Biometric systems attempt to solve a matching problem through live measurements of human body features. However, one main barrier that prevents the widespread application of biometrics is the concern about the security and privacy of biometric information. To address this concern biometrics need to be protected with cryptography. However, the specific problems with biometrics, namely Number Limitation, Non-secrecy, Non-reproducibility and Non-cancelability, make it a challenge to secure biometrics effectively with existing cryptographic algorithms, especially on how to match two biometric templates in encrypted or hashed formats.

In biology genetic information is typically transformed with two processes: DNA (Source code, base 4) is transcribed into RNA (Intermediate code, base 4) and then RNA is translated into protein (Executables, base 20). In fact, the natural processes contain two ciphers. The first one we called Intronization Cipher refers to the splicing or removal of introns from gene to form RNA. The second one is the Substitution Cipher referring to the 1-to-N mapping from codon to amino acid. We carried out research with both ciphers in mind, but focused on the Intronization Cipher with an effort to develop it into a practical and stand-alone method called Artificial Intronization Method (AIM) to approach the problems of biometrics. AIM is a method of inserting introns into exon sequence to obtain ciphertext. Three methods are proposed to introduce introns into plaintext: PRNG, Integer Sequence and Geometric Key. The major advantage to use AIM is to prevent error propagation. One disadvantage of AIM is that security may depend on large Message Expansion Rate. Therefore, three methods are proposed to control the Message Expansion Rate: Intron Compression, Intron Removal, and Exon Elimination.

With AIM, the four problems can be approached as the following. Both Number Limitation and Non-cancelability problems can be solved by incorporating different sets of introns; Non-secrecy problem can be solved by using PIN-controlled intron sets; in theory, AIM could achieve zero-error propagation – a solution to the Non-reproducible problem. AIM, due to the intentional suppress of the diffusion property in favor of zero-error propagation, is vulnerable to known-plaintext attack. Therefore it has its limitation as a stand-alone cipher. However, we believe AIM can be an effective hashing mechanism for protecting fuzzy biometrics. Our testing results support this belief.

AIM, bearing some similarity with salting but is different from salting with the way we developed and applied it, can also be used as a preprocessing step for other cryptographic algorithms to enhance security.