Gérald Gavin

Inference and parameter estimation on hierarchical belief networks for image segmentation

We introduce a new causal hierarchical belief network for image segmentation. Contrary to classical tree structured (or pyramidal) models, the factor graph of the network contains cycles. Each level of the hierarchical structure features the same number of sites as the base level and each site on a given level has several neighbors on the parent level. Compared to tree structured models, the (spatial) random process on the base level of the model is stationary which avoids known drawbacks, namely visual artifacts in the segmented image. We propose different parameterizations of the conditional probability distributions governing the transitions between the image levels. A parametric distribution depending on a single parameter allows the design of a fast inference algorithm on graph cuts, whereas for arbitrary distributions, we propose inference with loopy belief propagation. The method is evaluated on scanned documents, showing an improvement of character recognition results compared to other methods.

Oblivious Multi-variate Polynomial Evaluation

In this paper, we propose a protocol for Oblivious Polynomial Evaluation (OPE) considering a multi-variate polynomial. There are two parties, Alice who has a secret multi-variate polynomial f and Bob who has an input x = (x 1,...,x T ). Thus, Bob wants to compute f(x) without any information leakage: Alice learns nothing about x and Bob learns only what can be inferred from f(x). In [4], the authors proposed a solution for this problem using Oblivious Transfer (OT) protocol only. In this paper, we propose efficient OPE protocols for the multi-variate case based upon additive and multiplicative homomorphic encryption schemes defined on the same domain. Our protocol only reveals the number of monomials.

Enforcing Privacy in Cloud Databases

Outsourcing databases, i.e., resorting to Database-as-a-Service (DBaaS), is nowadays a popular choice due to the elasticity, availability, scalability and pay-as-you-go features of cloud computing. However, most data are sensitive to some extent, and data privacy remains one of the top concerns to DBaaS users, for obvious legal and competitive reasons.In this paper, we survey the mechanisms that aim at making databases secure in a cloud environment, and discuss current pitfalls and related research challenges.

An Efficient Somewhat Homomorphic Encryption Scheme Based on Factorization

Surprisingly, most of existing provably secure FHE or SWHE schemes are lattice-based constructions. It is legitimate to question whether there is a mysterious link between homomorphic encryptions and lattices. This paper can be seen as a first (partial) negative answer to this question. We propose a very simple private-key (partially) homomorphic encryption scheme whose security relies on factorization. This encryption scheme deals with a secret multivariate rational function \(\phi _D\) defined over \(\mathbb {Z}_n\), n being an RSA-modulus. An encryption of x is simply a vector c such that \(\phi _D(c)=x+\textsf {noise}\). To get homomorphic properties, nonlinear operators are specifically developed. We first prove IND-CPA security in the generic ring model assuming the hardness of factoring. We then extend this model in order to integrate lattice-based cryptanalysis and we reduce the security of our scheme (in this extended model) to an algebraic condition. This condition is extensively discussed for several choices of parameters. Some of these choices lead to competitive performance with respect to other existing homomorphic encryptions. While quantum computers are not only dreams anymore, designing factorization-based cryptographic schemes might appear as irrelevant. But, it is important to notice that, in our scheme, the factorization of n is not required to decrypt. The factoring assumption simply ensures that solving nonlinear equations or finding non-null polynomials with many roots is difficult. Consequently, the ideas behind our construction could be re-used in rings satisfying these properties.

Short-Term Urban Rail Passenger Flow Forecasting: A Dynamic Bayesian Network Approach

We propose a dynamic Bayesian network approach to forecast the short-term passenger flows of the urban rail network of Paris. This approach can deal with the incompleteness of the data caused by failures or lack of collection systems. The structure of the model is based on the causal relationships between the adjacent flows and is designed to take into account the transport service. To reduce the number of arcs and find the maximum likelihood estimate of the parameters, we perform the structural expectation-maximization (EM) algorithm. Then short-term forecasting is conducted by inference, using the bootstrap filter. An experiment is carried out on an entire metro line, using ticket validation, count and transport service data. Overall, the forecasting results outperform historical average and last observation carried forward (LOCF). They illustrate the potential of the approach, as well as the key role of the transport service.

Comparison of two metaheuristics to solve a 2-D cutting stock problem with set-up cost in the paper industry

This paper deals with the two-dimensional cutting stock problem with set-up cost 2CSP-S. This problem is composed of three optimisation sub-problems: a 2-D bin packing 2BP problem to place images on patterns, a linear programming LP problem to find for each pattern the number of stock sheets to be printed and a combinatorial problem to find the number of each image on each pattern. We have already developed two different metaheuristics to solve the 2CSP-S focusing on this third sub-problem: a simulated annealing and a genetic algorithm. In this article, we propose to compare these two approaches. It is important to notice that our approaches are not new packing techniques. This work was conducted for a paper industry company and experiments were realised on real and artificial data sets.

A Genetic Algorithm to Solve a Real 2-D Cutting Stock Problem with Setup Cost in the Paper Industry

This paper deals with the Two-Dimensional Cutting Stock Problem with Setup Cost (2CSP-S). This problem is composed of three optimization sub-problems: a 2-D Bin Packing (2BP) problem (to place images on patterns), a Linear Programming (LP) problem (to find for each pattern the number of stock sheets to be printed) and a combinatorial problem (to find the number of each image on each pattern). In this article, we solve the 2CSP-S focusing on this third sub-problem. A genetic algorithm was developed to automatically find the proper number of each image on patterns. It is important to notice that our approach is not a new packing technique. This work was conducted for a paper industry company and experiments were realized on real and artificial datasets.

Quadratic error minimization in a distributed environment with privacy preserving

In this paper, we address the issue of privacy preserving datamining. Specifically, we consider a scenario where each member j of T parties has its own private database. The party j builds a private classifier hj for predicting a binary class variable y. The aim of this paper consists in aggregating these classifiers hj in order to improve the individual predictions. Precisely, the parties wish to compute an efficient linear combinations over their classifier in a secure manner.