arXiv subjects

Massimo Bernaschi

Publications and source records attributed to Massimo Bernaschi.

At least 37 records · Page 2Linked to original sources

arXiv2021

LBcuda: a high-performance CUDA port of LBsoft for simulation of colloidal systems

We present LBcuda, a GPU accelerated version of LBsoft, our open-source MPI-based software for the simulation of multi-component colloidal flows. We describe the design principles, the optimization and the resulting performance as compared to the CPU version, using both an average cost GPU and high-end NVidia GPU cards (V100 and the latest A100). The results show a substantial acceleration for the fluid solver reaching up to 200 GLUPS (Giga Lattice Updates Per Second) on a cluster made of 512 A100 NVIDIA cards simulating a grid of eight billion lattice points. These results open attractive prospects for the computational design of new materials based on colloidal particles.

physics.comp-ph↗

arXiv2021

TLBfind: a Thermal Lattice Boltzmann code for concentrated emulsions with FINite-size Droplets

In this paper, we present TLBfind, a GPU code for simulating the hydrodynamics of droplets along with a dynamic temperature field. TLBfind hinges on a two-dimensional multi-component lattice Boltzmann (LB) model simulating a concentrated emulsion with finite-size droplets evolving in a thermal convective state, just above the transition from conduction to convection. The droplet concentration of the emulsion system is tenable, and at the core of the code lies the possibility to measure a large number of physical observables characterising the flow and droplets. Furthermore, TLBfind includes a parallel implementation on GPU of the Delaunay triangulation useful for the detection of droplets' plastic rearrangements, and several types of boundary conditions, supporting simulations of channels with structured rough walls.

physics.flu-dyn↗

arXiv2021

Unleashing the Tiger: Inference Attacks on Split Learning

We investigate the security of Split Learning -- a novel collaborative machine learning framework that enables peak performance by requiring minimal resources consumption. In the present paper, we expose vulnerabilities of the protocol and demonstrate its inherent insecurity by introducing general attack strategies targeting the reconstruction of clients' private training sets. More prominently, we show that a malicious server can actively hijack the learning process of the distributed model and bring it into an insecure state that enables inference attacks on clients' data. We implement different adaptations of the attack and test them on various datasets as well as within realistic threat scenarios. We demonstrate that our attack is able to overcome recently proposed defensive techniques aimed at enhancing the security of the split learning protocol. Finally, we also illustrate the protocol's insecurity against malicious clients by extending previously devised attacks for Federated Learning. To make our results reproducible, we made our code available at https://github.com/pasquini-dario/SplitNN_FSHA.

cs.CR↗

arXiv2021

Forensics for Microsoft Teams

Microsoft Teams is a collaboration and communication platform developed by Microsoft that replaces and extends Microsoft Skype for Business. It differs from Skype for Business by the fact that it exists only as part of the Microsoft 365 products whereas Skype for Business can be deployed completely or partly on-premise. During the pandemic emergency in 2020 and 2021 Microsoft Teams has increased dramatically its base of users as most of the meetings and the communications had to be conducted in virtual environments by users working remotely. Microsoft Teams allows users to collaborate sending and sharing information virtually with anyone internal or external to the an organization with PCs and mobile devices, therefore it requires a careful review of all the security configurations and procedures within the organization. Microsoft Teams infrastructure can also be integrated with PSTN telephone services, natively within the Microsoft 365 services or by integrating other PSTN service providers. Therefore, its architecture extends the perimeter that could be exploited for an attack. Microsoft Teams features can also be extended by Apps. There are hundreds of Apps developed by Microsoft and by other companies to address the various needs of modern collaboration. "Walkie Talkie", one of those Apps, transforms the Teams client installed in a mobile phone into a Walkie Talkie communication device for registered users. The goal of this paper is to describe different Teams' usage scenarios and to analyse Teams' PSTN and Teams' Walkie Talkie communication scenarios describing forensics procedures to investigate inappropriate users' conduct.

cs.CR↗

arXiv2021

Interpretable Probabilistic Password Strength Meters via Deep Learning

Probabilistic password strength meters have been proved to be the most accurate tools to measure password strength. Unfortunately, by construction, they are limited to solely produce an opaque security estimation that fails to fully support the user during the password composition. In the present work, we move the first steps towards cracking the intelligibility barrier of this compelling class of meters. We show that probabilistic password meters inherently own the capability of describing the latent relation occurring between password strength and password structure. In our approach, the security contribution of each character composing a password is disentangled and used to provide explicit fine-grained feedback for the user. Furthermore, unlike existing heuristic constructions, our method is free from any human bias, and, more importantly, its feedback has a probabilistic interpretation. In our contribution: (1) we formulate interpretable probabilistic password strength meters; (2) we describe how they can be implemented via an efficient and lightweight deep learning framework suitable for client-side operability.

cs.CR↗

arXiv2021

Effective modelling of the Rayleigh-Bénard convection of concentrated emulsions with finite-size droplets

We present mesoscale numerical simulations of Rayleigh-Bénard convection in a two-dimensional concentrated emulsion, confined between two parallel walls, heated from below and cooled from above, under the effect of buoyancy forces. The systems under study comprise finite-size droplets, whose concentration $Φ_0$ is varied, ranging from the dilute limit up to the point where the emulsion starts to be packed and exhibits non-Newtonian rheology. We focus on the characterisation of the convective heat transfer properties close to the transition from conductive to convective states. The convective flow is confined and heterogeneous, which causes the emulsion to exhibit concentration heterogeneities in space $ϕ_0(y)$, depending on the location in the wall-to-wall direction ($y$). With the aim of assessing quantitatively the heat transfer efficiency of such heterogeneous systems, we resort to a side-by-side comparison between the concentrated emulsion system and a single-phase (SP) system, whose local viscosity $η^{\mbox{SP}}(y)$ is suitably constructed from the shear rheology of the emulsion. Such comparison highlights that a suitable degree $Λ$ of coarse-graining needs to be introduced in the local viscosity $η_Λ^{\mbox{SP}}(y)$, in order for the single-phase system to attain the same heat transfer efficiency of the emulsion. Specifically, it is shown that a quantitative matching between the two systems is possible whenever the coarse-graining is performed over a scale of the order of the droplet size.

physics.flu-dyn↗

arXiv2021

Convective heat transfer of a model emulsion at the droplet scale

We numerically study the Rayleigh-Bénard (RB) convection in two-dimensional model emulsions confined between two parallel walls at fixed temperatures. The systems under study are heterogeneous, with finite-size droplets dispersed in a continuous phase. The droplet concentration is chosen to explore the convective heat transfer of both Newtonian (low droplet concentration) and non-Newtonian (high droplet concentration) emulsions, the latter exhibiting shear-thinning rheology, with a noticeable increase of viscosity at low shear rates. It is well known that the transition to convection of a homogeneous Newtonian system is accompanied by the onset of steady flow and time-independent heat flux; in marked contrast, the heterogeneity of emulsions brings in an additional and previously unexplored phenomenology. As a matter of fact, when the droplet concentration increases, we observe that the heat transfer process is mediated by a non-steady flow, with neat heat-flux fluctuations, obeying a non-Gaussian statistics. The observed findings are ascribed to the emergence of space correlations among distant droplets, which we highlight via direct measurements of the droplets displacement and the characterisation of the associated correlation functions.

physics.flu-dyn↗

arXiv2021

Inferring urban social networks from publicly available data

The emergence of social networks and the definition of suitable generative models for synthetic yet realistic social graphs are widely studied problems in the literature. By not being tied to any real data, random graph models cannot capture all the subtleties of real networks and are inadequate for many practical contexts -- including areas of research, such as computational epidemiology, which are recently high on the agenda. At the same time, the so-called contact networks describe interactions, rather than relationships, and are strongly dependent on the application and on the size and quality of the sample data used to infer them. To fill the gap between these two approaches, we present a data-driven model for urban social networks, implemented and released as open source software. Given a territory of interest, and only based on widely available aggregated demographic and social-mixing data, we construct an age-stratified and geo-referenced synthetic population whose individuals are connected by "strong ties" of two types: intra-household (e.g., kinship) or friendship. While household links are entirely data-driven, we propose a parametric probabilistic model for friendship, based on the assumption that distances and age differences play a role, and that not all individuals are equally sociable. The demographic and geographic factors governing the structure of the obtained network, under different configurations, are thoroughly studied through extensive simulations focused on three Italian cities of different size.

cs.SI↗

arXiv2021

Reducing Bias in Modeling Real-world Password Strength via Deep Learning and Dynamic Dictionaries

Password security hinges on an in-depth understanding of the techniques adopted by attackers. Unfortunately, real-world adversaries resort to pragmatic guessing strategies such as dictionary attacks that are inherently difficult to model in password security studies. In order to be representative of the actual threat, dictionary attacks must be thoughtfully configured and tuned. However, this process requires a domain-knowledge and expertise that cannot be easily replicated. The consequence of inaccurately calibrating dictionary attacks is the unreliability of password security analyses, impaired by a severe measurement bias. In the present work, we introduce a new generation of dictionary attacks that is consistently more resilient to inadequate configurations. Requiring no supervision or domain-knowledge, this technique automatically approximates the advanced guessing strategies adopted by real-world attackers. To achieve this: (1) We use deep neural networks to model the proficiency of adversaries in building attack configurations. (2) Then, we introduce dynamic guessing strategies within dictionary attacks. These mimic experts' ability to adapt their guessing strategies on the fly by incorporating knowledge on their targets. Our techniques enable more robust and sound password strength estimates within dictionary attacks, eventually reducing overestimation in modeling real-world threats in password security. Code available: https://github.com/TheAdamProject/adams

cs.CR↗

arXiv2021

Rayleigh-Bénard convection of a model emulsion: anomalous heat-flux fluctuations and finite-size droplets effects

We present mesoscale numerical simulations of Rayleigh-Bénard (RB) convection in a two-dimensional model emulsion. The systems under study are constituted of finite-size droplets, whose concentration Phi_0 is systematically varied from small (Newtonian emulsions) to large values (non-Newtonian emulsions). We focus on the characterisation of the heat transfer properties close to the transition from conductive to convective states, where it is known that a homogeneous Newtonian system exhibits a steady flow and a time-independent heat flux. In marked contrast, emulsions exhibit a non-steady dynamics with fluctuations in the heat flux. In this paper, we aim at the characterisation of such non-steady dynamics via detailed studies on the time-averaged heat flux and its fluctuations. To understand the time-averaged heat flux, we propose a side-by-side comparison between the emulsion system and a single-phase (SP) system, whose viscosity is constructed from the shear rheology of the emulsion. We show that such local closure works well only when a suitable degree of coarse-graining (at the droplet scale) is introduced in the local viscosity. To delve deeper into the fluctuations in the heat flux, we propose a side-by-side comparison between a Newtonian emulsion and a non-Newtonian emulsion, at fixed time-averaged heat flux. This comparison elucidates that finite-size droplets and the non-Newtonian rheology cooperate to trigger enhanced heat-flux fluctuations at the droplet scales. These enhanced fluctuations are rooted in the emergence of space correlations among distant droplets, which we highlight via direct measurements of the droplets displacement and the characterisation of the associated correlation function. The observed findings offer insights on heat transfer properties for confined systems possessing finite-size constituents.

physics.flu-dyn↗

arXiv2021

Onion under Microscope: An in-depth analysis of the Tor network

Tor is an anonymity network that allows offering and accessing various kinds of resources, known as hidden services, while guaranteeing sender and receiver anonymity. The Tor web is the set of web resources that exist on the Tor network, and Tor websites are part of the so-called dark web. Recent research works have evaluated Tor security, evolution over time, and thematic organization. Nevertheless, few information are available about the structure of the graph defined by the network of Tor websites. The limited number of Tor entry points that can be used to crawl the network renders the study of this graph far from being simple. In this paper we aim at better characterizing the Tor Web by analyzing three crawling datasets collected over a five-month time frame. On the one hand, we extensively study the global properties of the Tor Web, considering two different graph representations and verifying the impact of Tor's renowned volatility. We present an in depth investigation of the key features of the Tor Web graph showing what makes it different from the surface Web graph. On the other hand, we assess the relationship between contents and structural features. We analyse the local properties of the Tor Web to better characterize the role different services play in the network and to understand to which extent topological features are related to the contents of a service.

cs.SI↗

arXiv2020

A GPU-accelerated adaptive FSAI preconditioner for massively parallel simulations

The solution of linear systems of equations is a central task in a number of scientific and engineering applications. In many cases the solution of linear systems may take most of the simulation time thus representing a major bottleneck in the further development of scientific and technical software. For large scale simulations, nowadays accounting for several millions or even billions of unknowns, it is quite common to resort to preconditioned iterative solvers for exploiting their low memory requirements and, at least potential, parallelism. Approximate inverses have been shown to be robust and effective preconditioners in various contexts. In this work, we show how adaptive FSAI, an approximate inverse characterized by a very high degree of parallelism, can be successfully implemented on a distributed memory computer equipped with GPU accelerators. Taking advantage of GPUs in adaptive FSAI set-up is not a trivial task, nevertheless we show through an extensive numerical experimentation how the proposed approach outperforms more traditional preconditioners and results in a close-to-ideal behaviour in challenging linear algebra problems.

math.NA↗

arXiv2020

Improving Password Guessing via Representation Learning

Learning useful representations from unstructured data is one of the core challenges, as well as a driving force, of modern data-driven approaches. Deep learning has demonstrated the broad advantages of learning and harnessing such representations. In this paper, we introduce a deep generative model representation learning approach for password guessing. We show that an abstract password representation naturally offers compelling and versatile properties that can be used to open new directions in the extensively studied, and yet presently active, password guessing field. These properties can establish novel password generation techniques that are neither feasible nor practical with the existing probabilistic and non-probabilistic approaches. Based on these properties, we introduce:(1) A general framework for conditional password guessing that can generate passwords with arbitrary biases; and (2) an Expectation Maximization-inspired framework that can dynamically adapt the estimated password distribution to match the distribution of the attacked password set.

cs.CR↗

arXiv2020

LBsoft: a parallel open-source software for simulation of colloidal systems

We present LBsoft, an open-source software developed mainly to simulate the hydro-dynamics of colloidal systems based on the concurrent coupling between lattice Boltzmann methods for the fluid and discrete particle dynamics for the colloids. Such coupling has been developed before, but, to the best of our knowledge, no detailed discussion of the programming issues to be faced in order to attain efficient implementation on parallel architectures, has ever been presented to date. In this paper, we describe in detail the underlying multi-scale models, their coupling procedure, along side with a description of the relevant input variables, to facilitate third-parties usage. The code is designed to exploit parallel computing platforms, taking advantage also of the recent AVX-512 instruction set. We focus on LBsoft structure, functionality, parallel implementation, performance and availability, so as to facilitate the access to this computational tool to the research community in the field. The capabilities of LBsoft are highlighted for a number of prototypical case studies, such as pickering emulsions, bicontinuous systems, as well as an original study of the coarsening process in confined bijels under shear.

physics.comp-ph↗

arXiv2020

Towards Exascale Lattice Boltzmann computing

We discuss the state of art of Lattice Boltzmann (LB) computing, with special focus on prospective LB schemes capable of meeting the forthcoming Exascale challenge. After reviewing the basic notions of LB computing, we discuss current techniques to improve the performance of LB codes on parallel machines and illustrate selected leading-edge applications in the Petascale range. Finally, we put forward a few ideas on how to improve the communication/computation overlap in current large-scale LB simulations, as well as possible strategies towards fault-tolerant LB schemes.

physics.comp-ph↗

arXiv2019

Strong ergodicity breaking in aging of mean field spin glasses

Out of equilibrium relaxation processes show aging if they become slower as time passes. Aging processes are ubiquitous and play a fundamental role in the physics of glasses and spin glasses and in other applications (e.g. in algorithms minimizing complex cost/loss functions). The theory of aging in the out of equilibrium dynamics of mean-field spin glass models has achieved a fundamental role, thanks to the asymptotic analytic solution found by Cugliandolo and Kurchan. However this solution is based on assumptions (e.g. the weak ergodicity breaking hypothesis) which have never been put under a strong test until now. In the present work we present the results of an extraordinary large set of numerical simulations of the prototypical mean-field spin glass models, namely the Sherrington-Kirkpatrick and the Viana-Bray models. Thanks to a very intensive use of GPUs, we have been able to run the latter model for more than $2^{64}$ spin updates and thus safely extrapolate the numerical data both in the thermodynamical limit and in the large times limit. The measurements of the two-times correlation functions in isothermal aging after a quench from a random initial configuration to a temperature $T<T_c$ provides clear evidence that, at large times, such correlations do not decay to zero as expected by assuming weak ergodicity breaking. We conclude that strong ergodicity breaking takes place in mean-field spin glasses aging dynamics which, asymptotically, takes place in a confined configurational space. Theoretical models for the aging dynamics need to be revised accordingly.

cond-mat.dis-nn↗

arXiv2019

A Performance Study of the 2D Ising Model on GPUs

The simulation of the two-dimensional Ising model is used as a benchmark to show the computational capabilities of Graphic Processing Units (GPUs). The rich programming environment now available on GPUs and flexible hardware capabilities allowed us to quickly experiment with several implementation ideas: a simple stencil-based algorithm, recasting the stencil operations into matrix multiplies to take advantage of Tensor Cores available on NVIDIA GPUs, and a highly optimized multi-spin coding approach. Using the managed memory API available in CUDA allows for simple and efficient distribution of these implementations across a multi-GPU NVIDIA DGX-2 server. We show that even a basic GPU implementation can outperform current results published on TPUs and that the optimized multi-GPU implementation can simulate very large lattices faster than custom FPGA solutions.

cs.DC↗

arXiv2019

Adversarial Out-domain Examples for Generative Models

Deep generative models are rapidly becoming a common tool for researchers and developers. However, as exhaustively shown for the family of discriminative models, the test-time inference of deep neural networks cannot be fully controlled and erroneous behaviors can be induced by an attacker. In the present work, we show how a malicious user can force a pre-trained generator to reproduce arbitrary data instances by feeding it suitable adversarial inputs. Moreover, we show that these adversarial latent vectors can be shaped so as to be statistically indistinguishable from the set of genuine inputs. The proposed attack technique is evaluated with respect to various GAN images generators using different architectures, training processes and for both conditional and not-conditional setups.

cs.LG↗