An analysis of polling data from the Aug. 15
referendum in Venezuela to recall President Hugo Chavez
indicates that certain forms of computer fraud were
unlikely to have occurred during the electronic voting
process, according to a study by three computer science
researchers.
Groups opposed to Chavez charged that statistical
anomalies in polling data indicated that election results
were fraudulent. However, an independent analysis of the
same data by Princeton professor Edward Felten, Johns
Hopkins professor Aviel D. Rubin and Adam Stubblefield, a
doctoral student at Hopkins, did not detect any statistical
irregularities that would indicate fraud.
"The opposition's claims ... seem to be incorrect,"
Felten said. "However, this does not rule out the
possibility that other types of fraud, which would not have
left statistical traces, may have occurred." The
researchers classified the study as a statistical analysis
and not a comprehensive investigation or audit of election
procedures and documents.
Rubin said, "The types of fraud that would be most
likely to be employed by a cheating government would not
leave the kinds of statistical evidence that opposition
groups have been charging. Simply changing some number of
'yes' votes to 'no' votes inside the machines would not
produce statistical anomalies but could change the outcome
of the election."
The researchers warned that electronic voting is
susceptible to fraud and that electronic voting systems are
generally more susceptible than less automated polling
techniques.
Felten's research focuses on computer and Internet
security and on technology and the law. Rubin's areas of
research are systems and networking security and computer
privacy.
The study and related information are available at
www.venezuela-referendum.com.