Sec4AI4Sec project

Sec4AI4Sec Tutorial on Experimental Design Presented at ICSE and FSE

nikita — Wed, 29 Apr 2026 10:21:07 +0000

Fabio Massacci and Katja Tuma recently delivered the Sec4AI4Sec tutorial on orthogonal and cross-over design with human subjects (or LLMs) at the International […]

L'articolo Sec4AI4Sec Tutorial on Experimental Design Presented at ICSE and FSE proviene da Sec4AI4Sec project.

Paper: Evaluating and Improving the Robustness of Security Attack Detectors Generated by LLMs

nikita — Wed, 22 Apr 2026 12:51:40 +0000

On April 15, 2026, the paper “Evaluating and Improving the Robustness of Security Attack Detectors Generated by LLMs” was presented in Rio de […]

L'articolo Paper: Evaluating and Improving the Robustness of Security Attack Detectors Generated by LLMs proviene da Sec4AI4Sec project.

Paper: Evaluating Retrieval-Augmented Generation for LLM-Based Vulnerability Detection: An Empirical Study on Real-World Java Vulnerabilities

nikita — Wed, 15 Apr 2026 09:30:43 +0000

Software security is becoming increasingly difficult as modern applications grow in size and complexity. Detecting vulnerabilities early before they become real threats is […]

L'articolo Paper: Evaluating Retrieval-Augmented Generation for LLM-Based Vulnerability Detection: An Empirical Study on Real-World Java Vulnerabilities proviene da Sec4AI4Sec project.

Can RAG Make LLMs Better Vulnerability Detectors?

nikita — Wed, 15 Apr 2026 09:13:36 +0000

Software security continues to be one of the most pressing challenges in modern development. As codebases grow in complexity, automating the detection of […]

L'articolo Can RAG Make LLMs Better Vulnerability Detectors? proviene da Sec4AI4Sec project.

Paper: VulTerminator: Bringing Back Template-Based Automated Repair for Fixing Java Vulnerabilities

nikita — Wed, 08 Apr 2026 08:02:54 +0000

Overview: As part of our research efforts in Automated Vulnerability Repair (AVR), we introduce VulTerminator, a novel approach that combines heuristic-based techniques with […]

L'articolo Paper: VulTerminator: Bringing Back Template-Based Automated Repair for Fixing Java Vulnerabilities proviene da Sec4AI4Sec project.

Paper: Domain-aware graph neural networks for source code vulnerability detection

nikita — Tue, 31 Mar 2026 17:03:53 +0000

As software systems become more complex, detecting vulnerabilities early remains a major challenge. Traditional tools often generate too many false positives, while AI […]

L'articolo Paper: Domain-aware graph neural networks for source code vulnerability detection proviene da Sec4AI4Sec project.

If You Can Understand It, You Can Trust It

nikita — Mon, 23 Feb 2026 16:49:49 +0000

The @SEC4AI4SEC initiative promotes trustworthy, transparent, and secure AI systems. As illustrated in our comic, AI decisions can sometimes appear restrictive or overly […]

L'articolo If You Can Understand It, You Can Trust It proviene da Sec4AI4Sec project.

Paper: Leveraging GPT‑4 for Vulnerability‑Witnessing Unit Test Generation

nikita — Fri, 13 Feb 2026 11:27:42 +0000

Creating effective unit tests that expose known software vulnerabilities is a critical yet time‑consuming step in secure software development. Manual test creation requires […]

L'articolo Paper: Leveraging GPT‑4 for Vulnerability‑Witnessing Unit Test Generation proviene da Sec4AI4Sec project.

Paper: σ-zero: Gradient-based Optimization of ℓ0-norm Adversarial Examples

nikita — Fri, 13 Feb 2026 11:07:37 +0000

Generating adversarial examples — small perturbations to inputs that cause machine learning models to misclassify — is a central task in evaluating and […]

L'articolo Paper: σ-zero: Gradient-based Optimization of ℓ0-norm Adversarial Examples proviene da Sec4AI4Sec project.

Paper: AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples

nikita — Fri, 13 Feb 2026 11:03:31 +0000

Evaluating machine learning model robustness against adversarial examples often relies on gradient‑based attacks. However, existing research typically compares these attacks under inconsistent experimental […]

L'articolo Paper: AttackBench: Evaluating Gradient-based Attacks for Adversarial Examples proviene da Sec4AI4Sec project.