In physical-layer security, communication secrecy is commonly achieved by providing the legitimate receiver a physical-layer advantage over the eavesdropper or, equivalently, by making the eavesdropper's channel more degraded than the legitimate receiver's, a setting traditionally known as the degraded wiretap channel. There may be situations, however, where such advantages cannot be guaranteed. One example is the wireless channel where the legitimate receiver's channel may be more degraded than the eavesdropper's owing to a deep fade (outage), additive interference at the edge of a network cell, etc. To our knowledge, such scenarios have not been remedied in the related literature. Thus, we propose in this work the first physical-layer security scheme that achieves communication in a setting where the eavesdropper has the physical-layer advantage over the legitimate receiver, a scenario that we call the reversely-degraded wiretap channel. Precisely, we show that by constraining the transmitter's input to be discrete uniformly distributed (such as an on-off keying modulation) and by using nonlinear threshold-based signal detection, the legitimate receiver is guaranteed to have a positive secrecy capacity even though the eavesdropper has a better signal-to-noise ratio (SNR). The former requirement on the input distribution is to limit the capacity gains of the eavesdropper at high SNR. The latter requirement allows the communication channel to exhibit supra-threshold stochastic resonance, a phenomenon whereby noise improves signal detection and increases the mutual information. A numerical example is provided to confirm our claims.