Credit: Pixabay/CC0 Public Domain

About 70% of Americans voted in person in the 2024 presidential election, their ballots counted by machines called Precinct Count Optical Scanners (PCOS). Researchers at Towson University have systematically analyzed thousands of ways that PCOS machines could have process or security vulnerabilities—with the goal of helping local officials identify and mitigate those risks.

Their work was presented at the annual meeting of the Society for Risk Analysis, held Dec. 7–10 in Washington, D.C.

"In-person voting is more complex than mail voting, and that means there are more points in the process where things can go wrong," says Natalie M. Scala, Ph.D., professor in the College of Business and Economics and co-director of the Empowering Secure Elections Lab at Towson University.

How researchers mapped voting vulnerabilities

Her team built custom software to map out the "what-if" PCOS threat scenarios. Their Threat Modeling Analysis Tool visualizes and evaluates over 70,000 unique vulnerability pathways during the PCOS voting process—including the setup, voting, and teardown phases. (For example, a flowchart shows how an unauthorized change could occur within a local device.)

"Our model highlights each possible path in a scenario, so we can see weak spots and understand where extra protection would make the greatest impact," says Scala. "Seeing all those routes mapped out showed us how important it is to secure every link in the process—not just the machines themselves."

A sensitivity analysis allowed the researchers to see which security issues could have the biggest effect on the integrity of the election process. "This helps local election officials focus their time and resources on the fixes that matter most," says Scala.

Key findings on security vulnerabilities

The study showed that some of the most influential security issues are:

• Procedural or human-factor vulnerabilities. Ineffective poll-worker briefing and inconsistent interpretation of unclear markings repeatedly surfaced as top drivers of error likelihood. These represent training and communication gaps rather than technical flaws.

• Usability and interaction with voting equipment. The analysis showed that the way people interact with scanners matters. Mistimed feeds, misalignment , or confusion about prompts can all lead to delays or errors. These are simple, correctable issues that become important under Election Day pressure.

• Device handling and logistics. Unsecured or corrupt flash drives, unplugged or cut cables, and unsecured ports remained significant—no matter how the researchers shifted the relative importance of different factors. These emphasize the need for chain-of-custody checks and equipment protections.

The study showed that small procedural gaps can have large consequences. "It’s not always about software or code," Scala explains. "Sometimes it’s about missing paperwork or incomplete chain-of-custody documentation. Our work showed that tightening those everyday details greatly improves overall election security."

Next steps for practical application

The team is also developing an interactive version of the model that lets election officials explore "what-if" scenarios in their own voting processes—showing how even small procedural improvements can strengthen overall security.

They’ve also begun a new phase of research that looks at how total election risk changes when different mitigations are added into the process. That work will help identify which safeguards have the biggest impact, and where resources make the most difference.

"The long-term goal is to turn this research into something practical that election officials can use to visualize, measure, and manage risk before it becomes a problem," says Scala.

Provided by Society for Risk Analysis

Citation: Research aims to strengthen the security of in-person voting machines (2025, December 8) retrieved 8 December 2025 from https://phys.org/news/2025-12-aims-person-voting-machines.html

This document is subject to copyright. Apart from any fair dealing for the purpose of private study or research, no part may be reproduced without the written permission. The content is provided for information purposes only.