A wireless sensor network (WSN) refers to a network of sensor nodes that collaboratively work to sense, monitor, and control their surrounding environments. As WSNs are integrated into the Internet of Things, it is crucial to protect the network against malicious security attacks considering its wide applicability, such as military monitoring, healthcare, and civilian applications. Thus, in this paper, we consider a physical-layer security technique exploiting analog cooperative beamforming (ACB), where multiple sensor nodes create a virtual antenna array (VAA) and locally adapt their phases. As WSNs inherently have clustered topology, we model the VAA elements' locations by Gaussian distributions. The secrecy capacity of the ACB with the Gaussian-distributed elements is derived in a closed-form expression. The theoretical and numerical results indicate that the ACB-based schemes provide a better secrecy rate compared to the conventional co-located antenna array-based schemes. In addition, the ACB with Gaussian-distributed elements can better suppress side-lobe, which causes undesired information leakage to certain directions, compared to the ACB with uniformly distributed elements. Furthermore, we investigate the impacts of fading channel and phase estimation error in the ACB.