We propose a novel inversion algorithm to image buried objects in inhomogeneous media from the electromagnetic data on the outer boundary. Our method is based on exploiting the Cauchy data to derive a new Born-like linearization of the inverse problem. The main advantage of this formulation is to avoid the use of the background Green function and therefore is computationally more efficient. It also provides better accuracy than classical Born approximation. In the case of stratified media, our approach can be coupled with any appropriate continuation method. We discuss here the coupling with a continuation method based on the use of approximate transmission conditions. The feasibility and robustness of our methodology is validated through numerical experiments for single and multiple targets.