Word-based writer identification and authentication have been under investigation for years. In this paper, we present a new Arabic online/offline handwriting dataset for writer authentication and identification. The created dataset includes two parts: AHWDB1 and AHWDB2 which are made freely available for the research community. Each part of the dataset consists of 2000 (10 trials X 200 writers) samples captured using HUAWEI MediaPad M3. Dynamic information such as pressure, timestamp and the coordinates(X, Y) have been collected and involved for both parts of the dataset. In addition, age, gender and education level have been added to the dataset for future investigation. Several experiments are conducted on the dataset. The preliminary identification results are obtained using K-Nearest Neighbor (KNN) with Dynamic time warping (DTW) and support vector machines (SVM). We recorded 81.35% identification rate as the best result using SVM classifier on AHWDB2.