This document summarizes an algorithm for efficiently mining association rules from heterogeneous databases. The algorithm distributes the database across multiple sites while ensuring privacy. Each site locally mines frequent itemsets using an algorithm like Apriori. The sites then securely combine candidate itemsets and check rule confidence to find globally frequent rules meeting minimum support and confidence thresholds. The algorithm uses commutative encryption and a map-reduce model to parallelize the distributed mining efficiently while limiting data disclosure between sites.