The datasets proposed in Lib2Desc: Automatic generation of security-centric Android app descriptions using third-party libraries paper are published in this repository.
DATASETS
TPL-Dataset:
This is a structured dataset including information about 1900 Android third-party libraries.
-
TPL_category_textual-description.csv
It includes the name, category, general textual description of third-party libraries(TPL).
-
TPL_textual_features.csv
It includes rhe name and textual feature vector extracted from the general textual description of TPLs.
-
TPL_API_features.csv
It includes the name and API feature vector obtained from static analysis of TPLs.
-
TPL_name_url.csv
It includes the name and the website resource of the description.
AAPI Dataset:
This dataset consists of two parts codes of Android API calls and Javadoc comment of the APIs extracted from original Android API documentation. The dataset is split in train, validation, and test sets.
-
code.csv
It includes the Android API calls preprocessed as mentioned in the paper.
-
javadoc.csv
It includes the matching API descriptions for the API calls. This textual information is preprocessed as mentioned in the paper.
The repository hierarchy:
├── AAPI-Dataset
│ ├── dev
│ │ ├── code.csv
│ │ └── javadoc.csv
│ ├── test
│ │ ├── code.csv
│ │ └── javadoc.csv
│ └── train
│ ├── code.csv
│ └── javadoc.csv
├── LICENSE
├── README.md
└── TPL-Dataset
├── TPL_API_features.csv
├── TPL_category_textual-description.csv
├── TPL_name_url.csv
└── TPL_textual_features.csv
Usage
The datasets can be used for developing machine learning, deep learning, NLG, NLP models. These datasets are constructed and preprocessed for developing classification, NLG, and transformer based seq2seq models.
Related Sections
You can find more information about dataset collection, preprocessing and models in the following sections.
Section 3 Data Collection
Section 3.1 Collecting Third-Party Libraries
Section 4.3.3 Datasets
Contact
Author: Beyza Cevik
Contact: beyzaccevik@gmail.com