Here is a pitfall to be avoided in Steganography using low bits of AD converter output. Such converters may be biased in their low bit. [...] Falling too close to 50% would be a clue that the data
That's not too risky for graphics, since many scanned pictures aren't based on raw A/D converter output; they've been processed and squashed down to some smaller number of bits. A more serious concern is compressibility - a real image file is probably more compressible than a file with the low-order bit replaced by a crypto-bit, since the real data has moderate correlation and the crypto-bits are random. I doubt the Feds will immediately start looking to see if you're shipping GIF files that have significantly worse compression than average, but they'd probably find something if they did it. Bill Stewart
A more serious concern is compressibility - a real image file is probably more compressible than a file with the low-order bit replaced by a crypto-bit, since the real data has moderate correlation and the crypto-bits are random. I doubt the Feds will immediately start looking to see if you're shipping GIF files that have significantly worse compression than average, but they'd probably find something if they did it.
how many bits are we talking about here? suppose it's two in sixteen. 7/8 of the compressible bits remain. so if the normal compressibility is 2:1, taking two out of sixteen bits would leave 1.75:1 compression. is that a "notable" difference? i haven't been paying close enough attention -- is two out of sixteen a realistic amount? it seems high to me. if it's one out of sixteen, the effect is only a 6.25% reduction in compression. is that notable? peter
participants (2)
-
peter honeyman -
wcs＠anchor.ho.att.com