Limitations of scanned human copresence encounters for modelling proximity-borne malware

Patterns of human encounters, which are difficult to observe directly, are fundamental to the propagation of mobile malware aimed at infecting devices in spatial proximity. We investigate errors introduced by using scanners that detect the presence of devices on the assumption that device copresence at a scanner corresponds to a device encounter. We show in an ideal static model that only 59% of inferred encounters correspond to actual device copresence. To investigate the effects of mobility, we use a simulator to compare encounters between devices with those inferred by scanners. We show that the statistical properties of scanned encounters differ from actual device encounters in ways which impact malware propagation dynamics, a form of aggressive data dissemination. In addition to helping us understand the limitations of encounter data gathered by scanners in the field, our use of virtual scanners suggests a practical method for using these empirical datasets to better inform simulations of proximity malware outbreaks and similar data dissemination applications.