In recent years, messages and text posted on the Internet are used in\ncriminal investigations. Unfortunately, the authorship of many of them remains\nunknown. In some channels, the problem of establishing authorship may be even\nharder, since the length of digital texts is limited to a certain number of\ncharacters. In this work, we aim at identifying authors of tweet messages,\nwhich are limited to 280 characters. We evaluate popular features employed\ntraditionally in authorship attribution which capture properties of the writing\nstyle at different levels. We use for our experiments a self-captured database\nof 40 users, with 120 to 200 tweets per user. Results using this small set are\npromising, with the different features providing a classification accuracy\nbetween 92% and 98.5%. These results are competitive in comparison to existing\nstudies which employ short texts such as tweets or SMS.\n