators on the issue of Card Readers and PVCs on the 18th of February 2015 the INEC Chairman was reported to have “stunned many of the lawmakers when he revealed that One million Permanent Voter Cards had yet to be received by the commission from the manufacturers” (Punch Newspaper of 19 February 2015). The report of ThisDay newspaper of the same day quoted the INEC Chairman as having “assured the legislators that the commission would organize mock tests of the card readers in each of the six geopolitical zones of the country to ascertain their functionality ahead of the (rescheduled) elections”. He also told the senators that though the card readers were manufactured in China ‘they had been tested in Texas USA where their functionality and durability had been ascertained”. But at the same time, however, he admitted that out of the “182 000” card readers that he had bought
“503 have failed to work”.

Now the facts are there for everyone to judge using information from the horse’s mouth. As of 14the February 2015 delivery of one million PVCs was still outstanding. The PVCs had not, and still have not, been tested in a real life environment. And even though the functionality and durability of the card readers had been ascertained in Texas 503 of them failed to work, “dead on arrival”, he said. And yet with these glaring evidences of his unpreparedness for the election of 14th February Jega blamed everyone else, but himself, for postponing the elections.

Another fact worth mentioning is that the new election system being introduced by INEC includes a module for SMS (Short Messaging Service). Results are to be transmitted by SMS from Polling Units directly to, perhaps, INEC servers somewhere. People have been preoccupied by issues surrounding the workability, or otherwise, of the Card Readers. In fact this should be the lesser of the problems because if the card reader malfunctions it will be obvious to everyone watching. However news about hacking of SIM cards is more worrisome. It will also not be easy to know when SMS messages fail to arrive their destinations.

I have read comments by INEC staff on how their new system will reduce the waiting time to the announcement of results to 48 hours. It was also said that the system worked in Ghana and that there is no reason why it should not work in Nigeria. However that argument is not tenable. That the system worked in Ghana with a voter population of 14 million and 23000 polling units does not necessarily mean that it will work in Nigeria with a voter population of over 73 million and 180 000 polling units. An IT system generally degrades in performance when processing a large-sized database.

Additionally, and in fact more importantly, the system was only partially successful in Ghana even though Ghana’s Electoral Commission had earlier trained the operators and had also conducted a two-day test of the system before live deployment. According to an article in the Africa Report “the biometric machines brought forth significant problems with reliability. In two polling stations visited by the Africa Report, biometric machines had broken down and had held up voting for over an hour. In some cases, voters, having to wait in line for so long, considered skipping the vote and heading home”. As a result of this voting was, in fact, extended till the next day. There were also allegations that “fraud was taking place regarding the biometric identification cards (PVCs) of unsuspecting voters”. The result of the election was challenged in court.

Contrary to what an INEC staff would want us to believe the voting model that is about to be replicated here is that of Kenya not Ghana. Ghana’s model did not incorporate SMS module but Kenya’s own did. Kenyan voters were to identify themselves biometrically and to “ensure complete transparency, each returning officer would transmit the results, using handsets provided by the country’s biggest mobile phone network, directly to a giant screen at the tallying center in Nairobi”. This is similar to what INEC is proposing to do. There is also no doubt in my mind that UNDP which virtually runs the ICT department in INEC with Kenyan consultants would certainly recommend Kenya’s model.

So how did Kenya fare with this system during its election of 4 March 2013? This information is contained in an article in the SPECTATOR of 20th April 2013. “Biometric kits failed to recognize thumbs, forcing ID card (PVC) numbers to be typed laboriously by hand……
As for electronic transmission, exhausted returning officers forgot identification numbers needed to access the system or found that their figures would not transmit. The main server had crashed, overwhelmed by the relatively minor amount of data it was being asked to process. No backup had been catered for”. And the impressions and reactions that underlined this fiasco were revealing:” the electoral commission’s slowness in purchasing the equipment, testing it or training staff triggered accusations that the system was always intended to fail, allowing old fashioned rigging in by the backdoor”. To justify the fears and impressions of Kenyan’s on this form of electronic rigging the supreme court in confirming Uhuru Kenyata’s win indicated that it was a “cock-up” (of technology) not a “conspiracy”.

The method adopted for procuring and deploying the election system in Kenya- no prior training of personnel and no prior testing of card reader and SMS messaging services - is exactly what most Nigerians do not want. A fiasco similar to Kenya’s might have occurred on the 14th of February if the elections had not been postponed. INEC of course would not have been able to escape the damning accusation that it allowed “old fashioned rigging in by the backdoor” because INEC’s system was simply a dress rehearsal of the Kenyan one.

In conclusion INEC should be advised to avoid the temptation of deploying its new election system without adequate testing. Global best practices recommend this strongly especially for large-scale projects of the size that INEC is about to embark on. The consequences of deploying an IT solution without adequate testing are very evident in the Voter Registration system which performs very much below expectations. If I had my way I would recommend to INEC to run their new election system in parallel with the existing one to properly evaluate and confirm its performance before deploying it live, perhaps, in the 2019 election. Referring to the failure of the election systems in Ghana and Kenya, Jonathan Bhalla, from the African Research Institute think tank, stated very appropriately that “technology itself is rarely to blame; the issue is one of poor preparation and inappropriateness”. Every IT professional will agree with this.

*Dr. Chris Nwannenna is the Managing Director, Condata Systems Limited and former president of Nigeria Computer Society (NCS). He wrote this piece from Lagos.

ITREALMS ... delivering news for ICT4D
Short URLs: goo.gl, mcaf.ee, cli.gs

No comments:

Post a Comment