A complex situation in data recovery

Abstract

The research considers an unusual situation in data recovery. Data recovery is the process of recovering data from recording media that is not accessible by normal means. Providing that the data has not been overwritten or the recording medium physically damaged, this is usually a relatively simple process of either repairing the file system so that the file(s) may be accessed as usual or finding the data on the medium and copying it directly from the medium into normal file(s). The data in this recovery situation is recorded by specialist call centre recording equipment and is stored on the recording medium in a proprietary format whereby simultaneous conversations are multiplexed together and can only be accessed by using associated metadata records. The value of the recorded data may be very high especially in the financial sector where it may be considered a legal audit of business transactions. When a failure occurs and data needs to be recovered, both the data and metadata information must be recreated before a single call can be replayed. A key component to accessing this information is the location metadata that identifies the location of the required components on the medium. If the metadata is corrupted, incomplete or wrong then a repair cannot proceed until it is corrected. This research focuses on the problem of verifying this location metadata. Initially it was believed that only a small set of errors would exist and work centred on detecting these errors by presenting the information to engineers in an at-a-glance image. When the extent of the possible errors was realised, an attempt was made to deduce location metadata by exploring the content of the recorded medium. Although successful in one instance, the process was not able to distinguish between current and previous uses. Eventually insights gained from exploration of the recording application's source code, permitted an intelligent trial and error process which deduced the underlying medium apportioning formula. It was then possible to incorporate this formula into the heuristics, generating the at-a-glance image, to create an artefact that could verify the location metadata for any given repair. After discovering the formula, the research returned to the media exploration and the produced disk fingerprinting technique. The disk fingerprinting technique gave valuable insights into error states in call centre recording and provided a new way of seeing the contents of a hard drive. This research provided the following contributions: 1. It has provided a means by which the recording systems' location metadata can be verified and repaired. 2. As a result of this verification, greater automation of the recovery process is now possible before the need for human verification is required. 3. The disk fingerprinting process. This has already given insights into the recording system's problems and is able to provide a new way of seeing the contents of recording media.