In this paper, the fundamental insecurities hampering a scalable, wide-spread deployment of biometric authentication are examined, and cryptosystem capable using fingerprint data as its key is presented. For our application, we focus on situations where private stored smartcard used for in networked environment, assume an attacker can launch o -line attacks against stolen card.Juels Sudan's fuzzy vault starting point building analyzing secure scheme fingerprints smartcards called figerprint...

We propose a graphical model for representing networks of stochastic processes, the minimal generative graph. It is based on reduced factorizations joint distribution over time. show that under appropriate conditions, it unique and consistent with another type model, directed information graph, which generalization Granger causality. demonstrate how quantifies causality in particular sequential prediction setting. also develop efficient methods to estimate topological structure from data...

Explicit nonasymptotic upper bounds on the sizes of multiple-deletion correcting codes are presented. In particular, largest single-deletion code for q-ary alphabet and string length is shown to be size at most (q <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">n</sup> -q)/{(q-1)(n-1)}. An improved bound asymptotic rate function obtained as a corollary. Upper also derived constrained source that does not necessarily comprise all strings...

We consider the problem of perfectly recovering vertex correspondence between two correlated Erdos-Renyi (ER) graphs. For a pair graphs on same set, vertices can be obscured by randomly permuting labels one In some cases, structural information in allow this to recovered. investigate information-theoretic threshold for exact recovery, i.e. conditions under which entire correctly recovered given unbounded computational resources. Pedarsani and Grossglauser provided an achievability result...

We introduce the problem of best arm identification (BAI) with post-action context, a new BAI in stochastic multi-armed bandit environment and fixed-confidence setting. The addresses scenarios which learner receives $\textit{post-action context}$ addition to reward after playing each action. This context provides additional information that can significantly facilitate decision process. analyze two different types context: (i) $\textit{non-separator}$, where depends on both action (ii)...

Recent work has shown that traffic analysis of data carried on encrypted tunnels can be used to recover important semantic information. As one example, attackers find out which website, or page a user is accessing simply by monitoring the patterns. We show much greater threat privacy than previously thought, as such attacks remotely. In particular, we that, perform analysis, adversaries do not need directly observe Instead, they send probes from far-off vantage point exploit queuing side...

Digital fingerprinting is a framework for marking media files, such as images, music, or movies, with user-specific signatures to deter illegal distribution. Multiple users can collude produce forgery that potentially overcome system. This paper proposes an equiangular tight frame fingerprint design which robust collusion attacks. We motivate this by considering digital in terms of compressed sensing. The attack modeled linear averaging multiple marked copies before adding Gaussian noise...

We consider the problem of perfectly recovering vertex correspondence between two correlated Erdos-Renyi (ER) graphs. For a pair graphs on same set, vertices can be obscured by randomly permuting labels one In some cases, structural information in allow this to recovered. investigate information-theoretic threshold for exact recovery, i.e. conditions under which entire correctly recovered given unbounded computational resources. Pedarsani and Grossglauser provided an achievability result...

Automated decision making systems are increasingly being used in real-world applications. In these for the most part, rules derived by minimizing training error on available historical data. Therefore, if there is a bias related to sensitive attribute such as gender, race, religion, etc. data, say, due cultural/historical discriminatory practices against certain demographic, system could continue discrimination decisions including said its rule. We present an information theoretic framework...

Nonconvex minimax problems appear frequently in emerging machine learning applications, such as generative adversarial networks and learning. Simple algorithms the gradient descent ascent (GDA) are common practice for solving these nonconvex games receive lots of empirical success. Yet, it is known that vanilla GDA with constant step size can potentially diverge even convex setting. In this work, we show a subclass nonconvex-nonconcave objectives satisfying so-called two-sided...

We consider deletion correcting codes over a q-ary alphabet. It is well known that any code capable of s deletions can also correct combination total insertions and deletions. To obtain asymptotic upper bounds on size, we apply packing argument to channels perform different mixtures Even though the set identical for all these channels, vary. Prior this paper, only corresponding all-insertion case all-deletion were known. recover as special cases. The bound from case, due Levenshtein, has...

We demonstrate the presence of a novel scheduler side-channel in preemptive, fixed-priority real-time systems (RTS); examples such can be found automotive systems, avionic power plants and industrial control among others. This leak important timing information as future arrival times tasks.This then used to launch devastating attacks, two which are demonstrated here (on real hardware platforms). Note that it is not easy capture this due runtime variations schedules, multiple other tasks...

Linking network flows is an important problem in intrusion detection as well anonymity. Passive traffic analysis can link flows, but requires long periods of observation to reduce errors. Active analysis, also known flow watermarking, allows for better precision and more scalable. Previous watermarks introduce significant delays the a side effect using blind scheme; this enables attacks that detect remove watermark, while at same time slowing down legitimate traffic. We propose first...

Recently, directed information graphs have been proposed as concise graphical representations of the statistical dynamics among multiple random processes. A edge from one node to another indicates that past process statistically affects future another, given all other When number processes is large, computing those conditional dependence tests becomes difficult. Also, when interactions too graph no longer facilitates visual extraction relevant for decision-making. This work considers...

This paper presents the design and implementation of spyware communication circuits built into widely used carrier sense multiple access with collision avoidance (CSMA/CA) protocol. The components are embedded within sequential combinational circuit structure during synthesis, rendering distinction or dissociation from original impossible. We take advantage timing channel resulting transmission packets to implement a new practical coding scheme that covertly transfers spied data. Our codes...

We consider the problem of aligning a pair databases with correlated entries. introduce new measure correlation in joint distribution that we call cycle mutual information. This has operational significance: it determines whether exact recovery correspondence between database entries is possible for any algorithm. Additionally, there an efficient algorithm alignment achieves this information theoretic threshold.

We propose an approach for learning latent directed polytrees as long there exists appropriately defined discrepancy measure between the observed nodes. Specifically, we use our information where samples are available from only a subset of processes. Directed trees new type probabilistic graphical models that represent causal dynamics among set random processes in stochastic system. prove is consistent minimal trees. analyze sample complexity task when empirical estimator mutual used measure.

Learning the influence structure of multiple time series data is great interest to many disciplines. This paper studies problem recovering causal in network multivariate linear Hawkes processes. In such processes, occurrence an event one process affects probability new events some other Thus, a natural notion causality exists between processes captured by support excitation matrix. We show that resulting equivalent Directed Information graph (DIG) which encodes factorization joint...