Thread Links Date Links
Thread Prev Thread Next Thread Index Date Prev Date Next Date Index

RE: [Fwd: RE: MTBF for voting machines] part 2



Title: Message
The current test defined in the FEC 2002 standards for precinct tabulators that concurrently covers reliability and environmental requirements is run for 163 consecutive hours.  Each hour 100 ballots are voted/run for a precinct based unit (16,300 total ballots).  These 100 ballots are to exercise "all possible voting locations".  Temperature and line voltage are varied in a prescribed pattern every 4 hours for a 24 hour period (minimum and maximum operating temperature each at nominal, minimum and maximum line voltage).  This pattern is repeated in the 2nd 24 hour period.  The remainer of the test is run at nominal temperature and line voltage.
 
According to the standard, the ballot content that is contained must aggregate to 1,549,702 positions processed.  "If the system makes 1 error before counting 26,997 ballot positions correctly, it will be rejected."  "If the system reads at least 1,549,702 positions correctly, it will be accepted."   It also prescibes that if 1 error occurs between the these limits "testing will have to be continued until another 1,549,702 positions are counted without error (a total of 3,126,404 with one error)." 
 
 
-----Original Message-----
From: SAVIOCvs@AOL.COM [mailto:SAVIOCvs@AOL.COM]
Sent: Thursday, February 17, 2005 10:02 PM
To: sklein@CPCUG.ORG; stds-1583-disc@IEEE.ORG
Subject: Re: [Fwd: RE: MTBF for voting machines] part 2

In a message dated 2/15/2005 2:36:45 PM Eastern Standard Time, sklein@CPCUG.ORG writes:
>
> Herb -
>
> Here is a possible approach to think about:
>
> 1.  The number of cast vote records unrecoverably lost or altered as a
> result of machine malfunction shall be less than 1 in N ballots cast,
> where N is something like a million.
>
> (Note this recognizes that inaccuracy is not the same as malfunction,
> which is defined as in the current spec.  ....)
>
I'm not sure what the note in parentheses really means; however, here are my thoughts on the matter:
 
1.  Loss or alteration of votes in any non-random manner is a MUCH more serious matter than random loss or alteration of votes.
 
2.  Loss of 300 votes in one machine is much more serious than loss of one vote in each of 300 machines. 
 
3.  ANY detection of errors in "normal operation" (unrelated to component failures) should be clear grounds for rejection of a system.  (But is drift in touch-screen adjustment considered a component failure?  There seems to be a tolerance for that type of error.)
 
4.  In my experience, there is a significant amount of arbitrariness or subjectivity in estimating the probability of very unlikely events.  Does anyone really believe that a practical procedure exists for distinguishing between an expected loss of one vote in 500,000 and one vote in 2,000,000?  What level of confidence is required?  
 
5.  Calculating an expected number of lost or altered votes "as a result of machine malfunction" requires estimating the probability of each of many types of malfunctions as a function of usage (what about aging effects?) along with the expected number of votes lost or altered for each type of malfunction.  Does this seem practical?  I know that NASA did (and maybe still does) this kind of analysis for complex space systems, but it's very expensive -- and yields only a general idea of the probability of any failure.  As far as I know, NONE of the dramatic failures of space equipment was identified as high-risk by such studies.
 
6.  A-priori calculations of MTBF are, in my opinion, very unreliable.  Statistical evaluations of MTBF can be very accurate, if enough failures actually are observed.  If large MTBF values are required, that would imply very long (and expensive) tests. 
 
 
Most of the above is negative, yet some type of practical reliability/accuracy standard is required.  Let me suggest something along the following lines:
 
1.  Each of (12?) significantly different ballots is programmed for the voting system.  For each ballot, an independent automatic system simulates the actions of (1000?) voters completing their ballots in (10?) different ways, in random sequence.  Automatic testing is done on each ballot (8?) times, simulating a total of (96,000?) voters. 
2.  If all the vote totals from the (96?) voting system runs match what the independent automatic testing system generated, the voting system passes this accuracy/reliability criterion.
3.  If there is any type of failure during the above testing, then:
    a.  if the failure caused no alteration of recorded votes, and no loss of votes other than the last one cast before the failure, and
    b.  if a repaired or replaced system finishes (2?) more complete tests (steps 1 & 2) successfully, it passes;
    c.  otherwise, the system must be modified to reduce failures before retesting from the start. 
 
I'm strongly tempted to DOUBLE the testing requirements each time a system is forced into modification via step 3c.
 
The above obviously is much less stringent than the one-in-a-million requirement originally proposed, but I argue that "proving" one-in-a-million is either impractically expensive or largely subjective.  This suggested criterion attempts to address software errors, hardware failures and any problem that affects many votes at once.
 
I don't know how to enforce it, or who would collect the data, but rare problems should be documented wherever and whenever the systems are used.  There should be a clear mechanism to de-certify a system.