Les attaques par canaux auxiliaires sont apparues dans la deuxième moitié des années 1990. Ces attaques exploitent différentes informations qu’il est possible de collecter lors de l’exécution d’un algorithme sur un appareil cryptographique. Il est ainsi possible, entre autres, de mesurer la consommation d’énergie d’un appareil cryptographique, ou encore d’observer le temps d’exécution d’un certain algorithme sur un appareil. C’est à ces deux sources d’in- formation que nous nous intéressons dans ce travail. Après une présentation des concepts utiles à la lecture du travail et de l’état de l’art des attaques et des contre-mesures du domaine, nous abordons les résultats de nos recherches effectuées lors de ce travail de thèse. Nous présentons d’abord nos contributions aux attaques par mesure de consommation d’énergie :(1) une approche com- binant apprentissage semi-supervisé et attaques par templates pour retrouver le poids de Hamming des différents bytes d’une clé de chiffrement et (2) une approche utilisant des techniques d’apprentissage automatique pour attaquer une implantation protégée d’AES. Ensuite, nous abordons les contre-mesures investiguées durant nos recherches qui se résument (1) en la possibilité de rendre l’ordre des instructions d’AES le plus aléatoire possible en jouant sur la relation de dépendance entre celles-ci ainsi qu’en (2) l’étude de l’application partielle (sur un sous-ensemble de données) de certaines contre-mesures, afin de protéger les données sensibles d’un algorithme. Enfin, nous terminons ce travail par l’emploi de la programmation orientée aspects comme manière d’implanter des contre-mesures pour les attaques temporelles (sur RSA) et pour les attaques par mesures de consommation d’énergie (sur AES). / Doctorat en Sciences / info:eu-repo/semantics/nonPublished

We often have to face the fact that several signals are mixed together in unknown environment. The signals must be first extracted from the mixture in order to interpret them correctly. This problem is in signal processing society called blind source separation. This dissertation thesis deals with multi-channel separation of audio signals in real environment, when the source signals outnumber the sensors. An introduction to blind source separation is presented in the first part of the thesis. The present state of separation methods is then analyzed. Based on this knowledge, the separation systems implementing fuzzy time-frequency mask are introduced. However these methods are still introducing nonlinear changes in the signal spectra, which can yield in musical noise. In order to reduce musical noise, novel methods combining time-frequency binary masking and beamforming are introduced. The new separation system performs linear spatial filtering even if the source signals outnumber the sensors. Finally, the separation systems are evaluated by objective and subjective tests in the last part of the thesis.

OBJECTIVE: To determine in an interrupted word paradigm (Maryland CNCs; 10 ips, 50% duty cycle) if different locations of the interruption pattern produce different recognition performances. DESIGN: Repeated measures using two interruption patterns that were complementary halves referenced to word onset; one started with word onset (0-ms), the other started 50 ms later. The hypothesis was that recognition performance would be better on the 0-ms condition than on the 50-ms condition, but there would be some words with the reverse relation. STUDY SAMPLE: Twenty-four young adults with normal hearing for pure tones and 32 older adults (mean = 67 years) with sensorineural hearing loss participated. RESULTS: With the young listeners mean recognition performance on the 0-ms condition (63.1%) was significantly better than the mean performance on the 50-ms condition (47.8%). About twice as many words had better performance on the 0-ms condition. With the older listeners, who were given only stimuli on which performances were > 58% by the young normals, performances on the two conditions were the same. CONCLUSIONS: The hypothesis was supported with the young listeners. The equal performance by the older listeners on the two conditions was attributed to the manner in which the words were selected.

在迴歸分析中，當資料中存在很多離群值時，偵測的工作變得非常不容易。 在此狀況下，我們無法使用傳統的殘差分析正確地偵測出其是否存在，此現象稱為遮蔽效應(The Masking Effect)。 而為了避免此效應的發生，我們利用最小中位數穩健迴歸估計值(Least Median Squares Estimator)正確地找出這些群集離群值，此估計值擁有最大即50﹪的容離值 (Breakdown point)。 在這篇論文中，用來求出最小中位數穩健迴歸估計值的演算法稱為步進搜尋演算法 (the Forward Search Algorithm)。 結果顯示，我們可以利用此演算法得到的穩健迴歸估計值，很快並有效率的找出資料中的群集離群值；另外，更進一步的結果顯示，我們只需從資料中隨機選取一百次子集，並進行步進搜尋，即可得到概似的穩健迴歸估計值並正確的找出那些群集離群值。 最後，我們利用鐘乳石圖(Stalactite Plot)列出所有被偵測到的離群值。 在多變量資料中，我們若使用Mahalanobis距離也會遭遇到同樣的屏蔽效應。 而此一問題，隨著另一高度穩健估計值的採用，亦可迎刃而解。 此估計值稱為最小體積橢圓體估計值 (Minimum Volume Ellipsoid)，其亦擁有最大即50﹪的容離值。 在此，我們也利用步進搜尋法求出此估計值，並利用鐘乳石圖列出所有被偵測到的離群值。 這篇論文的第二部分則利用變數轉換的技巧將迴歸資料中的殘差項常態化並且加強其等變異的特性以利後續的資料分析。 在步進搜尋進行的過程中，我們觀察分數統計量(Score Statistic)和其他相關診斷統計量的變化。 結果顯示，這些統計量一起提供了有關轉換參數選取豐富的資訊，並且我們亦可從步進搜尋進行的過程中觀察出某些離群值對參數選取的影響。 / Detecting regression outliers is not trivial when there are many of them. The methods of using classical diagnostic plots sometimes fail to detect them. This phenomenon is known as the masking effect. To avoid this, we propose to find out those multiple outliers by using a highly robust regression estimator called the least median squares (LMS) estimator which has maximal breakdown point. The algorithm in search of the LMS estimator is called the forward search algorithm. The estimator found by the forward search is shown to lead to the rapid detection of multiple outliers. Furthermore, the result reveals that 100 repeats of a simple forward search from a random starting subset are shown to provide sufficiently robust parameter estimators to reveal multiple outliers. Finally, those detected outliers are exhibited by the stalactite plot that shows greatly stable pattern of them. Referring to multivariate data, the Mahalanobis distance also suffers from the masking effect that can be remedied by using a highly robust estimator called the minimum volume ellipsoid (MVE) estimator. It can also be found by using the forward search algorithm and it also has maximal breakdown point. The detected outliers are then displayed in the stalactite plot. The second part of this dissertation is the transformation of regression data so that the approximate normality and the homogeneity of the residuals can be achieved. During the process of the forward search, we monitor the quantity of interest called score statistic and some other diagnostic plots. They jointly provide a wealth of information about transformation along with the effect of individual observation on this statistic.