Several wide-area services are increasingly relying on bottleneck bandwidth estimation tools to enhance their network performance. Selfish hosts have, therefore, considerable incentives to fake their band-widths in order to increase their benefit in the network. In this paper, we address this problem and we investigate the vulnerabilities of current bottleneck bandwidth estimation techniques in adversarial settings. We show that finding "full-fledged" solutions for the multitude of attacks on the end-to-end bandwidth estimation process might not be feasible in the absence of trusted network components; we discuss solutions that make use of such trusted components. Nevertheless, we discuss other possible solutions that alleviate these threats without requiring trusted infrastructure support and we evaluate the effectiveness of our proposals on PlanetLab nodes.