The parameterized complexity and kernelization of resilience for database queries

Given a database instance and a query on it whose result is initially non-empty, the resilience decision problem is to decide if there exist a small enough number of facts in the database instance such that the deletion of these facts empties the result of the given query. In this paper, we revisit the resilience decision problem. We investigate the parameterized complexity for various classes of database queries. We consider the factors including the query size and the number of variables, and present several intractable cases even from the perspective of parameterized complexity. Meanwhile, we refine the characteristics of resilience for self-join-free conjunctive queries containing triads, and show that it is still NP-hard even if the structure of the input database instance is simple. This result implies the hardness essentially comes from the parity of triangle sequence instead of the complicate (non-planar) intersections of cycles. On the other hand, we also obtain some positive results showing that the resilience decision problem is still fixed parameter tractable for an important case through kernelization. Our work demonstrates a new insight for employing resilience computation in database operations.