The growing use of Internet in everyday life has been creating new challenges and opportunities to use data mining techniques. A relatively new trend in the Internet is the deep web. As a large number of deep web data sources tend to provide similar data, an important problem is to perform offline analysis to understand the differences in data available from different sources. This paper introduces data mining methods to extract a high-level summary of the differences in data provided by different deep web data sources.We consider pattern of values with respect to the same entity and we formulate a new data mining problem, which we refer to as differential rule mining. We have developed an algorithm for mining such rules. Our method includes a pruning method to summarize the identified differential rules. For efficiency, a hash-table is used to accelerate the pruning process. We show the effectiveness, efficiency, and utility of our methods by analyzing data across four travel-related web-sites.