Cum hoc ergo propter hoc | EPFL Graph Search

Cum hoc ergo propter hoc (latin signifiant avec ceci, donc à cause de ceci) est un sophisme qui consiste à prétendre que si deux événements sont corrélés, alors, il y a un lien de cause à effet entre les deux. La confusion entre corrélation et causalité est appelée effet cigogne en zététique (en référence à la corrélation trompeuse entre le nombre de nids de cigognes et celui des naissances humaines) ; en science et particulièrement en statistique cette erreur est rappelée par la phrase « la corrélation n'implique pas la causalité », en latin : cum hoc sed non propter hoc (avec ceci, cependant pas à cause de ceci). Principe L'argument fallacieux peut être résumé ainsi :

L'événement A est corrélé à l'événement B.
Donc A cause B.

Le sophisme consiste à conclure sur la causalité seulement après avoir constaté la corrélation. En soi, affirmer de façon asymétrique que l'un des événements est corrélé à l'autre, plutôt qu'une énonciation symétrique « les deu

L'événement proxémique : étude des relations de circulation entre piétons aux heures de pointe à Delhi, Los Angeles, Paris et Tokyo

Lucas François Tiphine

In a context of traffic networks saturation related to urbanisation acceleration, this research is aimed at contributing to the scientific debate on this issue by building a model to describe dynamic pedestriansâ physical distances regulation behaviours in public places. The urban dimension of Edward T. Hallâs theory of proxemics is taken as an initial inspiration. This author argues that a correlation exists between the organisation of space at the macro level of Society and the micro level of individual behaviours. It leads him to a synthetic distinction between âcontact culturesâ and ânon-contact culturesâ. The former tends to prefer smaller distances at all scales of space organisation compared to the latter. This hypothesis is considered seriously in the research. Nonetheless, Hallâs macro culturalist anthropological area distinction (e.g.: the âArab Worldâ = âcontact cultureâ) is not relevant anymore in the urban globalization era. Therefore, a new âmicromacroâ theory, based on 4 macro orientations expected to have an influence on micro behaviours, is proposed: equality between urban citizens, individualisation, relative urbanity and urban planning regulation. Two other scalar correlation theories of spatial relations description are also tested. One, called âmicromesoâ, is based on Roger Barkerâs behaviour setting theory. It states that proxemical behaviours are correlated with intermediate units related to a specific place and time such as âthe subway at peak hoursâ. The other theory, called âmicroâ, states that any form of scalar correlation is an ecological fallacy. Accordingly, the description process has to remain at the very level of physical distance relations between individuals. These different theories are tested empirically in Delhi, Los Angeles, Paris and Tokyo which all answer differently to the macro orientations of the micromacro theory. Behaviours inside the metro and outside of it are observed (micromeso theory). In each condition, similar levels of spatial competition for places are taken into consideration (micro theory). The results lead to conclude that traffic relations between pedestrians are best described when considered cumulatively as micro > micromeso > micromacro by order of efficiency. Eventually, I defend the thesis that urban dwellers should debate of traffic relationships micro conditions improvements rather than focusing on micromacro and micromeso explanations of the networks breakdowns which can be found for instance in the competition promoted between individuals on civil behaviours.

EPFL2018

Applied stream ciphers in mobile communications

Yi Lu

This dissertation is concerned with cryptanalysis of E0, the stream cipher used in the short-range wireless radio standard Bluetooth, and of its generalization by means of correlation attacks. It consists of three parts. In the first part, we propose an E0-like combiner with memory as the core stream cipher. First, we formulate a systematic and simple method to compute the correlations. An upper bound of the correlations is given. Second, we show how to build either a uni-bias-based or multi-bias-based distinguisher to distinguish the keystream produced by the combiner from a truly random sequence, once correlations are found. The data complexity of either distinguisher is analyzed for performance comparison. The keystream distinguisher is then upgraded for use in the key-recovery attack. The latter reduces to the well-known maximum likelihood decoding problem given the keystream long enough. In the second part, the core stream cipher is transformed into the dedicated stream cipher by attaching the one-level or two-level initialization scheme. We show that the correlation attack on the core stream cipher leads to the correlation attack on the dedicated stream cipher with the one-level initialization scheme (with equal bias), but not necessarily so with the two-level initialization scheme. In the last part, we generalize the existing concept of conditional correlations and study conditional correlation attacks against stream ciphers and other cryptosystems. A general framework is developed for smart distinguishers, which exploit those generalized conditional correlations. Based on the theory of the traditional distinguisher, we derive the number of samples necessary for a smart distinguisher to succeed. It allows to prove that the smart distinguisher improves on the traditional basic distinguisher. As an application of all our analysis, it leads to the fastest (and only) practical known-plaintext attack on Bluetooth encryption so far. Our attack recovers the encryption key using the first 24 bits of 223.8 frames and with 238 computations.

EPFL2006

High robustness of correlation-based alignment with Penrose patterns to marker damage in electron beam lithography

Correlation-based alignment is an alternative alignment method for electron beam lithography. Using complex marker patterns, such as Penrose patterns, which contain more positional information, greater alignment accuracy can be achieved. Correlation-based alignment with Penrose patterns is less susceptible to marker edge defects, such as rat bites, roughness and flagging, since many more edges contribute to determining the marker position. There are however other defects associated with fabricating markers and this paper investigates how defects that result in parts of the pattern being omitted or obscured affect the correlation process when using Penrose pattern markers. We show that in both cases severely damaged markers can be used successfully and demonstrate fabricated structures with sub-5 nm alignment using markers with up to 80% of the marker pattern missing. (C) 2008 Elsevier B.V. All rights reserved

Elsevier2009