LibvDiff-Dataset

We collected the source code of 9 widely used OSS from Github [13] and their official websites as Table 1 shows. These OSS can be categorized into different groups based on their functionality, such as document formatting, and compression. We compiled the source code into binaries using different compilation options, including4 architectures (ARM, X86, X64, PPC) and 4 optimization levels (O0, O1, O2, O3) with GCC v9.4.0. In total, we obtained 168 distinct versions of all OSS, resulting in 2688 (168 * 16) binaries as shown in Table 1. 

Accept for the binary dataset, an extra dataset with feature and dataset example for quick start is also provided.

 dataset_features_example.tar.gz: it includes features that we have already generated and can be evaluated quickly.

OSS_version_dataset.tar.gz: it includes the binary dataset mentioned in our paper, including 9 OSS, 168 versions and 2688 binaries, more details could be found in our paper (Sec 6). 

The source code of LibvDiff is available at heritage

The strucutre of dataset  is organized as:

- OSS # e.g. freetype
|-- Lib of OSS # e.g. libfreetype
|-- architecture # e.g. ARM
|-- optimization # e.g. O0
|-- version # e.g. VER-2-4-1
|-- binary # e.g. freetype-2.4.0