Secure Beamforming Design Against Sensing Eavesdroppers in Integrated Sensing and Communication Systems

Abstract

In integrated sensing and communication (ISAC) systems, sensing information privacy is vulnerable to leakage, as ISAC waveforms can be intercepted by sensing eavesdroppers to infer target-related information, such as direction, location, and presence. To enhance the sensing security of ISAC, this paper studies the secure ISAC beamforming design in the presence of a sensing eavesdropper, whose exact location is unknown but confined within a known uncertainty region. Considering that the eavesdropper employs beam scanning and energy detection to determine target presence, we derive a closed-form expression for its target detection probability, and introduce the average eavesdropping detection probability (A-EDP) over the uncertainty region as the sensing security metric. Then, we optimize the transmit beamforming vectors at the base station (BS) to minimize the A-EDP, subject to individual per-user achievable-rate constraints for communication users, the position error bound (PEB) for target positioning, and the total transmit power budget. To address the resulting non-convex problem, we propose a novel Warm-started Adaptive Riemannian Memetic Genetic Algorithm (WARM-GA), which combines geometry-guided initialization with feasibility-aware local refinement. Simulation results demonstrate that, compared with the projected gradient descent (PGD), $\epsilon $ -constrained differential evolution ( $\epsilon $ -DE), and Vanilla GA baselines, the proposed WARM-GA converges faster and achieves substantially lower A-EDP while satisfying all communication and sensing constraints.