Page 39 - GCN, October/November 2018

the electronic filing requirements that have applied to all other committees since Jan. 1, 2000,” an FEC official said. “So the vast majority of the reports that were handled through the paper automation project now will be filed electronically.”
PROJECT: Risk-Limiting Audits of Election Results
Colorado Department of State
Inspiring trust in
elections
The buzz around election security often focuses on front-end vulnerabilities, such as voting machines, state registration websites and online disinformation campaigns. However, one state has taken the lead on implementing a critical form of backend vote verification that can alert officials if their election has been hacked.
Colorado is the first — and so far the only — state to legislatively mandate
and implement risk-limiting audits of its elections. Security experts consider such audits to be the gold standard for ensuring accurate election results. Rhode Island is poised to become the second state to adopt the approach.
Colorado relies on open-source software called ColoradoRLA that compares a random sampling of a precinct’s paper ballots with their corresponding digital votes. If it discovers enough discrepancies, the software flags the ballots for a larger manual count.
Anyone can download ColoradoRLA
for free and deconstruct its code. Dwight Shellman, county support manager at the state’s Elections Division, said officials wanted to give technology-minded citizens and organizations the ability to validate the state’s results for themselves.
“The whole purpose of a risk-limiting audit is to obtain a statistical level of confidence that the outcome of an election is correct,” Shellman said. “We needed software to do it, but it’s hard to go back to the public and say, ‘Trust us. This software we developed shows that we’re correct.’
BEST IN CLASS - DEFENSE
PROJECT: Unstructured Data on Machinery Repair for Navy Ships Military Sealift Command, U.S. Navy
Preventive maintenance through predictive analytics
When the Navy’s Military Sealift Command realized that its decades-long horde of unstructured maintenance data was hampering strategic decision- making, officials turned to machine learning for help.
The command teamed up with Abeyon, a firm that specializes in artificial intelligence solutions, to create the data analysis tool Clarifi — a preventive capability that monitors the condition and reliability of all 100 of the command’s ships.
Using sample documents representing the larger unstructured dataset, the team built a machine learning-based text analysis model to explore and identify relationships among the equipment data and entities.
The tool’s pilot version turned nearly 30 years’ worth of data locked in Word documents into educated decisions regarding ships and their maintenance. Overall, the tool has increased operational efficiency and helped lower costs by millions of dollars.
As the technology matures, Clarifi could offer recommendations on equipment health, condition and potential for failure. Instead of guessing whether a machine has reached its end, the command can now look to the past to predict problems and preempt them.
GCN OCTOBER/NOVEMBER 2018 • GCN.COM 35