<title>Self-authentication of value documents</title>

Abstract

To prevent fraud it is critical to distinguish an authentic document from a counterfeit or altered document. Most current technologies rely on difficult-to-print human detectable features which are added to a document to prevent illegal reproduction. Fraud detection is mostly accomplished by human observation and is based upon the examiner's knowledge, experience and time allotted for examination of a document. Another approach to increasing the security of a value document is to add a unique property to each document. Data about that property is then encoded on the document itself and finally secured using a public key based digital signature. In such a scheme, machine readability of authenticity is possible. This paper describes a patent-applied-for methodology using the unique property of magnetic ink printing, magnetic remanence, that provides for full self- authentication when used with a recordable magnetic stripe for storing a digital signature and other document data. Traditionally the authenticity of a document is determined by physical examination for color, background printing, paper texture, printing resolution, and ink characteristics. On an initial level, there may be numerous security features present on a value document but only a few can be detected and evaluated by the untrained individual. Because security features are normally not standardized except on currency, training tellers and cashiers to do extensive security evaluation is not practical, even though these people are often the only people who get a chance to closely examine the document in a payment system which is back-end automated. In the context of this paper, one should be thinking about value documents such as commercial and personal checks although the concepts presented here can easily be applied to travelers cheques, credit cards, event tickets, passports, driver's licenses, motor vehicle titles, and even currency. For a practical self-authentication system, the false alarms should be less than 1% on the first read pass. Causes of false alarms could be the lack of robustness of the taggant discrimination algorithm, excessive document skew as it is being read, or errors in reading the recordable stripe. The false alarm rate is readily tested by reading the magnetic tags and digitally signing documents in one reader and performing authentication in at least two other reading devices. When reading the same check in the same reader where signed, the error metric is typically in the range of 0.0600. When comparing different checks in different readers, the error metric generally reports values in the range of 0.3930. It is clear from tests to date that the taggant patterns are randomly different for checks even when printed serially one after another using the same printing process. Testing results to date on the robustness of the taggant comparison and discrimination algorithms indicate that it is probable that low false alarms and very low false accept rates will be achieved.