A comparative study of file-type identification techniques

Research in file-type identification has employed a number of different approaches to classify unknown files according to their actual file type. However, due to the lack of implementation details in much of the published research and the use of private datasets for many of those projects, it is often not possible to compare new techniques with the prior work. In this paper, we present a comparison of five common file-type identification approaches, along with the parameters used to perform the comparisons. All approaches were evaluated with the same dataset which was drawn from public or widely-available sources. Our results show that each approach can produce good results with 88% to 97% classification rates, but achieving these results requires “tuning” the parameters of the inputs to the classifiers.