I really hate these "did you know that old thing + new trendy thing?!?!?!" articles. Yes, you can apply steganography to AI. Yes, you can use blockchain to secure grocery store coupons. Yes, you can train a neural network to operate Tinder for you.
This seems both novel and non-trivial to me (admittedly not a cryptographer)
> In this work, we consider the information-theoretic model of steganography introduced in (Cachin,1998). In Cachin (1998)’s model, the exact distribution of covertext is assumed to be known to all parties. Security is defined in terms of the KL divergence between the distribution of covertext and the distribution of stegotext. A procedure is said to be perfectly secure if it guarantees a divergence of zero.
In a practical scenario, the attacker likely does not know the distribution of covertext and therefore cannot detect a theoretically imperfect steg implementation. It's an area where the steg user has a huge advantage, as there are a practically infinite number of ways to do steganography and it's non-trivial to even detect that it is being used in the first place. And if you obtain the covertext distribution by say hacking the computer generating the steg, then why not just exfiltrate the plaintext directly?
I believe the gp was talking about hating a style of journalism.