Steganography 2 Flashcards
How secure is the stego-object
Works on the principle of security by obscurity
However if an attacker believes we are sending a hidden message it is relatively easy for him or her to guess the extraction algorithm and thereby obtain the secret
A better approach is to use the random interval method in which elements of the cover are chosen at random for hiding data
Random interval method
Use a psuedo random number generator (PRNG) to generate a sequence of random intervals
A number is used to seed the PRNG. This number is the stego-key
The stego key is used by the sender to generate the sequence of random intervals for encrypting the secret message
The stego key is also used by the reciever to generate the same sequence of random intervals that allows him or her to decrypt the secret message
Visual attacks on steganography
Often the image structure is visible in the least significant bits of an image
We can use this property to check for embedding in the LSBs of an image
Visual attack
A visual attack is a means of detecting embedded information
In general image stricture is lost when embedding has taken place
Can also tell us the proportion of the image that has been used for embedding
Can work well for sequential embedding
Does not work well for random interval embedding
statistical attacks on steganography
Used when referring to non-visual methods for determining embedding
Embedding secret info in the least significant bits of an image changes the image histogram
Image histogram
Is the frequency distribution of grey-levels within a digital image
Chi-squared statistic
Analyse an image histogram using the chi-squared statistic to obtain a probability of embedding
For simple 4-bit images we can determine the chi-squared statistic by hand and use a table to see if the value obtained is a significant at the 5% level
Statistics
Low value for chi-squared statistic suggests embedding
High value for chi-squared suggests no embedding
The null hypothesis is that embedding has taken place
The alternative hypothesis is that embedding has not taken place
If your calculated value is greater than the tabulated value we can be confident that the image doe snot contain embedded information