arXiv subjects

George T. Amariucai

Publications and source records attributed to George T. Amariucai.

8 recordsLinked to original sources

arXiv2026

Picture the Epsilon: Pursuing Identity-Level Privacy Guarantees for Images

Several methods for auditing privacy in embedding spaces report a number called "epsilon." The common name is misleading: one number may be a heuristic score, another may come from a valid population inequality but ignore sampling uncertainty, and a third may be a confidence bound. This paper asks when such a number is evidence about differential privacy. We study four approaches based on Gaussian calibration, marginal kernel-density ratios, maximum mean discrepancy (MMD), and classifier hypothesis tests. The first two are modeling diagnostics. The MMD and classifier approaches use valid population lower bounds, but only the classifier approach is given a separate test set and a finite-sample confidence calculation. To see how these distinctions matter, we build a synthetic benchmark in which the true privacy value is known. It includes pure-DP Laplace mechanisms, approximate-DP Gaussian mechanisms, an exact privacy null, two identity geometries, and three sample sizes. At the null, the Gaussian and kernel-density diagnostics remain large. A direct conversion of an empirical ROC curve often returns infinity even though the AUC is near chance and no threshold separates the samples perfectly. The MMD value increases as the distributions become easier to distinguish, but the sample estimate is small relative to the known reference and is not a lower confidence bound. By contrast, a classifier chosen on development data and evaluated on untouched test data gives simultaneous lower confidence bounds under the stated iid model. We also apply the methods to FaceFusion and InstantID. That case study shows how the methods behave on face data but does not calibrate them because the generators have no known privacy value. The main lesson is that an epsilon-like number is meaningful only together with its assumptions and its finite-sample interpretation.

cs.CR↗

arXiv2017

Layered Synthesis of Latent Gaussian Trees

A new synthesis scheme is proposed to generate a random vector with prescribed joint density that induces a (latent) Gaussian tree structure. The quality of synthesis is shown by vanishing total variation distance between the synthesized and desired statistics. The proposed layered and successive synthesis scheme relies on the learned structure of tree to use sufficient number of common random variables to synthesize the desired density. We characterize the achievable rate region for the rate tuples of multi-layer latent Gaussian tree, through which the number of bits needed to synthesize such Gaussian joint density are determined. The random sources used in our algorithm are the latent variables at the top layer of tree, the additive independent Gaussian noises, and the Bernoulli sign inputs that capture the ambiguity of correlation signs between the variables. We have shown that such ambiguity can further help in reducing the synthesis rates for the underlying Gaussian trees.

cs.IT↗

arXiv2017

Tree Structured Synthesis of Gaussian Trees

A new synthesis scheme is proposed to effectively generate a random vector with prescribed joint density that induces a (latent) Gaussian tree structure. The quality of synthesis is measured by total variation distance between the synthesized and desired statistics. The proposed layered and successive encoding scheme relies on the learned structure of tree to use minimal number of common random variables to synthesize the desired density. We characterize the achievable rate region for the rate tuples of multi-layer latent Gaussian tree, through which the number of bits needed to simulate such Gaussian joint density are determined. The random sources used in our algorithm are the latent variables at the top layer of tree, the additive independent Gaussian noises, and the Bernoulli sign inputs that capture the ambiguity of correlation signs between the variables.

cs.IT↗

arXiv2016

Synthesis of Gaussian Trees with Correlation Sign Ambiguity: An Information Theoretic Approach

In latent Gaussian trees the pairwise correlation signs between the variables are intrinsically unrecoverable. Such information is vital since it completely determines the direction in which two variables are associated. In this work, we resort to information theoretical approaches to achieve two fundamental goals: First, we quantify the amount of information loss due to unrecoverable sign information. Second, we show the importance of such information in determining the maximum achievable rate region, in which the observed output vector can be synthesized, given its probability density function. In particular, we model the graphical model as a communication channel and propose a new layered encoding framework to synthesize observed data using upper layer Gaussian inputs and independent Bernoulli correlation sign inputs from each layer. We find the achievable rate region for the rate tuples of multi-layer latent Gaussian messages to synthesize the desired observables.

cs.IT↗

arXiv2010

Half-Duplex Active Eavesdropping in Fast Fading Channels: A Block-Markov Wyner Secrecy Encoding Scheme

In this paper we study the problem of half-duplex active eavesdropping in fast fading channels. The active eavesdropper is a more powerful adversary than the classical eavesdropper. It can choose between two functional modes: eavesdropping the transmission between the legitimate parties (Ex mode), and jamming it (Jx mode) -- the active eavesdropper cannot function in full duplex mode. We consider a conservative scenario, when the active eavesdropper can choose its strategy based on the legitimate transmitter-receiver pair's strategy -- and thus the transmitter and legitimate receiver have to plan for the worst. We show that conventional physical-layer secrecy approaches perform poorly (if at all), and we introduce a novel encoding scheme, based on very limited and unsecured feedback -- the Block-Markov Wyner (BMW) encoding scheme -- which outperforms any schemes currently available.

cs.IT↗

arXiv2009

Optimal Strategies for Gaussian Jamming in Block-Fading Channels under Delay and Power Constraints

Without assuming any knowledge on source's codebook and its output signals, we formulate a Gaussian jamming problem in block fading channels as a two-player zero sum game. The outage probability is adopted as an objective function, over which transmitter aims at minimization and jammer aims at maximization by selecting their power control strategies. Optimal power control strategies for each player are obtained under both short-term and long-term power constraints. For the latter case, we first prove the non-existence of a Nash equilibrium, and then provide a complete solution for both maxmin and minimax problems. Numerical results demonstrate a sharp difference between the outage probabilities of the minimax and maxmin solutions.

cs.IT↗

arXiv2008

Jamming in Fixed-Rate Wireless Systems with Power Constraints - Part II: Parallel Slow Fading Channels

This is the second part of a two-part paper that studies the problem of jamming in a fixed-rate transmission system with fading. In the first part, we studied the scenario with a fast fading channel, and found Nash equilibria of mixed strategies for short term power constraints, and for average power constraints with and without channel state information (CSI) feedback. We also solved the equally important maximin and minimax problems with pure strategies. Whenever we dealt with average power constraints, we decomposed the problem into two levels of power control, which we solved individually. In this second part of the paper, we study the scenario with a parallel, slow fading channel, which usually models multi-carrier transmissions, such as OFDM. Although the framework is similar as the one in Part I \cite{myself3}, dealing with the slow fading requires more intricate techniques. Unlike in the fast fading scenario, where the frames supporting the transmission of the codewords were equivalent and completely characterized by the channel statistics, in our present scenario the frames are unique, and characterized by a specific set of channel realizations. This leads to more involved inter-frame power allocation strategies, and in some cases even to the need for a third level of power control. We also show that for parallel slow fading channels, the CSI feedback helps in the battle against jamming, as evidenced by the significant degradation to system performance when CSI is not sent back. We expect this degradation to decrease as the number of parallel channels $M$ increases, until it becomes marginal for $M\to \infty$ (which can be considered as the case in Part I).

cs.IT↗

arXiv2008

Jamming in Fixed-Rate Wireless Systems with Power Constraints - Part I: Fast Fading Channels

This is the first part of a two-part paper that studies the problem of jamming in a fixed-rate transmission system with fading. Both transmitter and jammer are subject to power constraints which can be enforced over each codeword short-term / peak) or over all codewords (long-term / average), hence generating different scenarios. All our jamming problems are formulated as zero-sum games, having the probability of outage as pay-off function and power control functions as strategies. The paper aims at providing a comprehensive coverage of these problems, under fast and slow fading, peak and average power constraints, pure and mixed strategies, with and without channel state information (CSI) feedback. In this first part we study the fast fading scenario. We first assume full CSI to be available to all parties. For peak power constraints, a Nash equilibrium of pure strategies is found. For average power constraints, both pure and mixed strategies are investigated. With pure strategies, we derive the optimal power control functions for both intra-frame and inter-frame power allocation. Maximin and minimax solutions are found and shown to be different, which implies the non-existence of a saddle point. In addition we provide alternative perspectives in obtaining the optimal intra-frame power control functions under the long-term power constraints. With mixed strategies, the Nash equilibrium is found by solving the generalized form of an older problem dating back to Bell and Cover \cite{bell}. Finally, we derive a Nash equilibrium of the game in which no CSI is fed back from the receiver. We show that full channel state information brings only a very slight improvement in the system's performance.

cs.IT↗