Machine-generated messages, such as text, images, and video, can now be used to achieve perfect security for steganography, a method of hiding secret messages within other messages. The use of large language models and minimum entropy coupling algorithms allows for statistically indistinguishable machine-generated content that conceals hidden messages. However, the practical implementation of perfect security in steganography is limited by computational efficiency.

4m read timeFrom quantamagazine.org
Post cover image
Table of contents
Shaved Heads and Other StrategiesFormalizing SecrecyAchieving Perfection

Sort: