Tomáš Pevný presents 53rd edition of the PRAGUE COMPUTER SCIENCE SEMINAR: Steganography
On 2023-03-09 16:15:00 at Auditorium S5, MFF UK Malostranské nám. 25, Praha 1
Steganography
The lecture will be followed by a discussion
ABSTRACT
Steganography communicates a secret message as an innocuous looking object
concealing the fact that something important is being communicated.
Steganography is therefore more devious than cryptography, since when someone
uses the latter, everyone knows that something is being hidden. Throughout
history, steganography has been used by spies, agents, dissidents, and
criminals. The oldest documented uses come from the old Greece at 5BC. With the
advent of digitization, steganography has moved to the new media such as digital
images, video, text, file systems, etc.
I will introduce the topic by talking about major breakthroughs in which I have
participated, which have ultimately led to the most principled design of
steganographic algorithms known to date, theoretically applicable to a diverse
set of media such as images or sound. We will explain how error-correcting codes
revolutionized the field. Since then, creating a new steganographic algorithm is
equal to finding a good model of the noise. This is surprisingly difficult. Our
first algorithm embracing this paradigm was broken in the first steganographic
contest that we have organized. The rest of the talk would be devoted to the
design of noise models by a combination of game theory and adversarial
machine learning.
The lecture will be followed by a discussion
ABSTRACT
Steganography communicates a secret message as an innocuous looking object
concealing the fact that something important is being communicated.
Steganography is therefore more devious than cryptography, since when someone
uses the latter, everyone knows that something is being hidden. Throughout
history, steganography has been used by spies, agents, dissidents, and
criminals. The oldest documented uses come from the old Greece at 5BC. With the
advent of digitization, steganography has moved to the new media such as digital
images, video, text, file systems, etc.
I will introduce the topic by talking about major breakthroughs in which I have
participated, which have ultimately led to the most principled design of
steganographic algorithms known to date, theoretically applicable to a diverse
set of media such as images or sound. We will explain how error-correcting codes
revolutionized the field. Since then, creating a new steganographic algorithm is
equal to finding a good model of the noise. This is surprisingly difficult. Our
first algorithm embracing this paradigm was broken in the first steganographic
contest that we have organized. The rest of the talk would be devoted to the
design of noise models by a combination of game theory and adversarial
machine learning.
External www: https://www.praguecomputerscience.cz/