Rohan Jaitley: Central Gov't Standing Counsel for Justice
Cara monical 2011 symposium powerpoint
1. Voting Machines: The Good, The Bad, and The Ugly Cara Monical, Centre College ‘13
2. Brown Fellow Project Exploration: Last summer: 6 weeks in Spain where I first discovered my passion for computer security and gained a more international outlook This summer: 8 weeks in South Carolina doing research in analyzing voting machines Next summer (intended): internship with NSA or FBI Partnership: working with election officials to improve the system Academic/Research Leadership: wanted to gain research experience in a larger environment
3. Importance of Elections Results of elections have a major impact on the people, the country, and the world The government’s right to rule is based on the people’s faith in the electoral process
5. The Good The blind community is able to vote independently for the first time Voting machines often make running an election easier for the poll workers, who are often volunteers well past retirement age with very little training Tabulating results is almost instantaneous The public, in general, prefers voting machines to past mechanisms for voting
6. The Bad Current voting technology is fundamentally insecure and flawed No meaningful audit trail for verifying results Elections are so close that machine malfunctions could change the result of the election PEB iVotronic PDA
7. The Ugly The already bad system is made worse by preventable human errors Votes counted without data support Votes in the data support but not counted Incorrect timestamps
8. Analyzing the Current System Worked on iVotronics made by ES&S Produce an event log that records iVotronicID, PEB ID, PEB type, date, time, event code, and event message for each event Try to find ways to improve from election to election with the current system
10. Normal event log Time Event 07:29:44 0001510 Vote cast by voter 07:41:02 0001510 Vote cast by voter 07:47:28 0001510 Vote cast by voter 07:49:16 0001510 Vote cast by voter 07:54:34 0001510 Vote cast by voter 08:00:13 0001510 Vote cast by voter Partial event log for iVotronic 5122975, used in Lexington County, with all events from 7:35 am to 8 am on November 2, 2011
11. Problematic Event Log Time Event 07:31:56 0001510 Vote cast by voter 07:32:03 0002400 PEB access failed 07:32:03 0002400 PEB access failed 07:32:15 0002400 PEB access failed 07:32:15 0000706 Failed to retrieve EQC from PEB 07:32:15 0001635 Terminal shutdown - IPS exit 07:35:30 0001510 Vote cast by voter 07:39:44 0001510 Vote cast by voter 07:39:51 0002400 PEB access failed 07:39:51 0002400 PEB access failed 07:40:03 0002400 PEB access failed 07:40:03 0000706 Failed to retrieve EQC from PEB Partial event log for iVotronic 5101203, used in Lexington County, with all events from 7:35 am to 7:40 am on November 2, 2011
12. Types of Results Identified… Specific iVotronics and PEBs likely to have a problem Lapses in election procedure Differences between counties Differences between elections Found evidence of poor programming practices in the software behind the event logs Found some evidence of hardware failures
13. Hardware Problems Sumter, general election, iVotronic 5137877 11/02/2010 14:25:43 0002504 CF - file read error 11/02/2010 14:25:43 0002513 File - Read or write error 11/02/2010 14:59:12 0002207 TF - chip vs chip crc error 11/02/2010 14:25:46 0002184 UNKNOWN 11/02/2010 14:25:46 0002449 UNKNOWN 11/02/2010 14:25:12 0001204 Terminal vote image is corrupted … 11/02/2010 14:57:20 0002504 CF - file read error 11/02/2010 14:57:20 0002513 File - Read or write error 11/02/2010 19:17:27 0002207 TF - chip vs chip crc error 11/02/2010 19:17:27 0002207 TF - chip vs chip crc error ….. 11/03/2010 11:23:00 0001404 Warning - no valid term audit data
14. Programming Problems UNKNOWN event codes Prints the hardware error messages for hours, pages of events The event “vote cancelled– printer problem” appears despite the lack of printers “Terminal shutdown” in some instances could be sleep mode, in others– definitely not The standard PEB access fail sequence is illogical
15. Future Work I hope to perform similar analysis on the iVotronics used in some counties in Kentucky I also plan to analyze the data from the upcoming South Carolina Republican primary The completed analysis will be submitted for the USENIX conference and to a computer science journal
16. Acknowledgements Dr. Buell Dean Crepes and Carolyn Bledsoe of the Lexington County Election Commission Mrs. Laura Boccanfuso, Dr. Eastman, and Dr. Bowles National Science Foundation University of South Carolina The Brown Foundation and Centre College
17. Questions or More Information For more information about my research, my project, or the voting machine situation, please contact me at cara.monical@centre.edu
Editor's Notes
Abstract: I spent 8 weeks of this summer at the University of South Carolina working with Dr. Buell on analyzing the event logs of the iVotronics made by ES&S. Current voting machines leave a lot to be desired in terms of reliability and security. My research consisted of looking for patterns in the event logs and seeing what kind of information could be extracted. While this presentation contains an outline of the methodology and results of my research, it also focuses on the general voting machine problem. Some of my “talking” points for each slide are included in the Notes section of this PowerPoint. The current working abstract for the paper (which will be submitted for a conference and journal) follows: The right to vote is fundamental to the workings of a modern democracy and as such elections should ensure that each citizen's vote is heard once and only once. However, most elections today use a computer system, and most election workers are volunteers with little training, and so human errors are introduced into the process which in the past has led to both votes in the record not being counted and votes being counted which were not in the vote record. Previous research in the area of voting iVotronics has focused on the shortcomings of current voting systems or developing better systems, but this research seeks to work with the current voting system in order to make future elections run smoother. This research delivers a system for evaluating event logs from elections in 2010 of the iVotronic machines used in South Carolina in order to identify anomalies and isolate potential causes. Additionally, this research raises serious issues about the quality of the iVotronic software due to issues found in the event logs.
Research in voting machines perfect combination of computer science major and government minorPlan on a career in computer security research with a US agencyThis summer confirmed my goal of going to graduate school
Picture from 1968 Chicago riots, what happened the last time people lost faith in the electoral process
All reputable computer scientists not working for a voting machine company agree that all current DREs are fundamentally flawedIn NC, lost 4500 votes because machines ran out of memoryIn Ohio, 2004 Presidential election, Diebold president promised to deliver Ohio to Bush and then a voting machine records 4000 votes in a precinct with 400 votersIn SC, didn’t set up machines correctly, couldn’t vote until 1 pm, recording votes on toilet paperIn the 2004 election, would have changed the results with changing 1 vote in 1% of the precincts
Average age of poll worker (in South Carolina) is 75, complex computer system, long days, little training
ES&S controls about 70% of the marketEvents related to normal operation of the machine, poll workers entering menus, error type messages
Basic outline of the analysis- will explain each step