Data hiding in other media

We describe briefly the data hiding techniques developed for other media than digital images.

The Steganographic File System

Ross Anderson et alii have developed a proposal for a steganographic file system. [ANS98]

Such a file system basically hides the user's documents into other seemingly random files. To extract a document, the user must provide the system its name and a password. Otherwise, the presence of any data is undetectable.This technique relies on encryption schemes and pseudo-random number generators to achieve stealth.

A steganographic file system can protect from some threats:

• Soldiers and intelligence agents can be captured and tortured into revealing cryptographic keys and other secret data.
• When conducting delicate negotiations, such as between a company and a trade union, informal offers may be made which will be denied in the event of later litigation. However, the other side might obtain court orders for access to documents.
• Police power may be abused. An individual may be arrested on "suspicion" of a crime, found with an encrypted hard disk partition, and told that if he doesn't supply the password, this will be taken as an evidence of guilt. But the encrypted files can well contain confidential business information sought by people who have bribed the police;
• Private individuals have been tortured by robbers into revealing informations such as the secret codes for their bank cards and the location of safes.

The steganographic file system would allow the owner in such situations to give the names and passwords for some of the files, while keeping secret the existence of the others. The enemy won't be able to tell if other documents are present or not. For example, one could give access to his love letters and tax records, but keep quiet about the secret plans he is developing.

The authors didn't resort in using classical data hiding in audio, video or still images because those techniques have a very low bit rate and are hard to implement in a transparent way for the OS. [ANS98]

Data hiding in audio

Bender et alii describe various techniques for data hiding in digital audio. [Ben96]

These techniques can't rely on many of the methods described for images because of the better performances of the Human Auditory System (HAS) with respect to the Human Visual System (HVS).

The authors have sought coding schemes resistant to most of the usual manipulation of sounds, like A/D-D/A conversion, radio broadcast, "over the air" playing, resampling, lossy compression.

The first proposed method is the easy and high bit rate LSB insertion, that is easily destroyed by anything else than pure digital transmission.

The second technique is based on the HAS sensitivity only for differential phase variation, but relative insensitivity to initial phase. Thus, the sound file is divided into blocks and each block's initial phase is modified using the embedded message, preserving the subsequent relative phase shifts. This is one of the better techniques with respect to perceived signal to noise ratio.

Third, spread spectrum schemes can be used, even if they usually add perceivable noise to the sound. However, the embedded signal can be filtered through a perceptual mask, so that the most audible components of the added noise are reduced in power.

Last, the authors hid data by adding echo to the audio signal, using two different delays to encode bits. Both of these delays were chosen small enough to be heard by the naked ear as enrichments in sound, and not as distortions. This technique had a good robustness and fairly high data rate. [Ben96] Moreover, is the only one that can resist a jitter attack. [PAK98]

Data Hiding in text

Bender et alii showed some of the techniques used to hide data in text files. [Ben96]

These methods are divided in open-space, syntactic and semantic.

The formers use white space to encode hidden informations. They can modify the spaces between words, or the spaces at the end of sentences. The problem with these approaches is that most of the word processors clean up or reformat spaces in files transparently, so the message can easily be destroyed.

Syntactic methods are based on modifying the structure of text by changing punctuation or word order. They are quite difficult to implement in an automatic way without being visible. In fact mispunctuation is usually readily visible after a certain extent, and the order of the words in a sentence can change its tone in a strange way.

Semantic methods exploit equivalences between words (synonyms) to encode a hidden message, by choosing one word or the other in a synonyms pair. The problem here is one of style, since sometimes words can't be exchanged for "equivalent" ones without changing profoundly the meaning of a sentence.

A common limitation to all of these techniques for data hiding in texts is the very low data rate they can achieve. [Ben96]

References:

[ANS98]

[Ben96]

[PAK98]

Previous: Moving images

Next: Invisible Digital Watermarking