The world of security research is fraught with ethical dilemmas, and open research on how to backdoor and subvert systems certainly brings its fair share. Releasing code for an intentionally flawed encryption implementation or a design for a system that appears to be secure, but in fact allows an attacker to easily recover secret data - this pushes the limits of doing more harm than good, yet is critically important for defenders. Without solid research into how systems could be effectively and efficiently backdoored, defenders quickly fall behind, giving clever attackers a strong upper hand. Defenders, those charged with reviewing code and designs may be able to easily detect the cliched sending emails with credit card numbers, but how prepared are they for a better class of attacker? This talk seeks to cover both the need for the information gained, especially through events such as the Underhanded Crypto Contest, and the risks that it presents.