The objective of this paper is to present a number of features that can be extracted from handwritten digits and used for author verification or identification of a person?s handwriting. The features under consideration are mainly computational features some of which cannot be easily evaluated by humans. On the other hand, these features can be extracted by computer algorithms with a high degree of accuracy. The eleven features used are described. All features were appropriately binarized so that binary feature vectors of constant lengths could be formed. These vectors were then used for author discrimination, using the Hamming distance measure. For this task a writer database consisting of 15 writers was created. Each writer was asked to write random strings of 0 to 9 at least 10 times. The results indicate that the combined features work well at discriminating writers and warrant further detailed investigation. Although the set of features was designed for dealing with handwritten digits (as may be written on cheques), it may also be used for isolated alphabetic characters.