The proliferation of data through various platforms and applications is in constant increase. The versatility of data and its omnipresence makes it very hard to detect the trustworthiness and intention of the source. This is very evident in dynamic environments such as mobile applications. As a result, designing mobile applications that will monitor, control and block any type of malintents is important. This paper makes an attempt in this direction by implementing a lightweight machine learning classification scheme for hate speech detection in Albanian Language for mobile applications. Initial testing and evaluations indicate good classifier accuracy in mobile environments where frequent and realtime training of the algorithm is required.