Association Rule Mining (ARM) in data mining provides quality association rules based on measures such as support and confidence. These rules are interpreted by domain experts for making well-informed decisions. However, there is an issue with ARM when the dataset is subjected to changes from time to time. Discovering rules by reinventing wheel, scanning entire dataset every time in other words, consumes more memory, processing power and time. This is still an open problem due to proliferation of different data structures being used for extracting frequent item sets. An algorithm is proposed for update of mined association rules when dataset changes occur. The proposed algorithm outperforms the traditional approach as it mines association rules incrementally and dynamically updates mined association rules.