THIRD_PARTY_LICENSES

# Third-Party Dataset Licenses

The CoBRA library is licensed under Apache-2.0 (see LICENSE). The test
datasets in test/datasets/ are redistributed from third-party research
projects under their original licenses, which may differ from the
library license. This file documents the provenance and license of each
dataset.

Users who redistribute CoBRA's test datasets must comply with the
license terms listed below.

## Summary

The "Effective License" column shows the most restrictive license in
the redistribution chain. Where the original project has no license
but was redistributed through a GPLv3 repository (SiMBA++ or GAMBA),
the GPLv3 terms of the redistributor apply.

| Directory / File        | Taken From  | Originally By   | Original License | Effective License |
|-------------------------|-------------|-----------------|------------------|-------------------|
| univariate64.txt        | Simplifier  | Simplifier      | GPLv3            | GPL-3.0-only      |
| multivariate64.txt      | Simplifier  | Simplifier      | GPLv3            | GPL-3.0-only      |
| permutation64.txt       | Simplifier  | Simplifier      | GPLv3            | GPL-3.0-only      |
| msimba.txt              | Simplifier  | Simplifier      | GPLv3            | GPL-3.0-only      |
| simba/e*_*vars.txt      | SiMBA++     | SiMBA (Denuvo)  | GPLv3            | GPL-3.0-only      |
| simba/pldi_*.txt        | SiMBA++     | MBA-Solver      | GPLv3            | GPL-3.0-only      |
| simba/blast_dataset*.txt| SiMBA++     | MBA-Blast       | Not specified    | GPL-3.0-only      |
| simba/test_data.txt     | SiMBA++     | NeuReduce       | Not specified    | GPL-3.0-only      |
| gamba/loki_tiny.txt     | GAMBA       | Loki            | AGPLv3           | AGPL-3.0-only     |
| gamba/neureduce.txt     | GAMBA       | NeuReduce       | Not specified    | GPL-3.0-only      |
| gamba/mba_obf_*.txt     | GAMBA       | MBA-Obfuscator  | MIT              | GPL-3.0-only      |
| gamba/syntia.txt        | GAMBA       | QSynth          | Not specified    | GPL-3.0-only      |
| gamba/qsynth_ea.txt     | GAMBA       | QSynth          | Not specified    | GPL-3.0-only      |
| gamba/mba_flatten.txt   | GAMBA       | MBA-Flatten     | Not specified    | GPL-3.0-only      |
| oses/oses_*.txt         | OSES        | OSES            | Not specified    | Not specified     |
| obfuscatorx.txt         | —           | ObfuscatorX     | Unknown          | Unknown           |

Note: `simba/test_data.txt` and `gamba/neureduce.txt` are both the
NeuReduce dataset with minor variable renaming (`b` vs `t`).

## Immediate Sources

### Simplifier (GPL-3.0-only)

Repository: https://github.com/mazeworks-security/Simplifier
License:    GPL-3.0-only

Source of univariate64.txt, multivariate64.txt, permutation64.txt,
and msimba.txt.

### SiMBA++ (GPL-3.0-only)

Repository: https://github.com/pgarba/SiMBA-
License:    GPL-3.0-only

The simba/ directory was sourced from the SiMBA++ repository. SiMBA++
itself aggregates datasets from several earlier projects (MBA-Solver,
MBA-Blast, NeuReduce). The SiMBA++ repository is licensed under GPLv3;
the original projects' licenses are listed below.

### GAMBA (GPL-3.0-only)

Repository: https://github.com/DenuvoSoftwareSolutions/GAMBA
License:    GPL-3.0-only

The gamba/ directory was sourced from the GAMBA repository. GAMBA
aggregates datasets from Loki, NeuReduce, MBA-Obfuscator, QSynth,
Syntia, and MBA-Flatten. The GAMBA repository is licensed under GPLv3;
the original projects' licenses are listed below.

## Original Sources

### SiMBA (GPL-3.0-only)

Repository:  https://github.com/DenuvoSoftwareSolutions/SiMBA
License:     GPL-3.0-only
Publication: SiMBA: Black-box Semantics-aware MBA Deobfuscation

Origin of the E-series expression templates (e1-e5, 2-5 variables).

### MBA-Solver (GPL-3.0-only)

Repository:  https://github.com/softsec-unh/MBA-Solver
License:     GPL-3.0-only
Publication: Boosting SMT Solver Performance on Mixed Bitwise-Arithmetic
             Expressions (PLDI'21)

Origin of the PLDI linear, polynomial, and nonpolynomial datasets.

### MBA-Blast (no license specified)

Repository:  https://github.com/softsec-unh/MBA-Blast
License:     No LICENSE file in repository
Publication: USENIX Security'21

Origin of blast_dataset1.txt and blast_dataset2.txt.

### NeuReduce (no license specified)

Repository:  https://github.com/fvrmatteo/NeuReduce
License:     No LICENSE file in repository
Publication: Semantics-Aware Machine Learning for Function Recognition
             in Binary Executables (EMNLP'20)

Origin of neureduce.txt and test_data.txt.

### Loki (AGPL-3.0-only)

Repository:  https://github.com/RUB-SysSec/loki
License:     AGPL-3.0-only
Publication: Loki: Hardening Code Obfuscation Against Automated Attacks
             (USENIX Security'22)

Origin of loki_tiny.txt.

### MBA-Obfuscator (MIT)

Repository:  https://github.com/nhpcc502/MBA-Obfuscator
License:     MIT
Publication: ICICS'21

Origin of mba_obf_linear.txt and mba_obf_nonlinear.txt.

### QSynth artifacts (no license specified)

Repository:  https://github.com/werew/qsynth-artifacts
License:     No LICENSE file in repository
Publication: QSynth: A Program Synthesis based Approach for Binary Code
             Deobfuscation (BAR'20)

Origin of qsynth_ea.txt and syntia.txt.

### MBA-Flatten (no license specified)

Repository:  https://tinyurl.com/y5l948pu
License:     No license identified
Publication: Security and Communication Networks'22

Origin of mba_flatten.txt.

### OSES (no license specified)

Repository:  https://github.com/fvrmatteo/oracle-synthesis-meets-equality-saturation
License:     No LICENSE file in repository

Origin of oses_fast.txt, oses_slow.txt, oses_all.txt.

### ObfuscatorX (unknown)

License:     Unknown (expressions extracted from a commercial binary)

Origin of obfuscatorx.txt.