A few studies have focused on the west Arabic (Maghreb) dialects for which resources are rare. To handle this problem, we devoped a web-based database of speech from Tunisian, Algerian and Moroccan speakers covering the diversity of Arabic dialects spoken in north Africa. Then speaker identification and verification experiments have been conducted in order to evaluate the performance of each dialect-based system. A baseline system using Timit database have also be developed for comparison purposes. The experiments show that the performances of dialectal speaker identification systems outperform those of the baseline identification system. The dialectal speaker verification systems performances are less better than the Timit-based sytem but are close to each other.