Добірка наукової літератури з теми "Modèle random probing"

La contre-mesure de masquage est l'une des contre-mesures les plus puissantes pour contrer les attaques side-channel. Des modèles de fuite ont été exposés pour raisonner théoriquement sur la sécurité de telles implémentations masquées. Jusqu'à présent, le modèle de fuite le plus largement utilisé est le probing model par Ishai, Sahai et Wagner (CRYPTO 2003). Bien qu'il soit avantageux pour les preuves de sécurité, il ne capture pas un adversaire exploitant des traces de fuite complètes, comme dans les attaques horizontales. Pour capturer une classe plus large d'attaques, un autre modèle a été introduit, appelé random probing model. A partir d'un paramètre de fuite p, chaque fil du circuit fuit sa valeur avec probabilité p. Le modèle random probing bénéficie d'une pertinence pratique grâce à une réduction au modèle noisy leakage, qui est admis comme la formalisation appropriée pour les attaques de puissance et électromagnétiques par canaux latéraux. De plus, le modèle random probing est beaucoup plus pratique que le modèle noisy leakage pour prouver la sécurité des schémas de masquage. Pendant ce temps, la communauté ne l'avait pas largement étudié avant cette thèse, et aucune construction pratique n'existait dans la littérature de masquage. Dans cette thèse, nous étudions de plus près le modèle random probing et définissons le premier cadre qui lui est dédié. Nous formalisons une propriété de composition pour les gadgets sécurisés et montrons sa relation avec la notion de non-interférence forte (SNI) utilisée dans le probing model. Nous revisitons ensuite l'idée d'extension proposée par Ananth, Ishai et Sahai (CRYPTO 2018) et introduisons un compilateur qui construit un circuit sécurisé dans le modèle random probing à partir de petits gadgets, obtenant une propriété d'extensibilité de random probing (RPE). Notre construction peut tolérer une probabilité de fuite jusqu'à 2^{-7.09}, contre 2^{-26} pour la construction précédente. On obtient également une meilleure complexité asymptotique de O(K^{7.5}) contre O(K^{7.87}) pour la construction précédente, où kappa est le paramètre de sécurité. Nous proposons ensuite une analyse approfondie de la notion de sécurité RPE. Nous exposons les premières bornes supérieures pour le paramètre principal d'un gadget RPE, connu sous le nom d'ordre d'amplification, et introduisons les premières constructions génériques de gadgets satisfiant RPE et avec des ordres d'amplification presque optimaux. Cela nous permet d'obtenir des instanciations beaucoup plus efficaces du compilateur en expansion. Dans d'autres efforts pour optimiser les constructions pour l'expansion random probing, nous généralisons l'approche RPE en considérant un choix dynamique des gadgets à chaque étape de l'expansion. Cette approche permet d'utiliser des gadgets avec un nombre élevé de "shares" de masquage – qui bénéficient d'une meilleure complexité asymptotique dans le cadre d'expansion – tout en tolérant le meilleur taux de fuite habituellement obtenu pour les petits gadgets. Nous étudions des stratégies pour choisir la séquence de compilateurs et montrons qu'elle peut réduire la complexité de la stratégie d'expansion. Enfin, nous présentons Ironmask, un nouvel outil de vérification polyvalent pour vérifier la sécurité. IronMask est le premier à vérifier les notions de sécurité standard basées sur la simulation dans le modèle probing et les notions récentes de composition et d'extensibilité dans le modèle random probing. Il prend en charge presque tous les gadgets de masquage dans l'état de l'art tout en fournissant des résultats de vérification exacts
The masking countermeasure is among the most potent countermeasures to counteract side-channel attacks. Leakage models have been exhibited to theoretically reason on the security of such masked implementations. So far, the most widely used leakage model is the probing model defined by Ishai, Sahai, and Wagner (CRYPTO 2003). While it is advantageous for security proofs, it does not capture an adversary exploiting full leakage traces, such as in horizontal attacks. To capture a broader class of attacks, another model was introduced, referred to as the random probing model. From a leakage parameter p, each wire of the circuit leaks its value with probability p. The random probing model enjoys practical relevance thanks to a reduction to the noisy leakage model, which is admitted as the suitable formalization for power and electromagnetic side-channel attacks. In addition, the random probing model is much more convenient than the noisy leakage model to prove the security of masking schemes. Meanwhile, the community had not widely studied it before this thesis, and no practical constructions existed in the masking literature. In this thesis, we study more closely the random probing model and define the first framework dedicated to it. We formalize a composition property for secure random probing gadgets and exhibit its relation to the strong non-interference (SNI) notion used in probing security. We then revisit the expansion idea proposed by Ananth, Ishai, and Sahai (CRYPTO 2018) and introduce a compiler that builds a random probing secure circuit from small base gadgets, achieving a random probing expandability (RPE) property. Our construction can tolerate a leakage probability of up to 2^{-7.09}, against 2^{-26} for the previous construction. We also obtain a better asymptotic complexity of O(K^{7.5}) against O(K^{7.87}) for the previous construction, where kappa is the security parameter. We then provide an in-depth analysis of the RPE security notion. We exhibit the first upper bounds for the main parameter of an RPE gadget, known as the amplification order, and introduce the first generic constructions of gadgets achieving RPE for any number of shares and with nearly optimal amplification orders. This allows us to obtain much more efficient instantiations of the expanding compiler. In further efforts to optimize constructions for the random probing expansion, we generalize the RPE approach by considering a dynamic choice of the base gadgets at each step in the expansion. This approach makes it possible to use gadgets with a high number of shares –which enjoy better asymptotic complexity in the expansion framework– while still tolerating the best leakage rate usually obtained for small gadgets. We investigate strategies for choosing the sequence of compilers and show that it can reduce the complexity of the expansion strategy. Finally, we introduce Ironmask, a new versatile verification tool for masking security. IronMask is the first to verify standard simulation-based security notions in the probing model and recent composition and expandability notions in the random probing model. It supports any masking gadgets with linear randomness (eg addition, copy, and refresh gadgets) as well as quadratic gadgets (eg multiplication gadgets) that might include non-linear randomness (eg by refreshing their inputs) while providing complete verification results for both types of gadgets. We report various benchmarks which show that IronMask is competitive with state-of-the-art verification tools in the probing model. IronMask is also several orders of magnitude faster than Vraps --the only previous tool verifying random probing composability and expandability- and Silver --the only previous tool providing complete verification for quadratic gadgets with non-linear randomness

Mestrado em Economia Monetária e Financeira
Este trabalho analisa que fatores, entre os determinantes de ratings soberanos encontrados na literatura, são responsáveis pelas diferenças entre os ratings de crédito soberanos de diferentes agências de rating, no período 1980-2015. Para tal, utilizaram-se modelos probit ordenados e simples de efeitos aleatórios com o objetivo de avaliar o poder explicativo de um conjunto de variáveis macroeconómicas e governamentais. Os resultados obtidos com os modelos estimados indicam que o saldo estrutural e a existência de um default nos últimos dez anos são as variáveis menos significativas enquanto o nível de dívida líquida, o saldo orçamental, o PIB per capita e a existência de um default nos últimos cinco anos são as variáveis que mais explicam as diferenças entre ratings de agências distintas.
In this work we study the factors, among the determinants of sovereign ratings found in the literature, leading to differences in sovereign credit ratings from different agencies, for the period 1980-2015. We employ random effects ordered and simple probit approaches to assess the explanatory power of different macroeconomic and government variables. Our results point to an average performance of the estimated models. Structural balance and the existence of a default in the last ten years were the least significant variables whereas the level of net debt, budget balance, GDP per capita and the existence of a default in the last five years were found to be the most relevant variables explaining the rating differences across agencies.

Mestrado em Econometria Aplicada e Previsão
Através de um modelo dinâmico probit de efeitos aleatórios, estimou-se a probabilidade de desemprego em Portugal de forma a avaliar se existem diferenças entre géneros nos efeitos parciais médios e na persistência do desemprego. Os dados utilizados provêm do Inquérito ao Rendimento e Condições de Vida (ICOR) para o período entre 2010 e 2013. A estimação é feita ao mesmo tempo que se controla pela heterogeneidade individual não observada e pelo problema das condições iniciais, que ocorre pelo fato de não se conhecer o processo estocástico que originou o estado de desemprego observado. Encontrámos forte evidência empírica de persistência do desemprego, e alguma evidência de que esta persistência é mais pronunciada para os homens. Através da inclusão de um efeito fixo especifico para as mulheres, que pretende captar o efeito da discriminação de género num período de instabilidade no mercado de trabalho, concluímos que existe evidência estatística de maior probabilidade de desemprego para as mulheres. Este trabalho tem como principais contributos o estudo dos determinantes da probabilidade de desemprego, que representa uma carência da literatura em economia do trabalho, no fato de o estudar num período de grande desemprego em Portugal, e no especial enfoque que dá à persistência do desemprego e à discriminação de género.
Using a dynamic random effects probit model we estimate the probability of unemployment in Portugal in order to assess gender differences in average partial effects and in unemployment persistence, with data from four waves of the Survey on Income and Living Conditions (ICOR), for the period between 2010 and 2013. The estimation occurs while controlling for unobserved individual heterogeneity and for the "initial conditions" problem, which arises from not knowing the stochastic process which originated the observed state of unemployment. We find strong evidence of persistence in unemployment, with some, although weak, evidence that men suffer more from the negative implications of previous unemployment. Simultaneously, we found evidence of higher probabilities of unemployment for women through a fixed effect that aimed to capture gender discrimination in an unstable labor market. The main contributions of the present work lie in the study of the determinants of the probability of unemployment, which represents a shortage in the current literature in labor economics, during a period of high unemployment in Portugal, and by having a special focus on unemployment persistence and gender discrimination.
info:eu-repo/semantics/publishedVersion

Made available in DSpace on 2016-06-02T20:06:01Z (GMT). No. of bitstreams: 1 2049.pdf: 560086 bytes, checksum: 32b955d6a93e81457f49b0418b1e9514 (MD5) Previous issue date: 2008-06-06
Financiadora de Estudos e Projetos
A summary of the state of the art relative to regression models for binary response variable and panel data is presented in this work. Those models may include efects from several sources: specific variables of interest, heterogeneity between individuals and lagged values of the response variable. The original contributions of the author are simulation studies to compare two diferent approaches to maximum likelihood estimation of parameters of dynamic models with all three kinds of efects, and also a study of properties of such estimators in group sequential analysis, using the bootstrap methodology. Original codes were developed in R for implementation of simulation studies. The relevance of the subject and the non availability of appropriate codes in commercial software for fitting dynamic models for binary response justify the choice of the theme.
Neste trabalho é apresentado inicialmente um levantamento da literatura referente a modelos de regressão não lineares quando a variável resposta é binária e as observações são um painel de dados. Tais modelos podem incluir efeitos de várias fontes: variáveis específicas de interesse, heterogeneidade não observável dos indivíduos e valores defasados da variável resposta. A parte original do trabalho consiste nos estudos por simulação usando programação criada para esse fim no software R, visando comparar duas propostas recentes da literatura para ajustar, por máxima verossimilhança condicional, modelos dinâmicos que incluem os três tipos de efeitos mencionados. Também é original o estudo empírico, usando a metodologia de reamostragem \bootstrap", de características da distribuição conjunta dos estimadores dos parâmetros em análises intermediárias dos dados. A justificativa do trabalho é a atualidade do tema e a inexistência de programas de ajuste de modelos dinâmicos de resposta binária na maioria dos softwares comerciais.

Abstract The logistic-linear model is well established as the basis for analysing regression data, or data from a designed experiment, when the response variable is a proportion. Suppose that the data consist of a set of n cases. In the ith case ri successes, mi - ri failures, and a vector Xi of p explanatory variables are recorded. The logistic-linear model assumes that the ri are realizations of random variables Ri distributed independently Binomial(mi, θi), where θi is related to a linear predictor through the logistic link logit(θi) = log{θi/(l - θi)} = ηi;• An alternative link is the probit, given by probit(θi) = <I>-1(θi) = ηi, where <1>-1 denotes the inverse Normal distribution function.

This chapter covers commonly-used supervised-learning techniques and combinations, with guidance on choices. (1) A view from on high—i) caveats—data dependency of predictive modelling; ii) language—in academic literature for representation of formulae and concepts. (2) Parametric—make assumptions about the relationships between predictors and predicted, and their distributions {linear regression, linear probability modelling (LPM), probit/logit, discriminant analysis, linear programming}. (3) Non-parametric—require few or no assumptions {k-nearest neighbours, Decisions Trees and Random Forests (RF)s, support vector machines, artificial neural networks, genetic algorithms}. (4) Conglomerations—of models and approaches, whether i) practical—for business reasons; ii) parallel—developed using the same data, and then fused; iii) residual—subsequent models predict what prior models could not. Machine learning is introduced. (5) Making the choice—factors affecting the choice {regulatory/compliance, transparency/opacity, suitability to the statistical and business problems, skills availability, longer-term maintenance, development and implementation timeframes, speed of execution once implemented}.

Access to microcredit is very vital in raising firm profitability and productivity, particularly for SMEs; it is a strategy for improved financial inclusion. Thus, the determinants of microcredit access to SMEs from Abia State University (ABSU) Microfinance Bank (MfB) were analyzed. Data generated via snowballing and simple random sampling was analyzed using descriptive statistics and Probit regression model. The socio-economic factors affected access to microcredit. The average amount of microcredit disbursed by the MfB was ₦109,500. Most (75%) invested in agricultural production and retail/petty trading (60%), respectively. The regression coefficient was 0.779. The identified constraints affected access to microcredit from the MfB. This study recommends adoption of measures to facilitate quick approvals and disbursement; improved financial information and trade union activities; provision of checklists and financial advisory services; mitigation of financial bottlenecks; repayment tenure restructuring; low interest rates; and improved financial inclusion.

The chapter applied a willingness to pay ability (WPA) approach to estimate the economic value of drinking water (EVM) using primary data. The sample of 350 households are randomly taken from four wards of Kanpur Nagar, Uttar Pradesh (India). Simple OLS, probit, logit, and mixed-effect logit regression models are applied to examine the effect of social- economic variables on WPA of households to receive better water supply from suppliers. It found that the municipality is charging a very low price to provide water supply of 30 Kilolitre/household per month. However, implicit cost (IC) of water/household is found higher as compared to WPA. IC/household appeared high due to consumption of unsafe drinking water, and incidence of various waterborne health issues and diseases. Family size, occupation, and annual income showed a positive impact on WPA of households.

Abstract The direct measurement of the memory state (i.e. bit at “0” or at “1”) on single magnetic tunnel junction (MTJ) in a commercial magnetic random access memory (MRAM) remains challenging. In this paper, we propose a probing approach to investigate the MTJ resistance and by this way determine the memory state. To reach this goal, the MRAM device needs to be prepared to create an electrical access to both sides of the MTJs. The suitable methodology consists in a backside preparation routine that creates a bevel allowing us to access the bottom side of the MTJs through vias and the top side to the bitlines. After that, two approaches are discussed to establish the electrical connection. First described is the nanoprobing technique where the electrical connection is created by two nanometric tips positioned in contact on vias and bitlines thanks to a scanning electron microscope. It is then possible to collect the current flowing through the MTJs and to evaluate the resistance. A resistance around 12 kΩ and 14 kΩ were determined for “0” and “1” bits respectively, which is in agreement with literature. Secondly, these measurements will be compared to those resulting from a near-field probing experiment done in a conductive mode. A resistance around 19 kΩ and 24 kΩ were determined for “0” and “1” bits respectively. The use of both methods allows for a cross-reference between the resistance values and a discussion on the advantages and drawbacks of both probing techniques.

Poultry farmers are faced with many problems that reduce their capacity utilisation. The challenges have made a lot of poultry entrepreneurs to leave the business compellingly. Many interventions have been applied to solve production problems of the poultry farmers but they have not given the desired results. Also, literatures on poultry production in Nigeria are many and growing focusing mainly on production, economic and marketing research questions. Few have linked the socio-economic factors of poultry farmers with production management and challenges. The research, therefore, analysed the effects of socio-economic characteristics of poultry farmers on selected problems of poultry farming. Two-stage sampling procedure was adopted. At the first stage, three states: Ekiti, Lagos and Ondo were purposefully selected and at the second stage, simple random sampling was used to select 320 poultry farmers in the three states out of which 307 were successful for analysis. Data were collected on 10 socio-economic characteristics and 5 production problems: adverse effect of climate change, disease incidence, high cost of transportation, high feeds cost and high cost of drugs. Frequency distribution, percentage and Probit model were used to analyse data. 76.55%, 63.84%, 68.73% and 67.10% of the farmers were male, married, into farming as main occupation and belonged to cooperative in that order. Poultry farmers using battery cage system have higher likelihood, 0.497, of facing adverse effect of climate change than the farmers using deep litter system while those in Ekiti State have lower likelihood (-0.506) of facing the problem of climate change than those in Ondo and Lagos States. The likelihood of having disease incidence declines with membership of cooperative and increase in years of experience among the poultry farmers with the coefficients of -0.705 and -0.035 respectively. Poultry farmers that are members of cooperative are less likely to have high transportation cost and high cost of feeds as problems from the coefficients of -0.525 and -0.726 correspondingly. Though the probability of having high cost of drugs as production problem increases with age of farmers, older farmers are less likely to have the problem and farmers that have poultry production as main occupation have higher probability of having high cost of drugs than those that are into poultry farming as secondary occupation with the coefficients of 0.156, -0.002 and 0.402 respectively. The study recommends that deep litter system of production should be encouraged among poultry farmers to reduce the adverse effects of climate change. Cooperative membership should be encouraged among the poultry farmers to reduce transportation cost and disease incidence on farms. Also, poultry drug related assistance should be given to the farmers that engage in the enterprise as main occupation which demands the identification of poultry farmers that are into the enterprise as main occupation. Key words: Climate change, Cost of feeds, Cost of drug, Cost of transportation, Poultry farmer, Southwest Nigeria

Nearly 5,000 people are killed and more than 418,000 are injured in weather-related traffic incidents each year. Assessments of the effectiveness of statistical models applied to crash severity prediction compared to machine learning (ML) and deep learning techniques (DL) help researchers and practitioners know what models are most effective under specific conditions. Given the class imbalance in crash data, the synthetic minority over-sampling technique for nominal (SMOTE-N) data was employed to generate synthetic samples for the minority class. The ordered logit model (OLM) and the ordered probit model (OPM) were evaluated as statistical models, while random forest (RF) and XGBoost were evaluated as ML models. For DL, multi-layer perceptron (MLP) and TabNet were evaluated. The performance of these models varied across severity levels, with property damage only (PDO) predictions performing the best and severe injury predictions performing the worst. The TabNet model performed best in predicting severe injury and PDO crashes, while RF was the most effective in predicting moderate injury crashes. However, all models struggled with severe injury classification, indicating the potential need for model refinement and exploration of other techniques. Hence, the choice of model depends on the specific application and the relative costs of false negatives and false positives. This conclusion underscores the need for further research in this area to improve the prediction accuracy of severe and moderate injury incidents, ultimately improving available data that can be used to increase road safety.

Active transportation can play an important role in promoting more physically active and positive public health outcomes. While walking and biking provide significant physical health benefits, their modal share remains low. As a new form of micromobility service, shared e-scooters can enhance the suite of options available in cities to promote active transportation and fill in the gaps when walking or biking are not preferred. Although e-scooters show potential as a mode of transportation, it is unclear whether people will adopt the technology for everyday use. Furthermore, shared micromobility (e.g., electric scooters) is gaining attention as a complementary mode to public transit and is expected to offer a solution to access/egress for public transit. However, few studies have analyzed integrated usage of shared e-scooters and public transit systems while using panel data to measure spatial and temporal characteristics. This study aims to examine the adoption and frequency of shared e-scooter usage and provide policy implementation. To do so, the researchers launched a survey in the Chicago region in late 2020 and collected a rich data set that includes residents’ sociodemographic details and frequency of shared e-scooter use. To characterize the frequency, the researchers used an ordered probit structure. The findings show that respondents who are male, low income, Millennials and Generation Z, or do not have a vehicle are associated with a higher frequency of shared e-scooter use. Furthermore, this study utilizes shared e-scooter trips for a 35-day measurement period from 10 shared e-scooter operators in Chicago, where the researchers used a random-parameter negative binomial modeling approach to analyze panel effects. The findings highlight the critical role of spatial and temporal characteristics in the integration of shared e-scooters with transit.