Adversarial Text Generation on CTI
97ASRattention-based fake text generation methodology
Evaluation results
| Method | Links | |||
|---|---|---|---|---|
| attention-based fake text generation methodologyTarget domain=CTI, Attack objective=Pipeline disruption, Generation method=Prompt-based generation, Semantic preservation=Enforced, Identifier-like patterns=Yes (CVE-like), Attack scope=Multi-stage CTI pipeline, Pipeline Impact=Explicitly evaluated2025.07 | 97 | — | 1 | |
| Ren et al.Target domain=CTI, Attack objective=Evasion (model-level), Generation method=LLM fine-tuning, Semantic preservation=Partial, Identifier-like patterns=Limited, Attack scope=Single model, Pipeline Impact=Not considered2025.07 | 87 | — | — | |
| EaTVulTarget domain=CTI, Attack objective=Evasion (model-level), Generation method=LLM + optimization, Semantic preservation=Enforced, Identifier-like patterns=Partial, Attack scope=Single model, Pipeline Impact=Not considered2025.07 | 80 | — | — |