IT Expert Reveals Three Sources of Sirekap Issues

IT expert Marsudi Wahyu Kisworo testifying for the KPU at a presidential election results disputes hearing, Wednesday (4/3/2024). Photo by MKRI/Ifa.

JAKARTA (MKRI) — Another hearing on the 2024 presidential election results disputes (presidential PHPU) was held by the Constitutional Court (MK) on Wednesday, April 3, 2024. The joint hearing for cases No. 1/PHPU.PRES-XXII/2024 and No. 2/PHPU.PRES-XXII/2024 was to hear the testimonies of experts and witnesses for the General Elections Commission (KPU) as Respondent as well as for the Elections Supervisory Body (Bawaslu).

Also read:

Anies-Muhaimin File Petition on Dispute Over Presidential Election Results

Anies-Muhaimin Demand Revote in the Presidential Election Without Prabowo-Gibran

Anies-Muhaimin’s Experts and Witnesses Highlight Gibran’s Candidacy and Social Assistance

Four Onward Indonesia Ministers to Testify in Court

The KPU presented IT expert Marsudi Wahyu Kisworo as an expert and witnesses Yudistira Dwi Wardhana Asnar and Andre Putra Hermawan. Before the hearing chaired by Chief Justice Suhartoyo, Marsudi said that since the first time computer was used in election in 2004, digital vote counting system had always been a point of contention.

“It happened in 2019 and again today, while we all know that the laws and regulations stipulates that the valid vote [count] is the tiered one. This means that at the extreme, even if Sirekap did not exist, it would actually have no effect on the vote count,” he said.

The first Indonesian IT professor further explained that Sirekap is in both mobile and web forms. Data was inputted into the web Sirekap from the mobile Sirekap. The web Sirekap serves for consolidation, visualizing, and exporting data onto the web, which was then displayed on the web.

Marsudi said that there are three problems in the mobile Sirekap. First, it obtained data in the form of C1 result forms whose contents were handwritten, using a technology called optical character recognition (OCR). This OCR is a sign of development compared to the 2019 system Situng, where the votes had been entered manually.

Marsudi said that there might have been uproar as if the votes had been intentionally inflated. The handwritten C1 result forms were scanned, captured, then converted into numbers.

“There lies the first problem, since the C1 forms were handwritten and we know that people’s handwritings differ. There were 822 thousand polling stations with different people and, thus, different handwritings. At this one polling station, the handwriting could have been easy to read, but at some others it could have been a lot less legible,” he added.

There could even be different styles of writing numbers. Marsudi gave an illustration of the number 4 where it looked like an upside-down chair where the top part is open, or one where the top part is closed.

“The OCR’s accuracy was 99%, so there was a one percent chance of error. However, in practice, the accuracy could have been lower. The highest was 93%. So, there was 7% chance of error when the OCR converted images into numbers,” he stressed.

The second issue, Marsudi added, was relating to camera. The mobile Sirekap was installed on each of the polling station working committee’s (KPPS) members’ mobile phones, which could have different camera quality. Consequently, the quality of C1 forms generated differed. Some were clear, some dark, some white, some yellowish.

The third issue, Marsudi said, was the issue of paper. A folder paper could have resulted in misinterpretation of data by the OCR.

“This OCR is not a human that can predict things. It only complies with the training data. So, this AI system was given data on various kinds of handwriting. From those handwritings it learned to recognize numbers, 1, 2, 3, and so on. However, when the image quality was [low], it became a problem. These three problems are the source of the issues, which explains why when displayed on the web, the numbers could be different from those on the C1 result forms,” he added.

Nevertheless, Marsudi asserted, Sirekap is one of the means for transparency. Therefore, in the event of discrepancies or public complaints, the KPU then immediately took corrective action so there were fewer and fewer errors.

“The OCR technology is well-established but not perfect and nor is it 100% accurate. We cannot accuse the software of cheating. The solution in the future is that there must be verification before the results are published,” he emphasized.

Also read:

Ganjar-Mahfud File Petition on Dispute Over Presidential Election Results

Ganjar-Mahfud Allege Abuse of Power as Key Violation of 2024 Presidential Election

Ganjar-Mahfud’s Expert: Court Can Investigate TSM Violations, Highlight Ethical Breach in Gibran’s Candidacy

Ganjar-Mahfud Witnesses Reveal Vote Inflation on Sirekap

Publishing and Recapitulation Medium

Yudistira Dwi Wardhana Asnar, Sirekap security analyst, affirmed that Sirekap serves as both a publication medium and a recapitulation platform. The final result is the decision of the plenary recapitulation at each level. On the first day of the election, data from 29.07% of polling stations had been transmitted to Sirekap.

Yudistira explained the business process of Sirekap 2024, which includes the adoption of open-source authentication. Furthermore, he expounded on common vulnerabilities and exposures (CVE), a project designed to disseminate information regarding vulnerabilities in all software. He mentioned that CVE entries are only made public once the vendor has rectified the associated issues.

“CVE is only released when [vulnerabilities] have been disclosed,” he said.

Also read:

KPU and Prabowo-Gibran Refutes Allegation of Violations

The Nepotism Allegations Made by the Ganjar-Mahfud Irrelevant

Andre Putra Hermawan, head of Infrastructure and Information Technology Division at the KPU, emphasized that Sirekap serves as a tool to swiftly present election results from polling stations to the public. “Previously, Situng (counting information system) was used. And, for the first time in the 2024 election, the polling station working committees (KPPS) only needed to take photos. So, previously, the KPPS had had to manually copy documents for each witness and distribute them. However, with Sirekap, they can quickly take a photo and then provide the witnesses with a PDF file. This can be easily sent using a registered cellphone, eliminating the need for manual copying for each witness,” he explained.

Continuing his testimony, Andre explained that Sirekap serves as the initial data source for tiered recapitulation, commencing from the polling station (TPS) level, progressing to the subdistrict, regency, province, and finally to central (national) levels. Additionally, it functions as a tool to generate digital copies of the C results forms, which can be submitted to witnesses and polling station supervisors.

“In Situng, we utilized what we call C1 copies, where the large plane was transcribed onto smaller sheets and then transported to the regency/city using a scanner. However, with Sirekap, the KPU aims for greater transparency, intending to convey more comprehensive information about the conditions at the polling stations. That’s why there were guidelines on how to photograph the results on the large C result forms. The large plane forms were impractical to scan directly, hence the necessity to capture them in photos,” Andre explained. 

Author          : Utami Argawati
Editor           :
Nur R.
Translator     : Yuniar Widiastuti, Fuad Subhan (RA)

Disclaimer: The original version of the news is in Indonesian. In case of any differences between the English and the Indonesian versions, the Indonesian version will prevail.

Wednesday, April 03, 2024 | 15:38 WIB 52