Nano-machines have the potential to achieve targeted drug delivery, improving efficacy and reducing side-effects and wastage. Coordinating multiple nano-machines require communication capability, and molecular communications (MC) is recognised as a key technology under the Internet-of-Nano Things (IoNT) ecosystem. Whilst secure communication is critical, it is limited by the low computational capability of nano-machines. Here, we leverage on the rapid advances in radio physical layer security (PLS) to propose cipher keys generated from dynamic molecular channel statistics for encryption. Like PLS, codewords do not need to be exchanged and a public codeword library is not required, decreasing the computational and architectural burden for ad-hoc nano-machine coordination. Our results show that we can achieve eavesdropper key disagreement rate (KDR) of 5-7 × higher than our intended transmission channel, demonstrating that any potential malicious eavesdropper cannot decipher the message and hence cannot go on to perform malicious actions.