Abstract-The need for secure communications has significantly increased with the explosive growth of the internet and mobile communications. The usage of text documents has doubled several times over the past years especially with mobile devices. In this paper, we propose a new steganography algorithm for Unicode language (Arabic). The algorithm employs some Arabic language characteristics which represent extension letters. Kashida letter is an optional property for any Arabic text and usually is not popularly used. Many algorithms tried to employ this property to hide data in Arabic text. In our method, we use this property to hide data and reduce the probability of suspicions. The proposed algorithm first introduces four scenarios to add Kashida letters. Then, random concepts are employed for selecting one of the four scenarios for each round. Message segmentation principles are also applied, enabling the sender to select more than one strategy for each block of message. At the other end, the recipient can recognize which algorithm was applied and can then decrypt then message content and aggregate it. Kashida variation algorithm can be extended to other similar Unicode languages to improve robustness and capacity.