Can LLMs get help from other LLMs without revealing private information?

• URL: http://arxiv.org/abs/2404.01041v2
• Date: Tue, 2 Apr 2024 06:49:33 GMT
• Title: Can LLMs get help from other LLMs without revealing private information?
• Authors: Florian Hartmann, Duc-Hieu Tran, Peter Kairouz, Victor Cărbune, Blaise Aguera y Arcas,
• Abstract summary: Cascades are a common type of machine learning system in which a large, remote model can be queried if a local model is not able to label a user's data by itself. We show the feasibility of applying cascade systems in such setups by equipping the local model with privacy-preserving techniques.
• Score: 13.687385548330717
• License: http://arxiv.org/licenses/nonexclusive-distrib/1.0/
• Abstract: Cascades are a common type of machine learning systems in which a large, remote model can be queried if a local model is not able to accurately label a user's data by itself. Serving stacks for large language models (LLMs) increasingly use cascades due to their ability to preserve task performance while dramatically reducing inference costs. However, applying cascade systems in situations where the local model has access to sensitive data constitutes a significant privacy risk for users since such data could be forwarded to the remote model. In this work, we show the feasibility of applying cascade systems in such setups by equipping the local model with privacy-preserving techniques that reduce the risk of leaking private information when querying the remote model. To quantify information leakage in such setups, we introduce two privacy measures. We then propose a system that leverages the recently introduced social learning paradigm in which LLMs collaboratively learn from each other by exchanging natural language. Using this paradigm, we demonstrate on several datasets that our methods minimize the privacy loss while at the same time improving task performance compared to a non-cascade baseline.

Related papers

• CoSteer: Collaborative Decoding-Time Personalization via Local Delta Steering [68.91862701376155]
  CoSteer is a novel collaborative framework that enables decoding-time personalization through localized delta steering.<n>We formulate token-level optimization as an online learning problem, where local delta vectors dynamically adjust the remote LLM's logits.<n>This approach preserves privacy by transmitting only the final steered tokens rather than raw data or intermediate vectors.
  arXiv  Detail & Related papers  (2025-07-07T08:32:29Z)
• Learning Obfuscations Of LLM Embedding Sequences: Stained Glass Transform [1.8749305679160366]
  We introduce the Stained Glass Transform, a learned, sequence dependent transformation of the word embeddings of an AI model.<n>We calculate a-postiori privacy estimates, based on mutual information, and verify the privacy and utility of instances of transformed embeddings.
  arXiv  Detail & Related papers  (2025-06-11T06:56:12Z)
• Model Inversion in Split Learning for Personalized LLMs: New Insights from Information Bottleneck Theory [11.83473842859642]
  This work is the first to identify model inversion attacks in the split learning framework for personalized LLMs. We propose a two-stage attack system in which the first part projects representations into the embedding space, and the second part uses a generative model to recover text from these embeddings.
  arXiv  Detail & Related papers  (2025-01-10T13:47:13Z)
• Robust Utility-Preserving Text Anonymization Based on Large Language Models [80.5266278002083]
  Text anonymization is crucial for sharing sensitive data while maintaining privacy. Existing techniques face the emerging challenges of re-identification attack ability of Large Language Models. This paper proposes a framework composed of three LLM-based components -- a privacy evaluator, a utility evaluator, and an optimization component.
  arXiv  Detail & Related papers  (2024-07-16T14:28:56Z)
• Can LLMs Separate Instructions From Data? And What Do We Even Mean By That? [60.50127555651554]
  Large Language Models (LLMs) show impressive results in numerous practical applications, but they lack essential safety features. This makes them vulnerable to manipulations such as indirect prompt injections and generally unsuitable for safety-critical tasks. We introduce a formal measure for instruction-data separation and an empirical variant that is calculable from a model's outputs.
  arXiv  Detail & Related papers  (2024-03-11T15:48:56Z)
• Rethinking Privacy in Machine Learning Pipelines from an Information Flow Control Perspective [16.487545258246932]
  Modern machine learning systems use models trained on ever-growing corpora. metadata such as ownership, access control, or licensing information is ignored during training. We take an information flow control perspective to describe machine learning systems.
  arXiv  Detail & Related papers  (2023-11-27T13:14:39Z)
• PrivacyMind: Large Language Models Can Be Contextual Privacy Protection Learners [81.571305826793]
  We introduce Contextual Privacy Protection Language Models (PrivacyMind) Our work offers a theoretical analysis for model design and benchmarks various techniques. In particular, instruction tuning with both positive and negative examples stands out as a promising method.
  arXiv  Detail & Related papers  (2023-10-03T22:37:01Z)
• Recovering from Privacy-Preserving Masking with Large Language Models [14.828717714653779]
  We use large language models (LLMs) to suggest substitutes of masked tokens. We show that models trained on the obfuscation corpora are able to achieve comparable performance with the ones trained on the original data.
  arXiv  Detail & Related papers  (2023-09-12T16:39:41Z)
• Privacy Side Channels in Machine Learning Systems [87.53240071195168]
  We introduce privacy side channels: attacks that exploit system-level components to extract private information. For example, we show that deduplicating training data before applying differentially-private training creates a side-channel that completely invalidates any provable privacy guarantees. We further show that systems which block language models from regenerating training data can be exploited to exfiltrate private keys contained in the training set.
  arXiv  Detail & Related papers  (2023-09-11T16:49:05Z)
• Scalable Collaborative Learning via Representation Sharing [53.047460465980144]
  Federated learning (FL) and Split Learning (SL) are two frameworks that enable collaborative learning while keeping the data private (on device) In FL, each data holder trains a model locally and releases it to a central server for aggregation. In SL, the clients must release individual cut-layer activations (smashed data) to the server and wait for its response (during both inference and back propagation). In this work, we present a novel approach for privacy-preserving machine learning, where the clients collaborate via online knowledge distillation using a contrastive loss.
  arXiv  Detail & Related papers  (2022-11-20T10:49:22Z)
• How Much Privacy Does Federated Learning with Secure Aggregation Guarantee? [22.7443077369789]
  Federated learning (FL) has attracted growing interest for enabling privacy-preserving machine learning on data stored at multiple users. While data never leaves users' devices, privacy still cannot be guaranteed since significant computations on users' training data are shared in the form of trained local models. Secure Aggregation (SA) has been developed as a framework to preserve privacy in FL.
  arXiv  Detail & Related papers  (2022-08-03T18:44:17Z)
• Applied Federated Learning: Architectural Design for Robust and Efficient Learning in Privacy Aware Settings [0.8454446648908585]
  The classical machine learning paradigm requires the aggregation of user data in a central location. Centralization of data poses risks, including a heightened risk of internal and external security incidents. Federated learning with differential privacy is designed to avoid the server-side centralization pitfall.
  arXiv  Detail & Related papers  (2022-06-02T00:30:04Z)
• Privacy-Preserving Federated Learning via System Immersion and Random Matrix Encryption [4.258856853258348]
  Federated learning (FL) has emerged as a privacy solution for collaborative distributed learning where clients train AI models directly on their devices instead of sharing their data with a centralized (potentially adversarial) server. We propose a Privacy-Preserving Federated Learning (PPFL) framework built on the synergy of matrix encryption and system immersion tools from control theory. We show that our algorithm provides the same level of accuracy and convergence rate as the standard FL with a negligible cost while revealing no information about clients' data.
  arXiv  Detail & Related papers  (2022-04-05T21:28:59Z)
• Survey: Leakage and Privacy at Inference Time [59.957056214792665]
  Leakage of data from publicly available Machine Learning (ML) models is an area of growing significance. We focus on inference-time leakage, as the most likely scenario for publicly available models. We propose a taxonomy across involuntary and malevolent leakage, available defences, followed by the currently available assessment metrics and applications.
  arXiv  Detail & Related papers  (2021-07-04T12:59:16Z)

This list is automatically generated from the titles and abstracts of the papers in this site.

This site does not guarantee the quality of this site (including all information) and is not responsible for any consequences.