The upcoming presidential election in November may prove to be a referendum on the concept of e-voting more than a contest between diverse candidates. Can lessons be learned for our own IT implementations?

e-Voting Struggles in the Courts

Last week, on July 7, 2004, a federal judge upheld a decision made last April by the California Secretary of State to decertify touch-screen voting machines used in the upcoming election. Why did California decertify touch-screens? Because the systems did not include a verifiable paper audit trail similar to that provided by the traditional paper ballot systems.

The next day, on July 8, 2004, voters' rights groups in Florida sued Florida election administrators to overturn a rule that prohibits the manual recounting of ballots cast with touch-screen machines, a lawsuit with echoes of the state's disputed 2000 presidential election voting. Why doesn't Florida want to recount ballots? Because the Florida touch-screen voting machines cannot perform a recount; they do not produce a paper audit trail.

User Confidence Flagging

Clearly, with the contentious 2000 presidential election still in recent memory, citizens are concerned about the automation of a process that has not been adequately tested. Indeed, security experts around the United States have severely criticized most of the e-balloting systems for their lack of a verifiable audit trail, their proprietary software, their unproven security, and their lack of transparency in a critical process. (See "Weapons of Mass Election.")

But beyond the political and financial crises that these disputes have created for local and state municipalities, there are some underlying lessons that we in IT must learn if we're to be successful in every automation effort. Those lessons have to do with security and transparency. If we can learn anything from the current e-voting disputes, it's how not to apply a technical solution to a critical information system.

Security: User Perception Is Key

As an information system, the practice of voting has some unique qualities that make the design and implementation of a technical solution somewhat complex. Chief among these requirements is the anonymity of the voter: Voters must be guaranteed that their choices will not be made known to others. Of course, this policy is meant to ensure that the voter is not dealt recriminations and to prevent the possibility of brokering votes in an election.

If one looks at this problem from a systems perspective, however, it's not much different than an anonymous security profile in a public forum. In such a scheme, registered users can post anonymously on a forum, hiding their identities from others in the forum. Only the system administrator could theoretically track down the actual identity of the user.

However, security systems for the crop of e-voting machines that are currently being implemented are not known. Their designs are held as proprietary by the manufacturers of the machines, and the companies have not allowed security experts to inspect their workings. Consequently, no one except the manufacturers of the various e-voting machines will testify to their applicability for maintaining anonymous votes. Unfortunately, part of the source code for the e-voting security scheme of one manufacturer was recently illicitly posted to a public Web site, raising questions about the general level of security within the manufacturer's organization itself.

Clearly, such leaks would tend to diminish the voters' sense of security in the information system.

Is there a lesson for IT in this debacle? Treat security as your number one priority! If users don't trust it, you will need to go to great lengths to reestablish your organization's credibility. Without that trust--and an adequate means of proving its applicability--you've lost your system's credibility before you've even implemented a single line of code. Security must be based upon publicly acknowledged standards and--in critical applications--exceed those standards by a significant degree.

Transparent Verification Is Critical to User Acceptance

The second most important criteria that a critical system must provide is in the validation of the system itself. A key part of validation is in testing the transparency of the system's data and the verification of a user's actions to that data.

Certainly, it's in this area that the current e-voting machines are struggling within some states. For instance, the decision by the Secretary of State of California to decertify certain touch-screen machines for voting was a direct result of their inability to verify the actions of voters with an audit trail. And in Florida, the lawsuit against the election committee is based upon the fact that no verifiable audit trail can be provided by the machines in question.

Of course, we've all experienced problems in this area. In the critical systems that we implement within IT, we often mistakenly presume that every action initiated by the user is accurately recorded or translated by the processes we've programmed. Yet each one of us has been chagrined to discover that bugs and glitches do occur. That's why, on critical systems, we're afforded the time to not only verify the results of data translations or systems migrations, but also to involve the users in validation testing to enlist their confidence.

Yet, in an ironic twist, Congress has afforded more legal requirements to business process validation--through the Sarbanes-Oxley legislation--than to the current electoral automation process that they funded. Instead, state and municipalities were provided a mandate to automate the voting process but not the necessary funds to validate that machines they might purchase actually produced accurate voting results. And since there were no verifiable public standards, manufacturers were free to make whatever claims they wished. It's those claims of accuracy that are now being questioned, and as a result, some systems are now being disqualified.

e-Voting as a Management Primer on Automation

If there is a single lesson that we can learn from the e-voting debates, it's that new technology alone cannot revolutionize information systems that already have embedded traditions, requirements, and standards. Implementation costs--including testing the security and validating the system--will inherently eat into the budget and potentially dwarf the initial cost of the technology itself.

Yet if appropriately implemented as an element of a total information system, new technology can significantly speed the processes and deliver a real return on investment over time. E-voting is one such element, but it is not yet a proven component, and it is certainly not the entire voting information system itself.

Congress envisioned e-voting in 2004 as a panacea for the malfunctions of the past presidential election. Unfortunately, because of poor implementation practices and poor management oversight, how that promise will be delivered is, unfortunately, still open to doubt.

Thomas M. Stockwell is Editor in Chief of MC Press Online, LP.