Trends in steganography¶
Classical methods (LSB substitution, DCT coefficient tweaking) are baseline comparisons rather than current practice. The field has moved toward neural and generative approaches that change what detection even means.
Embedding is now learned, not hand-crafted
Generation has replaced modification in the most advanced systems
Steganalysis and embedding are both AI-driven, and locked in an arms race
Covert channels have expanded beyond images to video, audio, network traffic, and language
The shift is structural, not incremental: