Jump to content

Steganography

From GAMAYUN+ Wiki

Cryptography hides what a message says. Steganography hides that a message exists at all — an encrypted file announces "there's a secret here," a steganographic one doesn't. The word is Greek for "covered writing," and the technique predates computers by millennia: Herodotus records a Greek noble tattooing a message onto a slave's shaved scalp, then waiting for the hair to grow back over it before sending him through enemy territory.[1] The digital version most modern tools use — LSB (least-significant-bit) encoding — swaps the lowest, least visually significant bit of each color channel in an image for a bit of the hidden message, nudging every affected pixel's color by an amount too small to see. It's also fragile on purpose: anything that recompresses the image (a JPEG re-save, an upload to a site that reprocesses photos) scrambles those low bits and destroys the message, which is why LSB tools generally only ever output PNG.

From pigeons to pixels

The technique kept reinventing itself with whatever medium was available. During the siege of Paris in the Franco-Prussian War (1870–71), photographer René Dagron used microfilm to let a single carrier pigeon smuggle huge volumes of correspondence past the blockade — not steganography in the strict "hidden" sense (everyone knew pigeons carried messages), but the direct ancestor of the microdot: a photograph reduced until it's the size of a printed period. Emanuel Goldberg formalized the two-stage reduction process for true microdots in 1925, and by World War II, German intelligence was mailing them concealed on ordinary-looking letters, embedded in the depth of a "." on a typed page, or hidden on couriers' clothing and even inside dolls carried as gifts — the FBI's wartime Special Intelligence Service spent years unraveling espionage rings across Latin America built partly around exactly this delivery method.[2]

A case the medium itself gave away

Digital image steganography isn't just a classroom exercise — it's shown up in real intelligence work. In 2010, the FBI's decade-long Operation Ghost Stories ended with the arrest of ten Russian intelligence officers living under deep-cover American identities. Part of how they communicated with Moscow was embedding encrypted messages into ordinary-looking images and posting them to public websites, where handlers would download the same images and extract the hidden payload with matching software.[3] The steganography itself reportedly held up — what didn't hold up was everything around it (tradecraft mistakes, physical surveillance, a cooperating source), which is the pattern in essentially every real steganography case that's ever become public: the hidden-message layer is rarely the point of failure.

How steganalysis actually catches it

Detecting LSB steganography specifically doesn't require breaking any code — it requires noticing that "real" pixel data (from a camera sensor, or ordinary JPEG compression) has statistical noise patterns that are subtly different from noise generated by flipping bits to encode a message. Techniques like chi-square attacks and RS (regular/singular) analysis look for exactly that mismatch: an image that's had data embedded in its low bits tends to have measurably different statistical structure in those bits than one that hasn't, even though nothing is visible to the eye.

Beyond images

LSB-in-images is the easiest version to build and to explain, but it's one technique among many. Audio steganography hides data in similarly inaudible bit-level or frequency-domain changes; video steganography gets an entire extra dimension (time) to spread a payload across; and text steganography ranges from crude (invisible Unicode characters, trailing whitespace patterns) to genuinely clever (choosing between synonymous words or sentence structures according to a hidden bit pattern). Statistical structure that looks meaningless but resists every attempt to explain it away is also the entire premise of the Voynich Manuscript — the open question there is whether anyone will ever find the equivalent of a working steganalysis technique for a 600-year-old handwritten codex.

See also