On the Privacy of LLMs: An Ablation Study
Fuente:
arXiv
Guardado en:
| Autores principales: | , , , , , , |
|---|---|
| Formato: | Preprint |
| Publicado: |
2026
|
| Materias: | |
| Acceso en línea: | |
| Etiquetas: |
Agregar Etiqueta
Sin Etiquetas, Sea el primero en etiquetar este registro!
|
| _version_ | 1866918480137158656 |
|---|---|
| author | Makhlouf, Karima Basyoni, Lamiaa Khaderi, Syed Marquez, Gabriel Sotomango, Peter Awawdah, Mahmoud Zhioua, Sami |
| author_facet | Makhlouf, Karima Basyoni, Lamiaa Khaderi, Syed Marquez, Gabriel Sotomango, Peter Awawdah, Mahmoud Zhioua, Sami |
| contents | Large language models (LLMs) are increasingly deployed in interactive and retrieval-augmented settings, raising significant privacy concerns. While attacks such as Membership Inference (MIA), Attribute Inference (AIA), Data Extraction (DEA), and Backdoor Attacks (BA) have been studied, they are typically analyzed in isolation, leaving a gap in understanding their behavior under common system factors. In this paper, we introduce a unified threat model and notation, reproduce a representative set of privacy attacks, and conduct a structured ablation study to evaluate the impact of key factors such as model architecture, scale, dataset characteristics, and retrieval configuration. Our analysis reveals clear differences across attack types. Membership inference attacks, particularly mask-based variants, exhibit strong and reliable signals, while backdoor attacks achieve consistently high success rates due to their trigger-based nature. In contrast, attribute inference and data extraction attacks remain more challenging, resulting in lower accuracy, yet they pose significant risks as they target sensitive personal information. Overall, these results highlight that privacy risks in LLM systems are highly context-dependent and driven by design choices, emphasizing the need for holistic evaluation and informed deployment practices. |
| format | Preprint |
| id |
arxiv_https___arxiv_org_abs_2605_02255 |
| institution | arXiv |
| publishDate | 2026 |
| record_format | arxiv |
| spellingShingle | On the Privacy of LLMs: An Ablation Study Makhlouf, Karima Basyoni, Lamiaa Khaderi, Syed Marquez, Gabriel Sotomango, Peter Awawdah, Mahmoud Zhioua, Sami Cryptography and Security Artificial Intelligence Large language models (LLMs) are increasingly deployed in interactive and retrieval-augmented settings, raising significant privacy concerns. While attacks such as Membership Inference (MIA), Attribute Inference (AIA), Data Extraction (DEA), and Backdoor Attacks (BA) have been studied, they are typically analyzed in isolation, leaving a gap in understanding their behavior under common system factors. In this paper, we introduce a unified threat model and notation, reproduce a representative set of privacy attacks, and conduct a structured ablation study to evaluate the impact of key factors such as model architecture, scale, dataset characteristics, and retrieval configuration. Our analysis reveals clear differences across attack types. Membership inference attacks, particularly mask-based variants, exhibit strong and reliable signals, while backdoor attacks achieve consistently high success rates due to their trigger-based nature. In contrast, attribute inference and data extraction attacks remain more challenging, resulting in lower accuracy, yet they pose significant risks as they target sensitive personal information. Overall, these results highlight that privacy risks in LLM systems are highly context-dependent and driven by design choices, emphasizing the need for holistic evaluation and informed deployment practices. |
| title | On the Privacy of LLMs: An Ablation Study |
| topic | Cryptography and Security Artificial Intelligence |
| url | https://arxiv.org/abs/2605.02255 |