One dimensional (1D) nanostructures offer a promising path towards a highly efficient heating and temperature control in integrated microsystems. The so called self-heating effect can be used to modulate the response of solid state gas sensor devices. Efficient self-heating was found to occur at random networks of nanostructured systems with similar power requirements than in highly ordered systems (e.g. individual nanowires, where its thermal efficiency attributed to the small dimensions of the objects). In this work, infrared thermography and Raman spectroscopy were used to map the temperature profiles of films based on random arrangements of carbon nanofibers during self-heating occurrence. Both techniques demonstrate consistently that heating concentrates in small regions, the here-called "hot-spots". Correlating dynamic temperature mapping with electrical measurements, we also observed that these minute hot-spots rule the resistance values observed macroscopically. A physical model of a random network of 1D resistors helped us to explain this observation. The model shows that, for a given random arrangement of 1D nanowires, current spreading though the network ends up defining a set of spots that dominate both the electrical resistance and the power dissipation. Such highly localized heating explains the high power savings observed in larger nanostructured systems. This understanding opens a path to design highly efficient self-heating systems, based on random or pseudo-random distributions of 1D nanostructures.