A simple way to cheat with PCOS machines

MANILA, Philippines - The spate of gross PCOS errors during the final field testing with a 10-ballot test set is actually a problem of misaligned ovals, although Smartmatic describes it as misconfiguration of the memory cards. Cheats can now exploit this problem to inject malicious files into the PCOS machine, without modifying the source code.

Before I explain how, let me go through the PCOS problem again.

According to Smartmatic, the problem was a mismatch between the PCOS configuration and the ballot layout for the local candidates. For some reason, they changed the ballot layout for the local candidates from single-spaced to double-spaced. However, they forgot to change the PCOS configuration to account for the new locations of the double-spaced ovals.

Thus, from the perspective of the PCOS, the ovals were misaligned. Ballot printing had changed the alignment of ovals due to the modified layout, but the PCOS was still looking for the ovals based on their original alignment. Thus the PCOS couldn’t find most of the ovals anymore.

However, the misalignment this caused is worse than the one-to-two millimeter creeping misalignment that happened with the ultraviolet security marks and possibly the ovals themselves. This misalignment was by a whole line. Instead of just reducing the accuracy of the PCOS machine, it simply caused the machine to miss many of the ovals.

So, you can describe the problem as either a misalignment of ovals, or a misconfiguration of the PCOS machine. Since they could not change the printed ballots anymore, the Smartmatic solution was to adjust the PCOS configuration.

While I have not seen the actual configuration file of the SAES 1800 PCOS, it would most probably involve a list of every oval on each side of the ballot, and the horizontal-vertical coordinates of each of these ovals, which is how other optical scanners do it. Smartmatic simply needs to adjust the coordinates of the ovals for the local candidates, to conform with the modified ballot layout, and voila, the misalignment is gone. I have grossly simplified of course but this is how the correction would essentially be done. Thus when the PCOS machine is started, it would read the corrected configuration file and now find the ovals exactly where they have been printed. I agree that, properly done, this would solve this particular problem.

But now, we’ve just discovered a way to do targetted cheating — mostly through vote-shaving — with the PCOS machine!

All that needs to be done is to slightly change the coordinates of the oval associated with a targetted candidate, so that the printed oval on the ballot associated with that candidate is slightly misaligned with respect to the coordinates on the configuration file. The larger the misalignment, the greater the possibility of missing the mark on the oval — this is called a false negative (or a “bawas”). Of course, the cheat would not make such a big change that the PCOS machine will miss the mark entirely — a consistently zero count would be too obvious and easily detected. But a slightly misaligned oval might be missed occasionally, say one in every ten marks, leading to a 10% shaving of votes for that particular targetted candidate.

And because configuration files are usually not considered part of the source code, this cheating method, which merely involves a minor change in one configuration file, will not change the source code. It will therefore escape detection through the usual method of checking for the program hash codes.

But it can be detected if the hash codes of configuration files are also double-checked. However, I have not seen any Smartmatic or Comelec statement claiming that they do so.

That this way of cheating can happen doesn’t mean it will happen. Murphy’s Law, however, says that if it can happen, then sooner or later it will, probably sooner than later. We’ve already seen this with the PCOS malfunctions that happened a few days before the elections. Those were so obvious they were easy to detect. This method of cheating is more subtle, and can be easily missed. The COMELEC must make sure it doesn’t happen.

Automating elections, it seems, will not eliminate cheating at all, but simply spawn new ways of doing so.

(Roberto Verzola has a background in engineering and economics and a passion for social and environmental issues. He is recognized by the IT industry as an Internet pioneer in the Philippines and works with NGOs on technical issues. He currently lectures at the Institute of Mathematics of the University of the Philippines.)