Firewalls are widely getting used for securing the private network. Firewalls check each incoming and outgoing packets and according the rules given by network administrator and it will take the decision whether to accept or discard the packet. As per the huge requirement of services on internet the rule set becomes large and takes more time to process one packet and it affects the throughput of firewall. So firewall optimization has a great demand to get good performance. Exiting research efforts developed techniques for either intra-firewall or inter-firewall optimization within a single administrative domain. In addition, existing techniques are inefficient in reducing packet processing delay, because they optimize firewall rules by only reducing the number of rules, but lack the intelligence to decide the order of rules. This paper proposes an adaptive cross-domain firewall policy optimization technique using statistical analysis, while protecting the policy confidentiality. To the best of our knowledge, we are the first to propose a technique that dynamically decides the order of rules based on the network statistics. The proposed technique not only identifies and removes redundant rules but also identifies the order of rules in the rule set to improve the performance of the system. The optimization process involves two tasks: First, collaboratively reduce the number of rules between multiple firewalls, while protecting confidentiality of them. Second, using network usage statistics, identify the order of rules in the rule set The feasibility of the proposed technique is shown with the help of the prototype implementation. The evaluation results show the effectiveness and efficiency of the proposed solution.