Privacy Personalization Trade offs in LLMs: The Impact of Stylometric Signal Reduction on User-Specific Text Generation
Summary
This paper investigates the privacy-personalization trade-off in LLMs by reducing stylistic signals in user-specific text generation, finding that anonymization lowers stylistic fidelity while preserving semantic meaning.
View Cached Full Text
Cached at: 09/22/26, 09:02 AM
# Privacy Personalization Trade offs in LLMs: The Impact of Stylometric Signal Reduction on User-Specific Text Generation Source: [https://arxiv.org/abs/2609.22112](https://arxiv.org/abs/2609.22112) [View PDF](https://arxiv.org/pdf/2609.22112) > Abstract:Large language models \(LLMs\) have demonstrated the ability to generate user\-specific text with high stylistic fidelity\. However, the personal data that enables such personalization frequently embeds demographic, cultural, and stylistic markers that raises concerns about stylometric re\- identification\. This paper investigates whether reducing identifiable stylistic signals affects personalization in text generation by LLMs\. We introduce a controlled framework to isolate stylometric signals in LLM personalization using the LaMP\-7 Twitter benchmark\. Experiments on 250 sampled users compare two settings: paraphrasing conditioned on the original profile and paraphrasing conditioned on an anonymized converted profile in which demographic identifiers, cultural references, personal details, and informal linguistic cues have been systematically neutralized\. Outputs are assessed by two independent LLM judges and a complementary human evaluation\. Our pairwise evaluation shows that outputs conditioned on original profiles are nearly indistinguishable from human\-authored ground truth, indicating that modern LLMs can closely reproduce an author's writing style with sufficient fidelity\. In contrast, preference for model outputs with anonymized profiles drops to 13\.0% on average, while semantic context preservation remains high at 94\.8%\. A study with human evaluators confirms the same pattern\. These findings reveal a clear privacy\-personalization trade\-off and highlight the need for privacy\-aware personalization methods that retain meaning while suppressing identifying stylistic signals\. ## Submission history From: Muhammed Nazmul Arefin \[[view email](https://arxiv.org/show-email/4ebd09df/2609.22112)\] **\[v1\]**Tue, 18 Aug 2026 14:27:06 UTC \(2,030 KB\)
Similar Articles
Demystifying the Privacy-Utility Trade-off in LLM Interactions
This paper analyzes the privacy-utility trade-off in LLM interactions, uncovering underlying mechanisms and introducing an intent-driven local protection framework with a lightweight model to enhance privacy while maintaining response utility.
Can Multi-Agent LLMs Identify Their Peers? Stylometric Fingerprinting in Role-Constrained Political Analysis
This paper investigates whether LLMs can identify their own model family from stylometric fingerprints in role-constrained political analysis texts, even after prompt-level anonymization. The findings confirm that anonymization is insufficient and have implications for EU AI Act compliance and multi-agent system validation.
Small Is Enough: Per-User Style Rewriting of AI-Edited Text via LoRA Adapters
This paper introduces InMyStyle, a privacy-first system that uses LoRA adapters on small language models (0.5B–7B) to rewrite AI-edited text toward an individual user's writing style without explicit prompts. Evaluations show quality plateaus across model sizes, suggesting compact models suffice for this task.
Lightweight Stylistic Consistency Profiling: Robust Detection of LLM-Generated Textual Content for Multimedia Moderation
Proposes LiSCP, a lightweight stylistic consistency profiling method for robust detection of LLM-generated textual content, focusing on feature stability under adversarial manipulation. Achieves superior performance on in-domain and cross-domain detection with notable robustness.
Re-Centering Humans in LLM Personalization
This paper studies the gap between synthetic and human data for evaluating LLM personalization across three stages: attribute extraction, relevance matching, and response generation. Results show models perform worse on real human data, and the authors introduce lightweight training interventions to improve alignment.