Search

We introduce a novel approach to text classification by combining doc2vec embeddings with advanced clustering techniques to improve the analysis of specialized, high-dimensional textual data. We integrate unsupervised methods such as Louvain, K-means, and Spectral clustering with doc2vec to enhance the detection of semantic patterns across a large corpus. As a case study, we apply this methodology to cybersecurity risk analysis using the MITRE ATT\&CK framework to structure and reduce the dimensionality of cyberattack tactics. Louvain clustering proved the most effective among the tested methods, achieving the best balance between cluster coherence and computational efficiency. Our approach identifies four "super tactics," demonstrating how clustering improves thematic coherence and risk attribution. The results validate the utility of combining doc2vec with clustering, particularly Louvain, for enhancing topic modeling and text classification., Comment: arXiv admin note: substantial text overlap with arXiv:2409.08728

We use a methodology based on a machine learning algorithm to quantify firms' cyber risks based on their disclosures and a dedicated cyber corpus. The model can identify paragraphs related to determined cyber-threat types and accordingly attribute several related cyber scores to the firm. The cyber scores are unrelated to other firms' characteristics. Stocks with high cyber scores significantly outperform other stocks. The long-short cyber risk factors have positive risk premia, are robust to all factors' benchmarks, and help price returns. Furthermore, we suggest the market does not distinguish between different types of cyber risks but instead views them as a single, aggregate cyber risk.

AI assistants are being increasingly used by students enrolled in higher education institutions. While these tools provide opportunities for improved teaching and education, they also pose significant challenges for assessment and learning outcomes. We conceptualize these challenges through the lens of vulnerability, the potential for university assessments and learning outcomes to be impacted by student use of generative AI. We investigate the potential scale of this vulnerability by measuring the degree to which AI assistants can complete assessment questions in standard university-level STEM courses. Specifically, we compile a novel dataset of textual assessment questions from 50 courses at EPFL and evaluate whether two AI assistants, GPT-3.5 and GPT-4 can adequately answer these questions. We use eight prompting strategies to produce responses and find that GPT-4 answers an average of 65.8% of questions correctly, and can even produce the correct answer across at least one prompting strategy for 85.1% of questions. When grouping courses in our dataset by degree program, these systems already pass non-project assessments of large numbers of core courses in various degree programs, posing risks to higher education accreditation that will be amplified as these models improve. Our results call for revising program-level assessment design in higher education in light of advances in generative AI., Comment: 20 pages, 8 figures

Mechanical degradation in electrode materials during successive electrochemical cycling is critical for battery lifetime and aging properties. A common strategy to mitigate electrode mechanical degradation is to suppress the volume variation induced by Li/Na intercalation/deintercalation, thereby designing strain-less electrodes. In this study, we investigate the electrochemically-induced volume variation in layered and spinel compounds used in Li-ion and Na-ion battery electrode materials through density functional theory computations. Specifically, we propose to decompose the volume variation into electronic, ionic, and structural contributions. Based on this analysis, we suggest methods to separately influence or control each contribution through strategies such as chemical substitution, doping, and polymorphism. Altogether, we conclude that volume variations can be controlled by designing either mechanically hard or compact electrode materials.

The Paris agreement is the first-ever universally accepted and legally binding agreement on global climate change. It is a bridge between today's and climate-neutrality policies and strategies before the end of the century. Critical to this endeavor is energy system modeling, which, while adept at devising cost-effective carbon-neutral strategies, often overlooks the broader environmental and social implications. This study introduces an innovative methodology that integrates life-cycle impact assessment indicators into energy system modeling, enabling a comprehensive assessment of both economic and environmental outcomes. Focusing on Switzerland's energy system as a case study, our model reveals that optimizing key environomic indicators can lead to significant economic advantages, with system costs potentially decreasing by 15% to 47% by minimizing potential impacts from operating fossil technologies to the indirect impact related to the construction of the renewable infrastructure. However, a system optimized solely for economic efficiency, despite achieving 63% reduction in carbon footprint compared to 2020, our results show a potential risk of burden shift to other environmental issues. The adoption of multi-objective optimization in our approach nuances the exploration of the complex interplay between environomic objectives and technological choices. Our results illuminate pathways towards more holistically optimized energy systems, effectively addressing trade-offs across environmental problems and enhancing societal acceptance of the solutions to this century's defining challenge.

We extract firms' cyber risk with a machine learning algorithm measuring the proximity between their disclosures and a dedicated cyber corpus. Our approach outperforms dictionary methods, uses full disclosure and not devoted-only sections, and generates a cyber risk measure uncorrelated with other firms' characteristics. We find that a portfolio of US-listed stocks in the high cyber risk quantile generates an excess return of 18.72% p.a. Moreover, a long-short cyber risk portfolio has a significant and positive risk premium of 6.93% p.a., robust to all factors' benchmarks. Finally, using a Bayesian asset pricing method, we show that our cyber risk factor is the essential feature that allows any multi-factor model to price the cross-section of stock returns., Comment: This document results from a research project funded by the Cyber-Defence Campus, armasuisse Science and Technology. We appreciate helpful comments from seminar participants at the Cyber Alp Retreat 2023. We also thank Pierre Collin-Dufresne and Michel Dubois for their invaluable comments

Along with the increasing frequency and severity of cyber incidents, understanding their economic implications is paramount. In this context, listed firms' reactions to cyber incidents are compelling to study since they (i) are a good proxy to estimate the costs borne by other organizations, (ii) have a critical position in the economy, and (iii) have their financial information publicly available. We extract listed firms' cyber incident dates and characteristics from newswire headlines. We use an event study over 2012--2022, using a three-day window around events and standard benchmarks. We find that the magnitude of abnormal returns around cyber incidents is on par with previous studies using newswire or alternative data to identify cyber incidents. Conversely, as we adjust the standard errors accounting for event-induced variance and residual cross-correlation, we find that the previously claimed significance of abnormal returns vanishes. Given these results, we run a horse race of specifications, in which we test for the marginal effects of type of cyber incidents, target firm sector, periods, and their interactions. Data breaches are the most detrimental incident type with an average loss of -1.3\% or (USD -1.9 billion) over the last decade. The health sector is the most sensitive to cyber incidents, with an average loss of -5.21\% (or USD -1.2 billion), and even more so when these are data breaches. Instead, we cannot show any time-varying effect of cyber incidents or a specific effect of the type of news as had previously been advocated., Comment: This document results from a research project funded by the Cyber-Defence Campus, armasuisse Science and Technology. Our code will be available at https://github.com/technometrics-lab/17-The-determinants-of-cyberattack-costs-An-event-study. Corresponding author: Daniel Celeny e-mail: daniel.celeny@alumni.epfl.ch, Cyber-Defence Campus, Innovation Park, EPFL, 1015 Lausanne

Early-stage firms play a significant role in driving innovation and creating new products and services, especially for cybersecurity. Therefore, evaluating their performance is crucial for investors and policymakers. This work presents a financial evaluation of early-stage firms' performance in 19 cybersecurity sectors using a private-equity dataset from 2010 to 2022 retrieved from Crunchbase. We observe firms, their primary and secondary activities, funding rounds, and pre and post-money valuations. We compare cybersecurity sectors regarding the amount raised over funding rounds and post-money valuations while inferring missing observations. We observe significant investor interest variations across categories, periods, and locations. In particular, we find the average capital raised (valuations) to range from USD 7.24 mln (USD 32.39 mln) for spam filtering to USD 45.46 mln (USD 447.22 mln) for the private cloud sector. Next, we assume a log process for returns computed from post-money valuations and estimate the expected returns, systematic and specific risks, and risk-adjusted returns of investments in early-stage firms belonging to cybersecurity sectors. Again, we observe substantial performance variations with annualized expected returns ranging from 9.72\% for privacy to 177.27\% for the blockchain sector. Finally, we show that overall, the cybersecurity industry performance is on par with previous results found in private equity. Our results shed light on the performance of cybersecurity investments and, thus, on investors' expectations about cybersecurity., Comment: This document results from a research project funded by the Cyber-Defence Campus, armasuisse Science and Technology. We appreciate helpful comments from seminar participants at the Cyber Alp Retreat 2022 and WEIS 2023

We have developed a poly(dimethylsiloxane) (PDMS) microfluidic chip to study the directional drying of a colloidal dispersion confined in a channel. Our measurements on a dispersion of silica nanoparticles once again revealed the phenomenology commonly observed for such systems: the formation of a porous solid with linear growth in the channel at short times, slowing down at longer times as the evaporation rate decreases. The growth of the solid is also accompanied by mechanical stresses that are released by the delamination of the solid from the channel walls and the formation of cracks. In addition to these observations, we report original measurements using hydrophilic filler in the PDMS formulation used (Sylgard-184). When the PDMS matrix is in contact with water, water molecules pool around these hydrophilic sites, resulting in the formation of microscopic water clusters whose size depends on the water potential $\psi$. In our work, we have used these water clusters to estimate the water potential profile in the channel as the porous solid grows. Using a transport model that also takes into account solid delamination in the channel, we then linked these water potential measurements to the hydraulic permeability of the porous solid. These measurements finally enabled us to show that the slowdown in the evaporation rate is due to the invasion of the porous solid by air/water nanomenisci at a critical capillary pressure $\psi_\text{cap}$.

The Swift-Hohenberg (SH) and Phase-Field Crystal (PFC) models are minimal yet powerful approaches for studying phenomena such as pattern formation, collective order, and defects via smooth order parameters. They are based on a free-energy functional that inherently includes elasticity effects. This study addresses how gradient elasticity (GE), a theory that accounts for elasticity effects at microscopic scales by introducing additional characteristic lengths, is incorporated into SH and PFC models. After presenting the fundamentals of these theories and models, we first calculate the characteristic lengths for various lattice symmetries in an approximated setting. We then discuss numerical simulations of stress fields at dislocations and comparisons with analytic solutions within first and second strain-gradient elasticity. Effective GE characteristic lengths for the elastic fields induced by dislocations are found to depend on the free-energy parameters in the same manner as the phase correlation length, thus unveiling how they change with the quenching depth. The findings presented in this study enable a thorough discussion and analysis of small-scale elasticity effects in pattern formation and crystalline systems using SH and PFC models and, importantly, complete the elasticity analysis therein. Additionally, we provide a microscopic foundation for GE in the context of order-disorder phase transitions., Comment: 31 pages, 5 figures

We present a mesoscale description of deformations and defects in thin, flexible sheets with crystalline order, tackling the interplay between in-plane elasticity, out-of-plane deformation, as well as dislocation nucleation and motion. Our approach is based on the Phase-Field Crystal (PFC) model, which describes the microscopic atomic density in crystals at diffusive timescales, naturally encoding elasticity and plasticity effects. In its amplitude expansion (APFC), a coarse-grained description of the mechanical properties of crystals is achieved. We introduce surface PFC and surface APFC models in a convenient height-function formulation encoding deformation in the normal direction. This framework is proven consistent with classical aspects of strain-induced buckling, defect nucleation on deformed surfaces, and out-of-plane relaxation near dislocations. In particular, we benchmark and discuss the results of numerical simulations by looking at the continuum limit for buckling under uniaxial compression and at evidence from microscopic models for deformation at defects and defect arrangements, demonstrating the scale-bridging capabilities of the proposed framework. Results concerning the interplay between lattice distortion at dislocations and out-of-plane deformation are also illustrated by looking at the annihilation of dislocation dipoles and systems hosting many dislocations. With the novel formulation proposed here, and its assessment with established approaches, we envision applications to multiscale investigations of crystalline order on deformable surfaces., Comment: 20 pages; 12 figures

With the rise of open data, identifiability of individuals based on 3D renderings obtained from routine structural magnetic resonance imaging (MRI) scans of the head has become a growing privacy concern. To protect subject privacy, several algorithms have been developed to de-identify imaging data using blurring, defacing or refacing. Completely removing facial structures provides the best re-identification protection but can significantly impact post-processing steps, like brain morphometry. As an alternative, refacing methods that replace individual facial structures with generic templates have a lower effect on the geometry and intensity distribution of original scans, and are able to provide more consistent post-processing results by the price of higher re-identification risk and computational complexity. In the current study, we propose a novel method for anonymised face generation for defaced 3D T1-weighted scans based on a 3D conditional generative adversarial network. To evaluate the performance of the proposed de-identification tool, a comparative study was conducted between several existing defacing and refacing tools, with two different segmentation algorithms (FAST and Morphobox). The aim was to evaluate (i) impact on brain morphometry reproducibility, (ii) re-identification risk, (iii) balance between (i) and (ii), and (iv) the processing time. The proposed method takes 9 seconds for face generation and is suitable for recovering consistent post-processing results after defacing., Comment: preprint

In this paper we study the inverse Laplace transform. We first derive a new global logarithmic stability estimate that shows that the inversion is severely ill-posed. Then we propose a regularization method to compute the inverse Laplace transform using the concept of mollification. Taking into account the exponential instability we derive a criterion for selection of the regularization parameter. We show that by taking the optimal value of this parameter we improve significantly the convergence of the method. Finally, making use of the holomorphic extension of the Laplace transform, we suggest a new PDEs based numerical method for the computation of the solution. The effectiveness of the proposed regularization method is demonstrated through several numerical examples., Comment: 22 pages, 10 figures

Generative Language Models gained significant attention in late 2022 / early 2023, notably with the introduction of models refined to act consistently with users' expectations of interactions with AI (conversational models). Arguably the focal point of public attention has been such a refinement of the GPT3 model -- the ChatGPT and its subsequent integration with auxiliary capabilities, including search as part of Microsoft Bing. Despite extensive prior research invested in their development, their performance and applicability to a range of daily tasks remained unclear and niche. However, their wider utilization without a requirement for technical expertise, made in large part possible through conversational fine-tuning, revealed the extent of their true capabilities in a real-world environment. This has garnered both public excitement for their potential applications and concerns about their capabilities and potential malicious uses. This review aims to provide a brief overview of the history, state of the art, and implications of Generative Language Models in terms of their principles, abilities, limitations, and future prospects -- especially in the context of cyber-defense, with a focus on the Swiss operational environment., Comment: 41 pages (without references), 13 figures; public report of Cyber-Defence Campus

The phase-field crystal (PFC) model describes crystal lattices at diffusive timescales. Its amplitude expansion (APFC) can be applied to the investigation of relatively large systems under some approximations. However, crystal symmetries accessible within the APFC model are limited to basic ones, namely triangular and square in two dimensions, and body-centered cubic and face-centered cubic in three dimensions. In this work, we propose a general, amplitudes-based description of virtually any lattice symmetry. To fully exploit the advantages of this model, featuring slowly varying quantities in bulk and localized significant variations at dislocations and interfaces, we consider formulations suitable for real-space numerical methods supporting adaptive spatial discretization. We explore approaches originally proposed for the PFC model which allow for symmetries beyond basic ones through extended parametrizations. Moreover, we tackle the modeling of non-Bravais lattices by introducing an amplitude expansion for lattices with a basis and further generalizations. We study and discuss the stability of selected, prototypical lattice symmetries. As pivotal examples, we show that the proposed approach allows for a coarse-grained description of the kagome lattice, exotic square arrangements, and the diamond lattice, as bulk crystals and, importantly, hosting dislocations., Comment: 14 pages, 7 figures

Grand Dyck paths with air pockets (GDAP) are a generalization of Dyck paths with air pockets by allowing them to go below the $x$-axis. We present enumerative results on GDAP (or their prefixes) subject to various restrictions such as maximal/minimal height, ordinate of the last point and particular first return decomposition. In some special cases we give bijections with other known combinatorial classes., Comment: 20 pages, 4 figures

We introduce TechRank, a recursive algorithm based on a bi-partite graph with weighted nodes. We develop TechRank to link companies and technologies based on the method of reflection. We allow the algorithm to incorporate exogenous variables that reflect an investor's preferences. We calibrate the algorithm in the cybersecurity sector. First, our results help estimate each entity's influence and explain companies' and technologies' ranking. Second, they provide investors with a quantitative optimal ranking of technologies and thus, help them design their optimal portfolio. We propose this method as an alternative to traditional portfolio management and, in the case of private equity investments, as a new way to price assets for which cash flows are not observable.

We introduce and study the new combinatorial class of Dyck paths with air pockets. We exhibit a bijection with the peakless Motzkin paths which transports several pattern statistics and give bivariate generating functions for the distribution of patterns as peaks, returns and pyramids. Then, we deduce the popularities and asymptotic expectations of these patterns and point out a link between the popularity of pyramids and a special kind of closed smooth self-overlapping curves, a subset of Fibonacci meanders. A similar study is conducted for non-decreasing Dyck paths with air pockets., Comment: 20 pages, 3 figures, 2 tables

Progress made in optical references, including ultra-stable Fabry-Perot cavities, optical frequency combs and optical atomic clocks, have driven the need for ultra-stable optical fiber networks. Telecom-wavelength ultra-pure optical signal transport has been demonstrated on distances ranging from the laboratory scale to the continental scale. In this manuscript, we present a Doppler-cancellation setup based on a digital phase-locked loop for ultra-stable optical signal dissemination over fiber. The optical phase stabilization setup is based on a usual heterodyne Michelson-interferometer setup, while the Software Defined Radio (SDR) implementation of the phase-locked loop is based on a compact commercial board embedding a field programmable gate array, analog-to-digital and digital-to-analog converters. Using three different configurations including an undersampling method, we demonstrate a 20 m long fiber link with residual fractional frequency instability as low as $10^{-18}$ at 1000 s, and an optical phase noise of $-70$ dBc/Hz at 1 Hz with a telecom frequency carrier., Comment: 11 pages, 6 figures

A sentinel network, Ob\'epine, has been designed to monitor SARS-CoV-2 viral load in wastewaters arriving at wastewater treatment plants (WWTPs) in France as an indirect macro-epidemiological parameter. The sources of uncertainty in such monitoring system are numerous and the concentration measurements it provides are left-censored and contain outliers, which biases the results of usual smoothing methods. Hence the need for an adapted pre-processing in order to evaluate the real daily amount of virus arriving to each WWTP. We propose a method based on an auto-regressive model adapted to censored data with outliers. Inference and prediction are produced via a discretised smoother which makes it a very flexible tool. This method is both validated on simulations and on real data from Ob\'epine. The resulting smoothed signal shows a good correlation with other epidemiological indicators and is currently used by Ob\'epine to provide an estimate of virus circulation over the watersheds corresponding to about 200 WWTPs., Comment: 18 pages, 9 figures

We propose and demonstrate an experimental method to measure by absorption imaging the size and local column density of a cloud of atoms, even when its smallest dimension is smaller than the resolution of the imaging system. To do this, we take advantage of the fact that, for a given total number of atoms, a smaller and denser cloud scatters less photons when the gas is optically thick. The method relies on making an ansatz on the cloud shape along the unresolved dimension(s), and on providing an additional information such as the total number of atoms. We demonstrate the method on \textit{in-situ} absorption images of elongated 87Sr Fermi gases. We find significant non-linear corrections to the estimated size and local density of the cloud compared to a standard analysis. This allows us to recover an un-distorted longitudinal density profile, and to measure transverse sizes as small as one fourth of our imaging resolution. The ultimate limit of our method is the wavelength that is used for imaging., Comment: 7 pages, 5 figures

The Cauchy problem for the inhomogeneous Helmholtz equation with non-uniform refraction index is considered. The ill-posedness of this problem is tackled by means of the variational form of mollification. This approach is proved to be consistent, and the proposed numerical simulations are quite promising.

We study the evolution of step bunches on vicinal surfaces using a thermodynamically consistent step-flow model that (i) circumvents the quasistatic approximation that prevails in the literature by accounting for the dynamics of adatom diffusion on terraces and attachment-detachment at steps (referred to as the dynamical effect), and (ii) generalizes the expression of the step chemical potential by incorporating the necessary coupling between the diffusion fields on adjacent terraces (referred to as the chemical effect). Having previously shown that these effects can explain the onset of step bunching without recourse to the inverse Ehrlich-Schwoebel (iES) barrier or other extraneous mechanisms, we are here interested in the evolution of step bunches beyond the linear-stability regime. In particular, the numerical resolution of the step-flow problem yields a robust power-law coarsening of the surface profile, with the bunch height growing in time as $H\sim t^{1/2}$ and the minimal interstep distance as a function of the number of steps in the bunch cell obeying $\ell_{min}\sim N^{-2/3}$. Although these exponents have previously been reported, this is the first time such scaling laws are obtained in the absence of an iES barrier or adatom electromigration. In order to validate our simulations, we take the continuum limit of the discrete step-flow system, leading to a novel nonlinear evolution equation for the surface height. We investigate the existence of self-similar solutions of this equation and confirm the 1/2 coarsening exponent obtained numerically for $H$. We highlight the influence of the combined dynamical-chemical effect and show that it can be interpreted as an effective iES barrier in the standard BCF theory. Finally, we use a Pad\'e approximant to derive an analytical expression for the velocity of steadily moving step bunches and compare it to numerical simulations., Comment: 14 pages, 9 figures

Using XMM-Newton, we undertook a dedicated project to search for X-ray bright wind-wind collisions in 18 WR+OB systems. We complemented these observations with Swift and Chandra datasets, allowing for the study of two additional systems. We also improved the ephemerides, for these systems displaying photometric changes, using TESS, Kepler, and ASAS-SN data. Five systems displayed a very faint X-ray emission ($\log [L_{\rm X}/L_{\rm BOL}]<-8$) and three a faint one ($\log [L_{\rm X}/L_{\rm BOL}]\sim-7$), incompatible with typical colliding wind emission: not all WR binaries are thus X-ray bright. In a few other systems, X-rays from the O-star companion cannot be excluded as being the true source of X-rays (or a large contributor). In two additional cases, the emission appears faint but the observations were taken with the WR wind obscuring the line-of-sight, which could hide a colliding wind emission. Clear evidence of colliding winds was however found in the remaining six systems (WR19, 21, 31, 97, 105, 127). In WR19, increased absorption and larger emission at periastron are even detected, in line with expectations of adiabatic collisions., Comment: accepted by MNRAS - the arxiv version has figures in low-resolution

We conduct an extensive study of nonlinear localized modes (NLMs), which are temporally periodic and spatially localized structures, in a two-dimensional array of repelling magnets. In our experiments, we arrange a lattice in a hexagonal configuration with a light-mass defect, and we harmonically drive the center of the chain with a tunable excitation frequency, amplitude, and angle. We use a damped, driven variant of a vector Fermi- Pasta-Ulam-Tsingou lattice to model our experimental setup. Despite the idealized nature of the model, we obtain good qualitative agreement between theory and experiments for a variety of dynamical behaviors. We find that the spatial decay is direction-dependent and that drive amplitudes along fundamental displacement axes lead to nonlinear resonant peaks in frequency continuations that are similar to those that occur in one-dimensional damped, driven lattices. However, driving along other directions leads to the creation of asymmetric NLMs that bifurcate from the main solution branch, which consists of symmetric NLMs. When we vary the drive amplitude, we observe such behavior both in our experiments and in our simulations. We also demonstrate that solutions that appear to be time-quasi-periodic bifurcate from the branch of symmetric time-periodic NLMs.