Thanks for introducing this interesting idea of reconfigurating the Huffman tree by exploiting the freedom on the choice of the actual Huffman tree shape (as contrary to a canonical Huffman tree). I would like to pose the following two questions:

Why is skipping a small constant number of bits after outputting a DC sufficient for protecting encryption?

Can your approach be extended to other statistical compression schemes using binary tree representations like Huffman trees for the codewords? I have the following paper in mind:

Antonio Fariña, Travis Gagie, Szymon Grabowski, Giovanni Manzini, Gonzalo Navarro, Alberto Ordóñez Pereira:
Efficient and compact representations of some non-canonical prefix-free codes. Theor. Comput. Sci. 907: 11-25 (2022)