In wireline networks, random network coding combined with path diversity provides inherent protection against eavesdropping since the eavesdropper has to compromise nodes along each diverse path, presumably at significant risk and cost per compromise, in order to acquire enough information to decode the source data. This touted ``free security'' does not necessarily apply, however, towireless networks. Since a sophisticated over-the-air eavesdropper at a protected site away from the network can receive transmissions from multiple nodes, without incurring significant additional risk or expense, any inherent protection that network coding provides in this context is unclear. In this paper, we investigate the eavesdropping vulnerability of a mobile ad-hoc network using the current state-of-the-art in random linear network coding and subgraph generation techniques. Specifically, we develop a simulation framework with proper metrics to quantify the eavesdropper's decoding success probability for a network-coded wireless network. Our results, obtained from extensive Monte Carlo simulations, demonstrate that such networks are indeed vulnerable to eavesdropping-and to a remarkable degree. Finally, we experimentally confirm the high degree of eavesdrop vulnerability using packets captured from an operational prototype testbed.