arXiv subjects

George Amariucai

Publications and source records attributed to George Amariucai.

15 recordsLinked to original sources

arXiv2026

Toward Interpretable Privacy Guarantees in Face-Swapping Anonymization

Face-swapping has emerged as a promising approach to facial privacy protection, replacing a target individual's appearance with that of a donor while preserving non-facial context. The resulting images visually resemble the donor, and face recognition systems tend to suppress the target's match scores -- ostensibly satisfying privacy requirements. Empirical evaluation across a range of face-swapping models, however, reveals that significant target identity leakage still occurs. This raises a deeper question: why does leakage occur, and can it be predicted? We propose a linear stochastic model that treats face-swappers as transformations on the space of identity embeddings, providing an interpretable account of the leakage mechanism. The model is fit to empirical observations and used to derive testable predictions. The aim is to ground privacy assessments in principled, interpretable analysis, thus making formal privacy guarantees explainable -- and perfectible -- rather than purely observational.

cs.CR↗

arXiv2025

LLM-Based Community Surveys for Operational Decision Making in Interconnected Utility Infrastructures

We represent interdependent infrastructure systems and communities alike with a hetero-functional graph (HFG) that encodes the dependencies between functionalities. This graph naturally imposes a partial order of functionalities that can inform the sequence of repair decisions to be made during a disaster across affected communities. However, using such technical criteria alone provides limited guidance at the point where the functionalities directly impact the communities, since these can be repaired in any order without violating the system constraints. To address this gap and improve resilience, we integrate community preferences to refine this partial order from the HFG into a total order. Our strategy involves getting the communities' opinions on their preferred sequence for repair crews to address infrastructure issues, considering potential constraints on resources. Due to the delay and cost associated with real-world survey data, we utilize a Large Language Model (LLM) as a proxy survey tool. We use the LLM to craft distinct personas representing individuals, each with varied disaster experiences. We construct diverse disaster scenarios, and each simulated persona provides input on prioritizing infrastructure repair needs across various communities. Finally, we apply learning algorithms to generate a global order based on the aggregated responses from these LLM-generated personas.

cs.SI↗

arXiv2024

Optimizing Privacy and Utility Tradeoffs for Group Interests Through Harmonization

We propose a novel problem formulation to address the privacy-utility tradeoff, specifically when dealing with two distinct user groups characterized by unique sets of private and utility attributes. Unlike previous studies that primarily focus on scenarios where all users share identical private and utility attributes and often rely on auxiliary datasets or manual annotations, we introduce a collaborative data-sharing mechanism between two user groups through a trusted third party. This third party uses adversarial privacy techniques with our proposed data-sharing mechanism to internally sanitize data for both groups and eliminates the need for manual annotation or auxiliary datasets. Our methodology ensures that private attributes cannot be accurately inferred while enabling highly accurate predictions of utility features. Importantly, even if analysts or adversaries possess auxiliary datasets containing raw data, they are unable to accurately deduce private features. Additionally, our data-sharing mechanism is compatible with various existing adversarially trained privacy techniques. We empirically demonstrate the effectiveness of our approach using synthetic and real-world datasets, showcasing its ability to balance the conflicting goals of privacy and utility.

cs.LG↗

arXiv2024

Initial Exploration of Zero-Shot Privacy Utility Tradeoffs in Tabular Data Using GPT-4

We investigate the application of large language models (LLMs), specifically GPT-4, to scenarios involving the tradeoff between privacy and utility in tabular data. Our approach entails prompting GPT-4 by transforming tabular data points into textual format, followed by the inclusion of precise sanitization instructions in a zero-shot manner. The primary objective is to sanitize the tabular data in such a way that it hinders existing machine learning models from accurately inferring private features while allowing models to accurately infer utility-related attributes. We explore various sanitization instructions. Notably, we discover that this relatively simple approach yields performance comparable to more complex adversarial optimization methods used for managing privacy-utility tradeoffs. Furthermore, while the prompts successfully obscure private features from the detection capabilities of existing machine learning models, we observe that this obscuration alone does not necessarily meet a range of fairness metrics. Nevertheless, our research indicates the potential effectiveness of LLMs in adhering to these fairness metrics, with some of our experimental results aligning with those achieved by well-established adversarial optimization techniques.

cs.LG↗

arXiv2024

Improving behavior based authentication against adversarial attack using XAI

In recent years, machine learning models, especially deep neural networks, have been widely used for classification tasks in the security domain. However, these models have been shown to be vulnerable to adversarial manipulation: small changes learned by an adversarial attack model, when applied to the input, can cause significant changes in the output. Most research on adversarial attacks and corresponding defense methods focuses only on scenarios where adversarial samples are directly generated by the attack model. In this study, we explore a more practical scenario in behavior-based authentication, where adversarial samples are collected from the attacker. The generated adversarial samples from the model are replicated by attackers with a certain level of discrepancy. We propose an eXplainable AI (XAI) based defense strategy against adversarial attacks in such scenarios. A feature selector, trained with our method, can be used as a filter in front of the original authenticator. It filters out features that are more vulnerable to adversarial attacks or irrelevant to authentication, while retaining features that are more robust. Through comprehensive experiments, we demonstrate that our XAI based defense strategy is effective against adversarial attacks and outperforms other defense strategies, such as adversarial training and defensive distillation.

cs.CR↗

arXiv2023

A Comprehensive and Reliable Feature Attribution Method: Double-sided Remove and Reconstruct (DoRaR)

The limited transparency of the inner decision-making mechanism in deep neural networks (DNN) and other machine learning (ML) models has hindered their application in several domains. In order to tackle this issue, feature attribution methods have been developed to identify the crucial features that heavily influence decisions made by these black box models. However, many feature attribution methods have inherent downsides. For example, one category of feature attribution methods suffers from the artifacts problem, which feeds out-of-distribution masked inputs directly through the classifier that was originally trained on natural data points. Another category of feature attribution method finds explanations by using jointly trained feature selectors and predictors. While avoiding the artifacts problem, this new category suffers from the Encoding Prediction in the Explanation (EPITE) problem, in which the predictor's decisions rely not on the features, but on the masks that selects those features. As a result, the credibility of attribution results is undermined by these downsides. In this research, we introduce the Double-sided Remove and Reconstruct (DoRaR) feature attribution method based on several improvement methods that addresses these issues. By conducting thorough testing on MNIST, CIFAR10 and our own synthetic dataset, we demonstrate that the DoRaR feature attribution method can effectively bypass the above issues and can aid in training a feature selector that outperforms other state-of-the-art feature attribution methods. Our code is available at https://github.com/dxq21/DoRaR.

cs.LG↗

arXiv2023

Hypersparse Traffic Matrix Construction using GraphBLAS on a DPU

Low-power small form factor data processing units (DPUs) enable offloading and acceleration of a broad range of networking and security services. DPUs have accelerated the transition to programmable networking by enabling the replacement of FPGAs/ASICs in a wide range of network oriented devices. The GraphBLAS sparse matrix graph open standard math library is well-suited for constructing anonymized hypersparse traffic matrices of network traffic which can enable a wide range of network analytics. This paper measures the performance of the GraphBLAS on an ARM based NVIDIA DPU (BlueField 2) and, to the best of our knowledge, represents the first reported GraphBLAS results on a DPU and/or ARM based system. Anonymized hypersparse traffic matrices were constructed at a rate of over 18 million packets per second.

cs.AR↗

arXiv2022

The Economics of Privacy and Utility: Investment Strategies

The inevitable leakage of privacy as a result of unrestrained disclosure of personal information has motivated extensive research on robust privacy-preserving mechanisms. However, existing research is mostly limited to solving the problem in a static setting with disregard for the privacy leakage over time. Unfortunately, this treatment of privacy is insufficient in practical settings where users continuously disclose their personal information over time resulting in an accumulated leakage of the users' sensitive information. In this paper, we consider privacy leakage over a finite time horizon and investigate optimal strategies to maximize the utility of the disclosed data while limiting the finite-horizon privacy leakage. We consider a simple privacy mechanism that involves compressing the user's data before each disclosure to meet the desired constraint on future privacy. We further motivate several algorithms to optimize the dynamic privacy-utility tradeoff and evaluate their performance via extensive synthetic performance tests.

cs.CR↗

arXiv2022

Uncertainty-Autoencoder-Based Privacy and Utility Preserving Data Type Conscious Transformation

We propose an adversarial learning framework that deals with the privacy-utility tradeoff problem under two types of conditions: data-type ignorant, and data-type aware. Under data-type aware conditions, the privacy mechanism provides a one-hot encoding of categorical features, representing exactly one class, while under data-type ignorant conditions the categorical variables are represented by a collection of scores, one for each class. We use a neural network architecture consisting of a generator and a discriminator, where the generator consists of an encoder-decoder pair, and the discriminator consists of an adversary and a utility provider. Unlike previous research considering this kind of architecture, which leverages autoencoders (AEs) without introducing any randomness, or variational autoencoders (VAEs) based on learning latent representations which are then forced into a Gaussian assumption, our proposed technique introduces randomness and removes the Gaussian assumption restriction on the latent variables, only focusing on the end-to-end stochastic mapping of the input to privatized data. We test our framework on different datasets: MNIST, FashionMNIST, UCI Adult, and US Census Demographic Data, providing a wide range of possible private and utility attributes. We use multiple adversaries simultaneously to test our privacy mechanism -- some trained from the ground truth data and some trained from the perturbed data generated by our privacy mechanism. Through comparative analysis, our results demonstrate better privacy and utility guarantees than the existing works under similar, data-type ignorant conditions, even when the latter are considered under their original restrictive single-adversary model.

cs.LG↗

arXiv2021

Review of the Security of Backward-Compatible Automotive Inter-ECU Communication

Advanced electronic units inside modern vehicles have enhanced the driving experience, but also introduced a myriad of security problems due to the inherent limitations of the internal communication protocol. In the last two decades, a number of security threats have been identified and accordingly, security measures have been proposed. While a large body of research on the vehicular security domain is focused on exposing vulnerabilities and proposing counter measures, there is an apparent paucity of research aimed at reviewing existing works on automotive security and at extracting insights. This paper provides a systematic review of security threats and countermeasures for the ubiquitous CAN bus communication protocol. It further exposes the limitations of the existing security measures, and discusses a seemingly-overlooked, simple, cost-effective and incrementally deployable solution which can provide a reasonable defense against a major class of packet injection attacks and many denial of service attacks.

cs.CR↗

arXiv2020

Implicit Crowdsourcing for Identifying Abusive Behavior in Online Social Networks

The increased use of online social networks for the dissemination of information comes with the misuse of the internet for cyberbullying, cybercrime, spam, vandalism, amongst other things. To proactively identify abuse in the networks, we propose a model to identify abusive posts by crowdsourcing. The crowdsourcing part of the detection mechanism is implemented implicitly, by simply observing the natural interaction between users encountering the messages. We explore the node-to-node spread of information on Twitter and propose a model that predicts the abuse level (abusive, hate, spam, normal) associated with the tweet by observing the attributes of the message, along with those of the users interacting with it. We demonstrate that the difference in users' interactions with abusive posts can be leveraged in identifying posts of varying abuse levels.

cs.SI↗

arXiv2020

A Practical Approach to Navigating the Tradeoff Between Privacy and Precise Utility

Due to the recent popularity of online social networks, coupled with people's propensity to disclose personal information in an effort to achieve certain gratifications, the problem of navigating the tradeoff between privacy and utility attracted a lot of recent interest and generated a rich body of research. A critical prerequisite to solving the problem is to appropriately capture the privacy and the utility aspects in the problem formulation. Most of the existing works' focus is on the notion of privacy, while utility loss is often treated as the undesirable but necessary distortion of the true data, introduced by the privacy mechanism. By contrast, we are interested in modelling utility differently, by associating it with specific attributes of a user, just like privacy is associated with specific private attributes in the literature. Our model of utility facilitates a better and more precise privacy mechanism, and achieves better privacy-utility tradeoffs. We further incorporate into our problem formulation a practical constraint on acceptable loss in utility per unit gain in privacy, which allows users to customize the privacy mechanisms in order to account for the relative values that each user associates with their own privacy and utility. This paper discusses the intricacies of our utility model and the corresponding privacy-utility tradeoff, and introduces a heuristic greedy algorithm to solve the problem.

cs.IT↗

arXiv2020

MIDMod-OSN: A Microscopic-level Information Diffusion Model for Online Social Networks

As online social networks continue to be commonly used for the dissemination of information to the public, understanding the phenomena that govern information diffusion is crucial for many security and safety-related applications, such as maximizing information spread and misinformation containment during crises and natural disasters. In this study, we hypothesize that the features that contribute to information diffusion in online social networks are significantly influenced by the type of event being studied. We classify Twitter events as either informative or trending and then explore the node-to-node influence dynamics associated with information spread. We build a model based on Bayesian Logistic Regression for learning and prediction and Random Forests for feature selection. Experimental results from real-world data sets show that the proposed model outperforms state-of-the-art diffusion prediction models, achieving 93% accuracy in informative events and 86% in trending events. We observed that the models for informative and trending events differ significantly, both in the diffusion process and in the user features that govern the diffusion. Our findings show that followers play an important role in the diffusion process and it is possible to use the diffusion and OSN behavior of users for predicting the trending character of a message without having to count the number of reactions.

cs.SI↗

arXiv2020

An Information Diffusion Approach to Rumor Propagation and Identification on Twitter

With the increasing use of online social networks as a source of news and information, the propensity for a rumor to disseminate widely and quickly poses a great concern, especially in disaster situations where users do not have enough time to fact-check posts before making the informed decision to react to a post that appears to be credible. In this study, we explore the propagation pattern of rumors on Twitter by exploring the dynamics of microscopic-level misinformation spread, based on the latent message and user interaction attributes. We perform supervised learning for feature selection and prediction. Experimental results with real-world data sets give the models' prediction accuracy at about 90\% for the diffusion of both True and False topics. Our findings confirm that rumor cascades run deeper and that rumor masked as news, and messages that incite fear, will diffuse faster than other messages. We show that the models for True and False message propagation differ significantly, both in the prediction parameters and in the message features that govern the diffusion. Finally, we show that the diffusion pattern is an important metric in identifying the credibility of a tweet.

cs.SI↗

arXiv2009

Feedback-Based Collaborative Secrecy Encoding over Binary Symmetric Channels

In this paper we propose a feedback scheme for transmitting secret messages between two legitimate parties, over an eavesdropped communication link. Relative to Wyner's traditional encoding scheme \cite{wyner1}, our feedback-based encoding often yields larger rate-equivocation regions and achievable secrecy rates. More importantly, by exploiting the channel randomness inherent in the feedback channels, our scheme achieves a strictly positive secrecy rate even when the eavesdropper's channel is less noisy than the legitimate receiver's channel. All channels are modeled as binary and symmetric (BSC). We demonstrate the versatility of our feedback-based encoding method by using it in three different configurations: the stand-alone configuration, the mixed configuration (when it combines with Wyner's scheme \cite{wyner1}), and the reversed configuration. Depending on the channel conditions, significant improvements over Wyner's secrecy capacity can be observed in all configurations.

cs.IT↗