Routing and spectrum allocation for connection requests in elastic optical networks are investigated. Considering the influence of physical nodes on the security, a global constrained optimization model is established, which is with the requirement of meeting the minimum security level as the constraint and with the maximum sequence number of frequency slots in networks as the optimization goal. In order to solve the constrained optimization model effectively, a global optimization algorithm is presented. The connection requests groomed are sorted according to a specific sorting policy. K paths, which satisfy the minimum security level requirement, are selected for every connection request. To minimize the maximum sequence number of frequency slots, the improved genetic algorithm is employed to determine the optimal routing and spectrum allocation scheme. In order to verify the validity of the proposed algorithm, simulations are conducted in different network topologies, and the results show that the efficient spectrum allocation can be achieved by the proposed algorithm.