Automating Signature Verification on Absentee Ballots and Mail-In Votes

Mail-in-voting is proving to be an increasingly popular method of casting votes that has put down its deepest roots on the West Coast of the United States, especially in Oregon, Washington and California. These states, among others, allow some or all elections to be conducted entirely by mail. Other states, such as Colorado are experiencing a rapid growth in voting by mail. Nearly half of the state's registered voters have requested ballots by mail in the 2008 presidential election. Oregon became the first state in the nation to conduct a presidential election entirely by mail in 2004. The state maintains a vigorous signature verification process for qualifying mail-in ballots and claims these procedures are key to its success.

Since the most common form of identification used in ballots and at the polls is a signature, many states and municipal governments are making a conscious effort to ensure that their processes include signature verification. However, there is a lack of uniformity in signature verification across all states. Manual signature verification procedures, still used in many, result in variable success rates of mail-in ballot fraud detection.

The weakness of manual signature verification processes lies in the fact that election workers, through no fault of their own, are not handwriting experts and cannot provide entirely reliable fraud detection capabilities even after having completed signature verification courses.

Still, manual detection was the only possible answer for many counties and states in early years of this century, since automated signature verification technology did not offer an industrially mature solution on par with visual signature verification. Detecting skilled forgeries was especially challenging as none of the existing technologies could present anything close to visual signature verification results. Today signature verification technology can detect random and skilled forgeries with an accuracy that not only equals - but far surpasses - visual signature verification.

In applications that deal with signed paper documents such as absentee ballots, only the document's static, two-dimensional image is available for signature verification. This poses a challenge, considering that detection has to address not only random forgeries that were produced without knowing the shape of the original signature, but also skilled forgeries, or those generated by people imitating the original signature. Therefore, viable automatic signature verification product must have the ability to detect all the different types of signature forgery by means of sophisticated algorithms. Parascript® SignatureXpert® is a leading product for signature verification that successfully handles this task.

Sometimes there are more deviations in genuine signatures than in skilled forgery, which can be created with a rather sophisticated technique, or be the result of thorough copying of every minor element. This is the reason that signature verification is based on the presumption that it is virtually impossible that two authentic signatures are exactly alike in terms of style, slant, and spacing.

Examples of inconsistent authentic signatures:

Signature Verification

In order to account for the missing important biometric data required to product highly accurate signature comparison results (such as speed, acceleration, deceleration, pen pressure, and others). Parascript's signature verification system imitates the methods and approaches used by human forensic document examiners. Parascript's automatic signature verification is executed by a powerful combination of verifiers using seven fundamentally different approaches. In particular, they combine a human-like holistic analysis of a signature and signature segmentation with a subsequent analysis of the signature elements. The whole signature verification process can be described as 'the work' of a group of highly skilled experts. Each of them focuses on the analysis of a certain range of informative signature features and their areas of expertise complement each other, resulting in improved overall performance. This signature verification approach allows unprecedented accuracy of verification by taking into consideration all informative data that can be extracted from a signature image, including biometric characteristics restored from the still image. During signature verification, signatures in question are compared with those on file to determine natural deviations in individual styles that the human eye often cannot detect. Parascript's signature verification software compares the image of a signature presented for verification against a reference signature image - a genuine signature previously collected, for example, from voters when they registered for voting-by-mail - and makes a conclusion about the authenticity of the input signature. Multiple reference signatures can be used for signature verification and many include signatures collected from voter registration cards, signature cards, driver licenses or signature snippets cut from any other document.

The major problem facing signature verification technology is differentiating between the consistent parts and unstable parts of the signature that vary with each signing. Therefore, the automated signature verification process has to have a certain tolerance, but it must also be sensitive enough to pick out a needle in a haystack when there is, in fact, distinctive evidence of forgery.



		HOME \| CONTACT US \| CAREERS \| CUSTOMER LOGIN \| SEARCH \| SITE MAP