The vast interconnectivity of devices in the Internet of Things (IoT) and Cyber-Physical systems under wireless environments facilitates information exchange, but challenges network security. In this paper, cooperative beamforming for secure communications is studied in wireless cyber-physical systems, where two legitimate devices communicate with the help of amplify-and-forward (AF) relays, and the eavesdropping devices around purpose to intercept the information. To achieve high secrecy sum rate, a secrecy sum rate maximization problem via cooperative beamforming with or without artificial noise is formulated under individual relay power constraint. The problem is shown to be non-convex. We then reformulate it into a difference of convex programming problem, relax it by a tight semidefinite relaxation (SDR), and approximate it into sequential convex problems by first-order Taylor expansion, which eventually leads to a high-rate beamforming solution. The simulations show that our developed beamforming scheme can achieve high secrecy sum rate and fast convergence.