Critics of DRE Machines, including both activists and computer security experts, claim that there is an increased risk of electoral fraud and if the security of the DRE software is compromised, election results could be tampered with in an undetectable fashion. This claim is countered by the position that review and testing procedures can detect fraudulent code, if such things are present, and that a thorough, verifiable chain of custody would prevent the insertion of such software.
Other critics charge that foreign hardware could be inserted into the machine, or between the user and the central mechanism of the machine itself, using a man in the middle attack technique, and thus even sealing DRE machines may not be sufficient protection. These claims are countered by the position that review and testing procedures can detect fraudulent code or hardware, if such things are present, and that a thorough, verifiable chain of custody would prevent the insertion of such hardware or software. Concerns like these have prompted the use of Voter Verified Paper Audit Trail. However these measures are often not or not properly taken. Another method to detect fraudulent voting machines are parallel test elections which are conducted on the election day with randomly picked machines. (The ACM published a study showing that, to change the outcome of the 2000 U.S. Presidential election, only 2 votes in each precinct would have needed to been changed.)
A workgroup of the National Institute of Standards and Technology (NIST) stated in a discussion draft, "Simply put, the DRE architecture’s inability to provide for independent audits of its electronic records makes it a poor choice for an environment in which detecting errors and fraud is important. The report does not represent the official position of NIST, and misinterpretations of the report has led NIST to explain that "Some statements in the report have been misinterpreted. The draft report includes statements from election officials, voting system vendors, computer scientists and other experts in the field about what is potentially possible in terms of attacks on DREs. However, these statements are not report conclusions.
A fundamental challenge with any voting machine is assuring the votes were recorded as cast and tabulated as recorded. Because there is no tangible ballot and the voter cannot check the computers memory by themselves, Non-document ballot voting system can have a greater burden of proof, and are even referred to as Black Box Voting machines by critics. This is often solved with an independent audit-able system that can also be used in recounts. These systems can include the ability for voters to verify how their votes were cast, or further to verify how their votes were tabulated.
Systems that allows the voter to prove how they voted have not been used in U.S. public elections, and are prohibited by most state constitutions. Voter intimidation and vote selling are the chief concerns that have led to prohibition of receipts.
Various technologies can be used to assure voters that their vote was cast correctly, detect possible fraud or malfunction, and to provide a means to audit the original machine. Some systems include technologies such as cryptography (visual or mathematical), paper (kept by the voter or only verified), audio verification, and dual recording systems (other than with paper).
Dr. Rebecca Mercuri, the creator of the voter verified paper audit trail (VVPAT) concept (as described in her Ph.D. dissertation in October 2000 on the basic voter verifiable ballot system), proposes to answer the audit-ability question by having the voting machine print a paper ballot or other paper facsimile that can be visually verified by the voter before being entered into a secure location. Subsequently, this is sometimes referred to as the "Mercuri method").
An audit system can be used in measured random recounts to detect possible malfunction or fraud. With the VVPAT method, the paper ballot is often treated as the official ballot of record. In this scenario, the ballot is primary and the electronic records are used only for an initial count. In any subsequent recounts or challenges the paper not electronic ballot would be used for tabulation. Whenever a paper record serves as the legal ballot, that system will be subject the same benefits and concerns of any paper ballot system.
To successfully audit any voting machine a strict chain of custody is required.
Additionally, with DRE voting systems there is no risk of exhausting the supply of paper ballots, and remove the need for printing of paper ballots, a significant cost. When administering elections in which ballots are offered in multiple languages (in some areas of the United States, public elections are required to by the National Voting Rights Act of 1965), DRE voting systems can be programmed to provide ballots in multiple languages on a single machine. For example, King County, Washington's demographics require them under U.S. federal election law to provide ballot access in Chinese. With any type of paper ballot, the county has to decide how many Chinese-language ballots to print, how many to make available at each polling place, etc. Any strategy that can assure that Chinese-language ballots will be available at all polling places is certain, at the very least, to result in a lot of wasted ballots.