As embedded devices, Internet of Things systems, and cyber-physical platforms become more widespread, there is a growing need to characterize software and hardware behavior without privileged access or intrusive instrumentation. In many real deployments, conventional profiling is impractical because instrumentation may be too costly, unavailable, or disruptive to the behavior being measured. In contrast, electromagnetic (EM) signals can be captured externally in a contactless, non-intrusive manner and still carry information about underlying computation and hardware activity. This thesis develops methods that use externally captured EM signals for software and hardware characterization. The first contribution is a hierarchical framework for detecting multiple periodicities in software code analysis, enabling recovery of nested periodic structures, usually caused by execution of loop nests in software, from EM side-channels for fine-grained behavioral analysis. The second contribution is an efficient dissimilarity detection approach for time-series with application to EM side-channel analysis, supporting sensitive detection of deviations from reference behavior for program monitoring. The third contribution extends EM-based characterization to the wireless domain through hardware model identification, showing that measurement signatures can be used to distinguish radio models based on consistent hardware-dependent patterns. Together, these contributions demonstrate that EM signals provide a practical foundation for non-intrusive characterization of both software behavior and hardware platforms.