No Data? No Problem: Synthesizing Security Graphs for Better Intrusion Detection

TL;DR

This paper introduces PROVSYN, a hybrid framework for synthesizing provenance graphs to improve intrusion detection, especially under data imbalance, by generating high-fidelity graphs and augmenting training data.

Contribution

PROVSYN is a novel hybrid synthesis framework combining graph and textual data generation, significantly enhancing graph fidelity and detection model performance.

Findings

PROVSYN produces higher-fidelity graphs than baselines across five evaluation metrics.

Augmenting datasets with synthesized graphs improves detection accuracy by up to 38%.

Synthetic graphs mitigate data imbalance, increasing normalized entropy by up to 35%.

Abstract

Provenance graph analysis plays a vital role in intrusion detection, particularly against Advanced Persistent Threats (APTs), by exposing complex attack patterns. While recent systems combine graph neural networks (GNNs) with natural language processing (NLP) to capture structural and semantic features, their effectiveness is limited by class imbalance in real-world data. To address this, we introduce PROVSYN, a novel hybrid provenance graph synthesis framework, which comprises three components: (1) graph structure synthesis via heterogeneous graph generation models, (2) textual attribute synthesis via fine-tuned Large Language Models (LLMs), and (3) five-dimensional fidelity evaluation. Experiments on six benchmark datasets demonstrate that PROVSYN consistently produces higher-fidelity graphs across the five evaluation dimensions compared to four strong baselines. To further demonstrate the practical utility of PROVSYN, we utilize the synthesized graphs to augment training datasets for downstream APT detection models. The results show that PROVSYN effectively mitigates data imbalance, improving normalized entropy by up to 35%, and enhances the generalizability of downstream detection models, achieving an accuracy improvement of up to 38%.