Modified honey encryption scheme for encoding natural language message

Abstract

Conventional encryption schemes are susceptible to brute-force attacks. This is because bytes encode utf8 (or ASCII) characters. Consequently, an adversary that intercepts a ciphertext and tries to decrypt the message by brute-forcing with an incorrect key can filter out some of the combinations of the decrypted message by observing that some of the sequences are a combination of characters which are distributed non-uniformly and form no plausible meaning. Honey encryption (HE) scheme was proposed to curtail this vulnerability of conventional encryption by producing ciphertexts yielding valid-looking, uniformly distributed but fake plaintexts upon decryption with incorrect keys. However, the scheme works for only passwords and PINS. Its adaptation to support encoding natural language messages (e-mails, human-generated documents) has remained an open problem. Existing proposals to extend the scheme to support encoding natural language messages reveals fragments of the plaintext in the ciphertext, hence, its susceptibility to chosen ciphertext attacks (CCA). In this paper, we modify the HE schemes to support the encoding of natural language messages using Natural Language Processing techniques. Our main contribution was creating a structure that allowed a message to be encoded entirely in binary. As a result of this strategy, most binary string produces syntactically correct messages which will be generated to deceive an attacker who attempts to decrypt a ciphertext using incorrect keys. We evaluate the security of our proposed scheme.