The new technique was first detailed in "Designing a cluster-based covert channel to evade disk investigation and forensics," a recent paper written by researchers from the University of Southern California at Los Angeles and the National University of Science and Technology (NUST) in Islamabad, Pakistan.
"There could be a number of potential uses for this technology, but the main strength of our technique lies in its ability to conceal information in cases where encryption cannot be used--e.g., where the presence of encrypted data would appear suspicious and may be deemed an unacceptable risk to the communicating parties," said report co-author Fauzan Mirza, a communication systems engineering professor at NUST, in an email interview. "The obvious application of these techniques would be among people or organizations that need to protect information against powerful--well-resourced--adversaries, such as spies, terrorists, whistleblowers, political groups, etc."
This type of covert channel could also have enterprise security applications, such as creating a covert password safe. Likewise, "it could also be used to implement a software copy protection mechanism or information tracking/watermarking mechanism," or even as part of a data leakage protection mechanism, said Mirza. "We did not go into these applications, since--as academics--we wanted to bring to light the simplicity and novelty of the idea, rather than dwell on the applications. We left that part to the readers."
How much data could be hidden with this technique? Based on a study of the disk fragmentation levels of 52 drives, the researchers estimate that the average 160GB drive could safely store about 20MB via this covert channel.
The technique works by exploiting how operating systems handle groups of consecutive sectors on a disk, which together form the clusters that store file and directory contents. "While allocation strategies vary, most operating systems allocate consecutive clusters to files consisting of multiple clusters," according to the report. "This approach works well until there are no consecutive unallocated clusters available. In that case, the contents of the file are scattered or fragmented across the file system." In other words, disk fragmentation is a natural phenomenon for these file systems and one that the researchers propose mimicking to hide secret data.
The researchers used the FAT32 file system to illustrate how their technique works (but said that it will also work on many other types of file systems, including NTFS). In FAT32, a cluster uses 32 bits, but reserves four of those. As a result, a covert channel can be created that uses the remaining bits to store information, including cluster location information so that the data distributed amongst clusters can be reassembled, provided you have the decryption key.
Interestingly, disk fragmentation turns out to be extremely common, especially with modern operating systems. For the report, the researchers studied the fragmentation of 52 disk drives, ranging in size from 60GB to 500GB, and found that only 4% lacked sufficient fragmentation, and that on average, 5.6% of all files were fragmented, and that data and log files--especially those created by recent Windows operating systems, or antivirus software--were especially prone to fragmentation.
Mirza said that forensic investigators could deconstruct the techniques the researchers use to illustrate their concept--two are in their report--to create an algorithm that detects the methods. But he said that "if the implementation of our scheme is well designed and used correctly, then it may be difficult for the investigator to determine decisively whether some information is hidden or not."