This paper discusses a practical protocol for text document authentication, applicable to digital and printed form documents. It uses the text characters information to determine a key used to generate an authentication vector. Based on this vector, a feature in each character of the document is modified, without affecting the character "meaning." The modifiable feature may be size, color, shape, relative position, among others. If any character on the text is changed, the character information is different and consequently the authentication vector is also different. The proposed system does not require database retrieval and it is extremely difficult to forge the authentication process. A correlation-based detector for the system is proposed, and because feature detection errors may occur, an analysis is performed to determine the false alarm error probability of the system. Experiments illustrate the applicability of the method, considering the digital and the printed cases.
展开▼
