In wireless sensor networks, preserving location privacy under successive inference attacks is extremely critical. Although this problem is NP-complete in general cases, we propose a dynamic programming based algorithm and prove it is optimal in special cases where the correlation only exists between p immediate adjacent observations.