Abstract: We present an alternative to the controversial ``key escrow'' techniques for enabling law-enforcement and national security access to encrypted communications. Our proposal allows such access with probability p for each message, for a parameter p between 0 and 1 to be chosen (say, by Congress) to provide an appropriate balance between concerns for individual privacy, on the one hand, and the need for such access by law-enforcement and national security, on the other. (For example, with p=0.4, a law-enforcement agency conducting an authorized wiretap which records 100 encrypted conversations would expect to be able to decrypt (approximately) 40 of these conversations; the agency would not be able to decrypt the remaining 60 conversations at all.) Our scheme is remarkably simple to implement, as it requires no prior escrowing of keys.
We implement translucent cryptography based on non-interactive oblivious transfer. Extending the schemes of Bellare and Micali, who showed how to transfer a message with probability 1/2, we provide schemes for non-interactive fractional oblivious transfer, which allow a message to be transmitted with any given probability p. Our protocol is based on the Diffie-Hellman assumption and uses just one El Gamal encryption (two exponentiations), regardless of the value of the transfer probability p. This makes the implementation of translucent cryptography competitive, in efficiency of encryption, with current suggestions for software key escrow.
Ref: Journal of Cryptology, Vol. 12, No. 2, 1999, pp. 117--140. (Early version was MIT Laboratory for Computer Science Technical Memo No. 683, February 1996.) Full paper of most recent version available below.
Full paper: Available as compressed postscript, postscript, or pdf. ( Help if this doesn't work).