Ninghui Li

Kaiyuan Zhang, Zian Su, Pin-Yu Chen et al.

Large Language Model (LLM) agents show considerable promise for automating complex tasks using contextual reasoning; however, interactions involving multiple agents and the system's susceptibility to prompt injection and other forms of context manipulation introduce new vulnerabilities related to privacy leakage and system exploitation. This position paper argues that the well-established design principles in information security, which are commonly referred to as security principles, should be employed when deploying LLM agents at scale. Design principles such as defense-in-depth, least privilege, complete mediation, and psychological acceptability have helped guide the design of mechanisms for securing information systems over the last five decades, and we argue that their explicit and conscientious adoption will help secure agentic systems. To illustrate this approach, we introduce AgentSandbox, a conceptual framework embedding these security principles to provide safeguards throughout an agent's life-cycle. We evaluate with state-of-the-art LLMs along three dimensions: benign utility, attack utility, and attack success rate. AgentSandbox maintains high utility for its intended functions under both benign and adversarial evaluations while substantially mitigating privacy risks. By embedding secure design principles as foundational elements within emerging LLM agent protocols, we aim to promote trustworthy agent ecosystems aligned with user privacy expectations and evolving regulatory requirements.

Kaiyuan Zhang, Siyuan Cheng, Hanxi Guo et al.

Large language models (LLMs) have achieved remarkable success and are widely adopted for diverse applications. However, fine-tuning these models often involves private or sensitive information, raising critical privacy concerns. In this work, we conduct the first comprehensive study evaluating the vulnerability of fine-tuned LLMs to membership inference attacks (MIAs). Our empirical analysis demonstrates that MIAs exploit the loss reduction during fine-tuning, making them highly effective in revealing membership information. These findings motivate the development of our defense. We propose SOFT (\textbf{S}elective data \textbf{O}bfuscation in LLM \textbf{F}ine-\textbf{T}uning), a novel defense technique that mitigates privacy leakage by leveraging influential data selection with an adjustable parameter to balance utility preservation and privacy protection. Our extensive experiments span six diverse domains and multiple LLM architectures and scales. Results show that SOFT effectively reduces privacy risks while maintaining competitive model performance, offering a practical and scalable solution to safeguard sensitive information in fine-tuned LLMs.

Yuntao Du, Zitao Li, Ninghui Li et al.

Large Language Models (LLMs) have achieved remarkable progress in natural language understanding, reasoning, and autonomous decision-making. However, these advancements have also come with significant privacy concerns. While significant research has focused on mitigating the data privacy risks of LLMs during various stages of model training, less attention has been paid to new threats emerging from their deployment. The integration of LLMs into widely used applications and the weaponization of their autonomous abilities have created new privacy vulnerabilities. These vulnerabilities provide opportunities for both inadvertent data leakage and malicious exfiltration from LLM-powered systems. Additionally, adversaries can exploit these systems to launch sophisticated, large-scale privacy attacks, threatening not only individual privacy but also financial security and societal trust. In this paper, we systematically examine these emerging privacy risks of LLMs. We also discuss potential mitigation strategies and call for the research community to broaden its focus beyond data privacy risks, developing new defenses to address the evolving threats posed by increasingly powerful LLMs and LLM-powered systems.

Joshua C. Zhao, Saurabh Bagchi, Salman Avestimehr et al.

Deep learning has shown incredible potential across a wide array of tasks, and accompanied by this growth has been an insatiable appetite for data. However, a large amount of data needed for enabling deep learning is stored on personal devices, and recent concerns on privacy have further highlighted challenges for accessing such data. As a result, federated learning (FL) has emerged as an important privacy-preserving technology that enables collaborative training of machine learning models without the need to send the raw, potentially sensitive, data to a central server. However, the fundamental premise that sending model updates to a server is privacy-preserving only holds if the updates cannot be "reverse engineered" to infer information about the private training data. It has been shown under a wide variety of settings that this privacy premise does not hold. In this survey paper, we provide a comprehensive literature review of the different privacy attacks and defense methods in FL. We identify the current limitations of these attacks and highlight the settings in which the privacy of an FL client can be broken. We further dissect some of the successful industry applications of FL and draw lessons for future successful adoption. We survey the emerging landscape of privacy regulation for FL and conclude with future directions for taking FL toward the cherished goal of generating accurate models while preserving the privacy of the data from its participants.

Aiping Xiong, Chuhao Wu, Tianhao Wang et al.

Proper communication is key to the adoption and implementation of differential privacy (DP). However, a prior study found that laypeople did not understand the data perturbation processes of DP and how DP noise protects their sensitive personal information. Consequently, they distrusted the techniques and chose to opt out of participating. In this project, we designed explanative illustrations of three DP models (Central DP, Local DP, Shuffler DP) to help laypeople conceptualize how random noise is added to protect individuals' privacy and preserve group utility. Following pilot surveys and interview studies, we conducted two online experiments (N = 595) examining participants' comprehension, privacy and utility perception, and data-sharing decisions across the three DP models. Besides the comparisons across the three models, we varied the noise levels of each model. We found that the illustrations can be effective in communicating DP to the participants. Given an adequate comprehension of DP, participants preferred strong privacy protection for a certain type of data usage scenarios (i.e., commercial interests) at both the model level and the noise level. We also obtained empirical evidence showing participants' acceptance of the Shuffler DP model for data privacy protection. Our findings have implications for multiple stakeholders for user-centered deployments of differential privacy, including app developers, DP model developers, data curators, and online users.

Shagufta Mehnaz, Sayanton V. Dibbo, Ehsanul Kabir et al.

Increasing use of machine learning (ML) technologies in privacy-sensitive domains such as medical diagnoses, lifestyle predictions, and business decisions highlights the need to better understand if these ML technologies are introducing leakage of sensitive and proprietary training data. In this paper, we focus on model inversion attacks where the adversary knows non-sensitive attributes about records in the training data and aims to infer the value of a sensitive attribute unknown to the adversary, using only black-box access to the target classification model. We first devise a novel confidence score-based model inversion attribute inference attack that significantly outperforms the state-of-the-art. We then introduce a label-only model inversion attack that relies only on the model's predicted labels but still matches our confidence score-based attack in terms of attack effectiveness. We also extend our attacks to the scenario where some of the other (non-sensitive) attributes of a target record are unknown to the adversary. We evaluate our attacks on two types of machine learning models, decision tree and deep neural network, trained on three real datasets. Moreover, we empirically demonstrate the disparate vulnerability of model inversion attacks, i.e., specific groups in the training dataset (grouped by gender, race, etc.) could be more vulnerable to model inversion attacks.

Shagufta Mehnaz, Ninghui Li, Elisa Bertino

Increasing use of ML technologies in privacy-sensitive domains such as medical diagnoses, lifestyle predictions, and business decisions highlights the need to better understand if these ML technologies are introducing leakages of sensitive and proprietary training data. In this paper, we focus on one kind of model inversion attacks, where the adversary knows non-sensitive attributes about instances in the training data and aims to infer the value of a sensitive attribute unknown to the adversary, using oracle access to the target classification model. We devise two novel model inversion attribute inference attacks -- confidence modeling-based attack and confidence score-based attack, and also extend our attack to the case where some of the other (non-sensitive) attributes are unknown to the adversary. Furthermore, while previous work uses accuracy as the metric to evaluate the effectiveness of attribute inference attacks, we find that accuracy is not informative when the sensitive attribute distribution is unbalanced. We identify two metrics that are better for evaluating attribute inference attacks, namely G-mean and Matthews correlation coefficient (MCC). We evaluate our attacks on two types of machine learning models, decision tree and deep neural network, trained with two real datasets. Experimental results show that our newly proposed attacks significantly outperform the state-of-the-art attacks. Moreover, we empirically show that specific groups in the training dataset (grouped by attributes, e.g., gender, race) could be more vulnerable to model inversion attacks. We also demonstrate that our attacks' performances are not impacted significantly when some of the other (non-sensitive) attributes are also unknown to the adversary.

Yu-Tsung Lee, William Enck, Haining Chen et al.

Android filesystem access control provides a foundation for Android system integrity. Android utilizes a combination of mandatory (e.g., SEAndroid) and discretionary (e.g., UNIX permissions) access control, both to protect the Android platform from Android/OEM services and to protect Android/OEM services from third-party apps. However, OEMs often create vulnerabilities when they introduce market-differentiating features because they err when re-configuring this complex combination of Android policies. In this paper, we propose the PolyScope tool to triage the combination of Android filesystem access control policies to vet releases for vulnerabilities. The PolyScope approach leverages two main insights: (1) adversaries may exploit the coarse granularity of mandatory policies and the flexibility of discretionary policies to increase the permissions available to launch attacks, which we call permission expansion, and (2) system configurations may limit the ways adversaries may use their permissions to launch attacks, motivating computation of attack operations. We apply PolyScope to three Google and five OEM Android releases to compute the attack operations accurately to vet these releases for vulnerabilities, finding that permission expansion increases the permissions available to launch attacks, sometimes by more than 10X, but a significant fraction of these permissions (about 15-20%) are not convertible into attack operations. Using PolyScope, we find two previously unknown vulnerabilities, showing how PolyScope helps OEMs triage the complex combination of access control policies down to attack operations worthy of testing.

Tianhao Wang, Joann Qiongna Chen, Zhikun Zhang et al.

In this paper, we study the problem of publishing a stream of real-valued data satisfying differential privacy (DP). One major challenge is that the maximal possible value can be quite large; thus it is necessary to estimate a threshold so that numbers above it are truncated to reduce the amount of noise that is required to all the data. The estimation must be done based on the data in a private fashion. We develop such a method that uses the Exponential Mechanism with a quality function that approximates well the utility goal while maintaining a low sensitivity. Given the threshold, we then propose a novel online hierarchical method and several post-processing techniques. Building on these ideas, we formalize the steps into a framework for private publishing of stream data. Our framework consists of three components: a threshold optimizer that privately estimates the threshold, a perturber that adds calibrated noises to the stream, and a smoother that improves the result using post-processing. Within our framework, we design an algorithm satisfying the more stringent setting of DP called local DP (LDP). To our knowledge, this is the first LDP algorithm for publishing streaming data. Using four real-world datasets, we demonstrate that our mechanism outperforms the state-of-the-art by a factor of 6-10 orders of magnitude in terms of utility (measured by the mean squared error of answering a random range query).

Aiping Xiong, Tianhao Wang, Ninghui Li et al.

Differential privacy protects an individual's privacy by perturbing data on an aggregated level (DP) or individual level (LDP). We report four online human-subject experiments investigating the effects of using different approaches to communicate differential privacy techniques to laypersons in a health app data collection setting. Experiments 1 and 2 investigated participants' data disclosure decisions for low-sensitive and high-sensitive personal information when given different DP or LDP descriptions. Experiments 3 and 4 uncovered reasons behind participants' data sharing decisions, and examined participants' subjective and objective comprehensions of these DP or LDP descriptions. When shown descriptions that explain the implications instead of the definition/processes of DP or LDP technique, participants demonstrated better comprehension and showed more willingness to share information with LDP than with DP, indicating their understanding of LDP's stronger privacy guarantee compared with DP.

Tianhao Wang, Bolin Ding, Min Xu et al.

When collecting information, local differential privacy (LDP) alleviates privacy concerns of users because their private information is randomized before being sent it to the central aggregator. LDP imposes large amount of noise as each user executes the randomization independently. To address this issue, recent work introduced an intermediate server with the assumption that this intermediate server does not collude with the aggregator. Under this assumption, less noise can be added to achieve the same privacy guarantee as LDP, thus improving utility for the data collection task. This paper investigates this multiple-party setting of LDP. We analyze the system model and identify potential adversaries. We then make two improvements: a new algorithm that achieves a better privacy-utility tradeoff; and a novel protocol that provides better protection against various attacks. Finally, we perform experiments to compare different methods and demonstrate the benefits of using our proposed method.

Tianhao Wang, Ninghui Li, Somesh Jha

The notion of Local Differential Privacy (LDP) enables users to answer sensitive questions while preserving their privacy. The basic LDP frequent oracle protocol enables the aggregator to estimate the frequency of any value. But when the domain of input values is large, finding the most frequent values, also known as the heavy hitters, by estimating the frequencies of all possible values, is computationally infeasible. In this paper, we propose an LDP protocol for identifying heavy hitters. In our proposed protocol, which we call Prefix Extending Method (PEM), users are divided into groups, with each group reporting a prefix of her value. We analyze how to choose optimal parameters for the protocol and identify two design principles for designing LDP protocols with high utility. Experiments on both synthetic and real-world datasets demonstrate the advantage of our proposed protocol.

Tianhao Wang, Jeremiah Blocki, Ninghui Li et al.

Protocols satisfying Local Differential Privacy (LDP) enable parties to collect aggregate information about a population while protecting each user's privacy, without relying on a trusted third party. LDP protocols (such as Google's RAPPOR) have been deployed in real-world scenarios. In these protocols, a user encodes his private information and perturbs the encoded value locally before sending it to an aggregator, who combines values that users contribute to infer statistics about the population. In this paper, we introduce a framework that generalizes several LDP protocols proposed in the literature. Our framework yields a simple and fast aggregation algorithm, whose accuracy can be precisely analyzed. Our in-depth analysis enables us to choose optimal parameters, resulting in two new protocols (i.e., Optimized Unary Encoding and Optimized Local Hashing) that provide better utility than protocols previously proposed. We present precise conditions for when each proposed protocol should be used, and perform experiments that demonstrate the advantage of our proposed protocols.

Dong Su, Jianneng Cao, Ninghui Li et al.

There are two broad approaches for differentially private data analysis. The interactive approach aims at developing customized differentially private algorithms for various data mining tasks. The non-interactive approach aims at developing differentially private algorithms that can output a synopsis of the input dataset, which can then be used to support various data mining tasks. In this paper we study the tradeoff of interactive vs. non-interactive approaches and propose a hybrid approach that combines interactive and non-interactive, using $k$-means clustering as an example. In the hybrid approach to differentially private $k$-means clustering, one first uses a non-interactive mechanism to publish a synopsis of the input dataset, then applies the standard $k$-means clustering algorithm to learn $k$ cluster centroids, and finally uses an interactive approach to further improve these cluster centroids. We analyze the error behavior of both non-interactive and interactive approaches and use such analysis to decide how to allocate privacy budget between the non-interactive step and the interactive step. Results from extensive experiments support our analysis and demonstrate the effectiveness of our approach.

Wahbeh Qardaji, Weining Yang, Ninghui Li

In this paper, we tackle the problem of constructing a differentially private synopsis for two-dimensional datasets such as geospatial datasets. The current state-of-the-art methods work by performing recursive binary partitioning of the data domains, and constructing a hierarchy of partitions. We show that the key challenge in partition-based synopsis methods lies in choosing the right partition granularity to balance the noise error and the non-uniformity error. We study the uniform-grid approach, which applies an equi-width grid of a certain size over the data domain and then issues independent count queries on the grid cells. This method has received no attention in the literature, probably due to the fact that no good method for choosing a grid size was known. Based on an analysis of the two kinds of errors, we propose a method for choosing the grid size. Experimental results validate our method, and show that this approach performs as well as, and often times better than, the state-of-the-art methods. We further introduce a novel adaptive-grid method. The adaptive grid method lays a coarse-grained grid over the dataset, and then further partitions each cell according to its noisy count. Both levels of partitions are then used in answering queries over the dataset. This method exploits the need to have finer granularity partitioning over dense regions and, at the same time, coarse partitioning over sparse regions. Through extensive experiments on real-world datasets, we show that this approach consistently and significantly outperforms the uniform-grid method and other state-of-the-art methods.