This page is used as a supplementary web page for paper “αDiff: Cross-Version Binary Code Similarity Detection with DNN” (ASE ‘18). This page contains the model configuration details and our dataset link.
ACM Reference Format
Bingchang Liu, Wei Huo, Chao Zhang, Wenchao Li, Feng Li, Aihua Piao, Wei Zou.2018. αDiff:Cross-Version Binary Code Similarity Detection with DNN. In Proceedings of the 2018 33rd ACM/IEEE International Conference on Automated Software Engineering (ASE ’18), September 3–7, 2018, Montpellier, France.ACM,NewYork,NY,USA,12pages.https://doi. org/10.1145/3238147.3238199
CNN Structure
Table I: Structure of our CNN. The output shapes are described in rows x cols x filters. The kernel is specified as rows x cols x rowstride x colstride . The input shape is 100x100x1 and output embedding is 64-dimensional.
| Layer | Kernel | Output shape | Param# |
|---|---|---|---|
| input_1 | 100x100x1 | 0 | |
| conv2d_1 | 3x3x1x1 | 100x100x32 | 320 |
| batch_normalization_1 | 100x100x32 | 128 | |
| activation_1 | 100x100x32 | 0 | |
| conv2d_2 | 3x3x1x1 | 100x100x32 | 9248 |
| batch_normalization_2 | 100x100x32 | 128 | |
| activation_2 | 100x100x32 | 0 | |
| max_pooling2d_1 | 2x2x2x2 | 50x50x32 | 0 |
| conv2d_3 | 3x3x1x1 | 50x50x64 | 18496 |
| batch_normalization_3 | 50x50x64 | 256 | |
| activation_3 | 50x50x64 | 0 | |
| conv2d_4 | 3x3x1x1 | 50x50x64 | 36928 |
| batch_normalization_4 | 50x50x64 | 256 | |
| activation_4 | 50x50x64 | 0 | |
| max_pooling2d_2 | 2x2x2x2 | 25x25x64 | 0 |
| conv2d_5 | 3x3x1x1 | 25x25x96 | 55392 |
| batch_normalization_5 | 25x25x96 | 384 | |
| activation_5 | 25x25x96 | 0 | |
| conv2d_6 | 3x3x1x1 | 25x25x96 | 83040 |
| batch_normalization_6 | 25x25x96 | 384 | |
| activation_6 | 25x25x96 | 0 | |
| max_pooling2d_3 | 2x2x2x2 | 12x12x96 | 0 |
| conv2d_7 | 3x3x1x1 | 12x12x96 | 83040 |
| batch_normalization_7 | 12x12x96 | 384 | |
| activation_7 | 12x12x96 | 0 | |
| conv2d_8 | 3x3x1x1 | 12x12x96 | 83040 |
| batch_normalization_8 | 12x12x96 | 384 | |
| activation_8 | 12x12x96 | 0 | |
| max_pooling2d_4 | 2x2x2x2 | 6x6x96 | 0 |
| dense_1 | 6x6x512 | 49664 | |
| flatten_1 | 18432 | 0 | |
| dense_2 | 64 | 1179712 | |
| Total params: | 1,601,184 | ||
| Trainable params: | 1,600,032 | ||
| Non-trainable params: | 1,152 |
Dataset
You can get our dataset from https://github.com/twelveand0/alphadiff-dataset.