<p>By leveraging radio frequency (RF) signals, uncertainty and instability of ambient signals result in vulnerable information security. To facilitate confidential data transmission, this paper proposes a robust artificial-noise (AN)-aided relay-enabled framework for energy harvesting backscatter communication (EH-BC) systems. In particular, the injected AN and amplified direct interference are jointly utilized to degrade the quality of multiple eavesdropping links. To maximize the secrecy rate, the cooperative beamforming (CB) vector, AN covariance matrix, and power splitting (PS) factor at the reader are jointly optimized. A three-level optimization approach is developed based on semidefinite relaxation (SDR) and one-dimensional search techniques. For enhancing the robustness of the constructed framework, the above result can be extended to the colluding eavesdropping scenario, where the channel state information (CSI) from the relay to the reader and the colluding eavesdropper are both imperfect. Besides, a robust optimization design based on worst-case is developed, where the CSI errors are norm-bounded. By using variable substitutions and the S-procedure, we derive equivalent forms of the constraints and then transform the non-convex robust design into a convex optimization problem. Simulation results show that the proposed robust AN scheme significantly enhances security performance compared with the benchmark schemes.</p>