OSS-Forge

OSS-Forge is an open research initiative focused on trustworthy, secure, and transparent AI-assisted software engineering.
We develop and publish:

• static and dynamic analyzers for AI-generated code
• benchmarks and datasets for software vulnerabilities, defects, exploits, and shellcode
• evaluation frameworks for correctness, robustness, and data poisoning
• models and reproducible pipelines for secure code generation
• experimental tools and artifacts from peer-reviewed scientific publications

Our mission is to build a transparent, verifiable, and secure ecosystem for integrating Large Language Models (LLMs) into software development, especially in safety-critical and security-sensitive contexts.

---

What You Will Find Here

This organization hosts resources from multiple research projects and publications in AI security, software engineering, and code generation. Current categories include:

Static Analyzers & Security Tools

• DeVAIC – Fast static analysis for detecting vulnerabilities in Python code
• PatchitPy – Automated patching of vulnerable Python code via pattern-based transformations
• ACCA – Automated correctness assessment of AI-generated code using symbolic execution

Datasets for Security & Software Engineering

• PyResBugs – 5,007 residual Python bugs with NL descriptions
• Shellcode_IA32 – The largest curated dataset of IA-32 shellcode snippets
• PoisonPy – Dataset supporting targeted data-poisoning attacks
• Human vs AI Code – Defects, vulnerabilities, and complexity analysis at scale

Robustness, Data Quality & Industrial Code Generation

• Residual Bug Generation from Natural Language – Frameworks for generating realistic residual defects from NL descriptions
• Impact of Data Quality on Code Models – Empirical studies on robustness, poisoning resilience, and dataset quality
• Industrial Code Generation – Models for domain-specific code synthesis (e.g., VHDL generation from natural language)

Our repositories include code, experimental scripts, datasets, and reproducibility materials.

---

Research Themes

Our work spans four interconnected areas:

1. Security of AI-generated Code
   Vulnerability detection, automated patching, exploit generation, and robustness testing.

2. Trustworthy LLM Evaluation
   Correctness, equivalence checking, symbolic execution, reproducible benchmarks.

3. Software Engineering with AI
   Defect analysis, complexity metrics, orthogonal defect classification (ODC).

4. Adversarial ML for Code Models
   Data poisoning, robustness stress-testing, unsafe pattern injection.

All research artifacts are peer-reviewed and associated with publications at DSN, ISSRE, ICPC, IST, EMSE, JSS, AUSE, and other venues.

---

Publications Powered by These Repositories

A non-exhaustive list includes works presented at:

• IEEE/IFIP DSN
• IEEE ISSRE
• IEEE/ACM ICPC
• Empirical Software Engineering (EMSE)
• Information and Software Technology (IST)
• Automated Software Engineering (AUSE)
• Journal of Systems and Software (JSS)

Full references are available inside each corresponding repository.

---

Contributing

We encourage contributions from the research and practitioner community.

You can contribute by:

• submitting new datasets
• improving static analysis rules
• adding benchmarks or experimental scripts
• reporting issues or proposing new features

Please open discussions or pull requests inside the relevant repository.

---

Contact

OSS-Forge is developed by a joint research team from the University of North Carolina at Charlotte (UNCC) and the University of Naples Federico II.

Scientific Leadership

• Prof. Domenico Cotroneo — UNCC

Core Research Contributors

• Dr. Pietro Liguori — University of Naples Federico II
• Cristina Improta — University of Naples Federico II
• Ph.D. students and graduate researchers and contributors from the DESSERT Research group — University of Naples Federico II

---