Electromagnetic side-channel measurements depend on where a probe sits above a chip. This paper studies how to make profiling attacks less sensitive to that placement.
A single neural network is trained using traces from multiple probe positions, allowing it to detect leakage across a wider area of the device. The evaluation also crosses laboratory boundaries: measurements from one lab train the model, while traces collected independently in another lab test it. This examines portability under conditions that are harder to reproduce than a fixed laboratory setup.
