Search

Experiments at synchrotron radiation sources and X-ray Free-Electron Lasers in the soft X-ray energy range ($250$eV--$2$keV) stand to benefit from the adaptation of the hybrid silicon detector technology for low energy photons. Inverse Low Gain Avalanche Diode (iLGAD) sensors provide an internal gain, enhancing the signal-to-noise ratio and allowing single photon detection below $1$keV using hybrid detectors. In addition, an optimization of the entrance window of these sensors enhances their quantum efficiency (QE). In this work, the QE and the gain of a batch of different iLGAD diodes with optimized entrance windows were characterized using soft X-rays at the Surface/Interface:Microscopy beamline of the Swiss Light Source synchrotron. Above $250$eV, the QE is larger than $55\%$ for all sensor variations, while the charge collection efficiency is close to $100\%$. The average gain depends on the gain layer design of the iLGADs and increases with photon energy. A fitting procedure is introduced to extract the multiplication factor as a function of the absorption depth of X-ray photons inside the sensors. In particular, the multiplication factors for electron- and hole-triggered avalanches are estimated, corresponding to photon absorption beyond or before the gain layer, respectively., Comment: 16 pages, 8 figures

Many new methodologies for the control of large-scale multi-agent systems are based on macroscopic representations of the emerging system dynamics, in the form of continuum approximations of large ensembles. These techniques, that are developed in the limit case of an infinite number of agents, are usually validated only through numerical simulations. In this paper, we introduce a mixed reality set-up for testing swarm robotics techniques, focusing on the macroscopic collective motion of robotic swarms. This hybrid apparatus combines both real differential drive robots and virtual agents to create a heterogeneous swarm of tunable size. We also extend continuification-based control methods for swarms to higher dimensions, and assess experimentally their validity in the new platform. Our study demonstrates the effectiveness of the platform for conducting large-scale swarm robotics experiments, and it contributes new theoretical insights into control algorithms exploiting continuification approaches.

AI-based code generators have become pivotal in assisting developers in writing software starting from natural language (NL). However, they are trained on large amounts of data, often collected from unsanitized online sources (e.g., GitHub, HuggingFace). As a consequence, AI models become an easy target for data poisoning, i.e., an attack that injects malicious samples into the training data to generate vulnerable code. To address this threat, this work investigates the security of AI code generators by devising a targeted data poisoning strategy. We poison the training data by injecting increasing amounts of code containing security vulnerabilities and assess the attack's success on different state-of-the-art models for code generation. Our study shows that AI code generators are vulnerable to even a small amount of poison. Notably, the attack success strongly depends on the model architecture and poisoning rate, whereas it is not influenced by the type of vulnerabilities. Moreover, since the attack does not impact the correctness of code generated by pre-trained models, it is hard to detect. Lastly, our work offers practical insights into understanding and potentially mitigating this threat., Comment: Accepted for publication at the International Conference on Program Comprehension 2024

The possible existence of primordial black holes (PBHs) is an open question in modern cosmology. Among the probes to test it, gravitational waves (GW) coming from their mergers constitute a powerful tool. In this work, we study how stellar mass PBH binaries could affect measurements of the clustering of merger events in future GW surveys. We account for PBH binaries formed both in the early and late Universe and show that the power spectrum modification they introduce can be detected at $\sim 2\sigma-3\sigma$ (depending on some assumptions) whenever PBH mergers make up at least $\sim 60\%$ of the overall number of detected events. By adding cross-correlations with galaxy surveys, this threshold is lowered to $\sim 40\%$. In the case of a poor redshift determination of GW sources, constraints are degraded by about a factor of 2. Assuming a theoretical model for the PBH merger rate, we can convert our results to constraints on the fraction of dark matter in PBHs, $f_{\rm PBH}$. Finally, we perform a Bayesian model selection forecast and confirm that the analysis we develop could be able to detect $\sim30M_\odot$ PBHs if they account for $f_{\rm PBH}\sim 10^{-4}-10^{-3}$, depending on the model uncertainty considered, being thus competitive with other probes., Comment: 41 pages, 17 figures, 4 tables. Accepted for publication on JCAP

We perform a detailed study of the cosmological bias of gravitational gave (GW) events produced by binary black hole mergers (BBHM). We start from a BBHM distribution modeled inside the EAGLE hydrodyamical simulation using the population synthesis code MOBSE. We then compare our findings with predictions from different Halo Occupation Distribution (HOD) prescriptions and find overall agreement, provided that the modeled properties of host galaxies and halos in the semi-analytical treatment match those in the simulations. By highlighting the sources of these discrepancies, we provide the stepping stone to build future more robust models that prevent the shortcoming of both simulation-based and analytical models. Finally, we train a neural network to build a simulation-based HOD and perform feature importance analysis to gain intuition on which host halo/galaxy parameters are the most relevant in determining the actual distribution and power spectrum of BBHM. We find that the distribution of BBHM in a galaxy does not only depend on its size, star formation rate and metallicity, but also by its kinetic state.

We study signatures of primordial non-Gaussianity (PNG) in the redshift-space halo field on non-linear scales, using a combination of three summary statistics, namely the halo mass function (HMF), power spectrum, and bispectrum. The choice of adding the HMF to our previous joint analysis of power spectrum and bispectrum is driven by a preliminary field-level analysis, in which we train graph neural networks on halo catalogues to infer the PNG $f_\mathrm{NL}$ parameter. The covariance matrix and the responses of our summaries to changes in model parameters are extracted from a suite of halo catalogues constructed from the Quijote-PNG N-body simulations. We consider the three main types of PNG: local, equilateral and orthogonal. Adding the HMF to our previous joint analysis of power spectrum and bispectrum produces two main effects. First, it reduces the equilateral $f_\mathrm{NL}$ predicted errors by roughly a factor $2$, while also producing notable, although smaller, improvements for orthogonal PNG. Second, it helps break the degeneracy between the local PNG amplitude, $f_\mathrm{NL}^\mathrm{local}$, and assembly bias, $b_{\phi}$, without relying on any external prior assumption. Our final forecasts for PNG parameters are $\Delta f_\mathrm{NL}^\mathrm{local} = 40$, $\Delta f_\mathrm{NL}^\mathrm{equil} = 210$, $\Delta f_\mathrm{NL}^\mathrm{ortho} = 91$, on a cubic volume of $1 \left(h^{-1}{\rm Gpc}\right)^3$, with a halo number density of $\bar{n}\sim 5.1 \times 10^{-5}~h^3\mathrm{Mpc}^{-3}$, at $z = 1$, and considering scales up to $k_\mathrm{max} = 0.5~h\,\mathrm{Mpc}^{-1}$., Comment: 17 pages, 11 figures. v3 (minor caption fix)

In real-world scenario, many phenomena produce a collection of events that occur in continuous time. Point Processes provide a natural mathematical framework for modeling these sequences of events. In this survey, we investigate probabilistic models for modeling event sequences through temporal processes. We revise the notion of event modeling and provide the mathematical foundations that characterize the literature on the topic. We define an ontology to categorize the existing approaches in terms of three families: simple, marked, and spatio-temporal point processes. For each family, we systematically review the existing approaches based based on deep learning. Finally, we analyze the scenarios where the proposed techniques can be used for addressing prediction and modeling aspects., Comment: Image replacements

Cloud computing systems fail in complex and unforeseen ways due to unexpected combinations of events and interactions among hardware and software components. These failures are especially problematic when they are silent, i.e., not accompanied by any explicit failure notification, hindering the timely detection and recovery. In this work, we propose an approach to run-time failure detection tailored for monitoring multi-tenant and concurrent cloud computing systems. The approach uses a non-intrusive form of event tracing, without manual changes to the system's internals to propagate session identifiers (IDs), and builds a set of lightweight monitoring rules from fault-free executions. We evaluated the effectiveness of the approach in detecting failures in the context of the OpenStack cloud computing platform, a complex and "off-the-shelf" distributed system, by executing a campaign of fault injection experiments in a multi-tenant scenario. Our experiments show that the approach detects the failure with an F1 score (0.85) and accuracy (0.77) higher than the ones provided by the OpenStack failure logging mechanisms (0.53 and 0.50) and two non--session-aware run-time verification approaches (both lower than 0.15). Moreover, the approach significantly decreases the average time to detect failures at run-time (~114 seconds) compared to the OpenStack logging mechanisms., Comment: Paper accepted for publication in The Journal of Systems and Software

AI-based code generators are an emerging solution for automatically writing programs starting from descriptions in natural language, by using deep neural networks (Neural Machine Translation, NMT). In particular, code generators have been used for ethical hacking and offensive security testing by generating proof-of-concept attacks. Unfortunately, the evaluation of code generators still faces several issues. The current practice uses output similarity metrics, i.e., automatic metrics that compute the textual similarity of generated code with ground-truth references. However, it is not clear what metric to use, and which metric is most suitable for specific contexts. This work analyzes a large set of output similarity metrics on offensive code generators. We apply the metrics on two state-of-the-art NMT models using two datasets containing offensive assembly and Python code with their descriptions in the English language. We compare the estimates from the automatic metrics with human evaluation and provide practical insights into their strengths and limitations.

We study primordial non-Gaussian signatures in the redshift-space halo field on non-linear scales, using a quasi-maximum likelihood estimator based on optimally compressed power spectrum and modal bispectrum statistics. We train and validate the estimator on a suite of halo catalogues constructed from the Quijote-PNG N-body simulations, which we release to accompany this paper. We verify its unbiasedness and near optimality, for the three main types of primordial non-Gaussianity (PNG): local, equilateral, and orthogonal. We compare the modal bispectrum expansion with a $k$-binning approach, showing that the former allows for faster convergence of numerical derivatives in the computation of the score-function, thus leading to better final constraints. We find, in agreement with previous studies, that the local PNG signal in the halo-field is dominated by the scale-dependent bias signature on large scales and saturates at $k \sim 0.2~h\,\mathrm{Mpc}^{-1}$, whereas the small-scale bispectrum is the main source of information for equilateral and orthogonal PNG. Combining power spectrum and bispectrum on non-linear scales plays an important role in breaking degeneracies between cosmological and PNG parameters; such degeneracies remain however strong for equilateral PNG. We forecast that PNG parameters can be constrained with $\Delta f_\mathrm{NL}^\mathrm{local} = 45$, $\Delta f_\mathrm{NL}^\mathrm{equil} = 570$, $\Delta f_\mathrm{NL}^\mathrm{ortho} = 110$, on a cubic volume of $1 \left({ {\rm Gpc}/{ {\rm h}}} \right)^3$, at $z = 1$, considering scales up to $k_\mathrm{max} = 0.5~h\,\mathrm{Mpc}^{-1}$., Comment: 17 pages, 10 figures. v2: minor updates to match published version

In this white paper, we present the MegaMapper concept. The MegaMapper is a proposed ground-based experiment to measure Inflation parameters and Dark Energy from galaxy redshifts at $2

We investigate how much can be learnt about four types of primordial non-Gaussianity (PNG) from small-scale measurements of the halo field. Using the QUIJOTE-PNG simulations, we quantify the information content accessible with measurements of the halo power spectrum monopole and quadrupole, the matter power spectrum, the halo-matter cross spectrum and the halo bispectrum monopole. This analysis is the first to include small, non-linear scales, up to $k_\mathrm{max}=0.5 \mathrm{h/Mpc}$, and to explore whether these scales can break degeneracies with cosmological and nuisance parameters making use of thousands of N-body simulations. We perform all the halo measurements in redshift space with a single sample comprised of all halos with mass $>3.2 \times 10^{13}~h^{-1}M_\odot$. For local PNG, measurements of the scale dependent bias effect from the power spectrum using sample variance cancellation provide significantly tighter constraints than measurements of the halo bispectrum. In this case measurements of the small scales add minimal additional constraining power. In contrast, the information on equilateral and orthogonal PNG is primarily accessible through the bispectrum. For these shapes, small scale measurements increase the constraining power of the halo bispectrum by up to $\times4$, though the addition of scales beyond $k\approx 0.3 \mathrm{h/Mpc}$ improves constraints largely through reducing degeneracies between PNG and the other parameters. These degeneracies are even more powerfully mitigated through combining power spectrum and bispectrum measurements. However even with combined measurements and small scale information, equilateral non-Gaussianity remains highly degenerate with $\sigma_8$ and our bias model., Comment: Updated to accepted version

Future Large Scale Structure surveys are expected to improve over current bounds on primordial non-Gaussianity (PNG), with a significant impact on our understanding of early Universe physics. The level of such improvements will however strongly depend on the extent to which late time non-linearities erase the PNG signal on small scales. In this work, we show how much primordial information remains in the bispectrum of the non-linear dark matter density field by implementing a new, simulation-based, methodology for joint estimation of PNG amplitudes ($f_{\rm NL}$) and standard $\Lambda$CDM parameters. The estimator is based on optimally compressed statistics, which, for a given input density field, combine power spectrum and modal bispectrum measurements, and numerically evaluate their covariance and their response to changes in cosmological parameters. We train and validate the estimator using a large suite of N-body simulations (QUIJOTE-PNG), including different types of PNG (local, equilateral, orthogonal). We explicitly test the estimator's unbiasedness, optimality and stability with respect to changes in the total number of input realizations. While the dark matter power spectrum itself contains negligible PNG information, as expected, including it as an ancillary statistic increases the PNG information content extracted from the bispectrum by a factor of order $2$. As a result, we prove the capability of our approach to optimally extract PNG information on non-linear scales beyond the perturbative regime, up to $k_{\rm max} = 0.5~h\,{\rm Mpc}^{-1}$, obtaining marginalized $1$-$\sigma$ bounds of $\Delta f_{\rm NL}^{\rm local} \sim 16$, $\Delta f_{\rm NL}^{\rm equil} \sim 77$ and $\Delta f_{\rm NL}^{\rm ortho} \sim 40$ on a cubic volume of $1~(\mathrm{Gpc}/h)^3$ at $z=1$. At the same time, we discuss the significant information on cosmological parameters contained on these scales., Comment: 22 pages, 12 figures

Primordial non-Gaussianity (PNG) is one of the most powerful probes of the early Universe and measurements of the large scale structure of the Universe have the potential to transform our understanding of this area. However relating measurements of the late time Universe to the primordial perturbations is challenging due to the non-linear processes that govern the evolution of the Universe. To help address this issue we release a large suite of N-body simulations containing four types of PNG: \textsc{quijote-png}. These simulations were designed to augment the \textsc{quijote} suite of simulations that explored the impact of various cosmological parameters on large scale structure observables. Using these simulations we investigate how much information on PNG can be extracted by extending power spectrum and bispectrum measurements beyond the perturbative regime at $z=0.0$. This is the first joint analysis of the PNG and cosmological information content accessible with power spectrum and bispectrum measurements of the non-linear scales. We find that the constraining power improves significantly up to $k_\mathrm{max}\approx 0.3 h/{\rm Mpc}$, with diminishing returns beyond as the statistical probes signal-to-noise ratios saturate. This saturation emphasizes the importance of accurately modelling all the contributions to the covariance matrix. Further we find that combining the two probes is a powerful method of breaking the degeneracies with the $\Lambda$CDM parameters., Comment: The simulation products are available at https://quijote-simulations.readthedocs.io/en/latest/png.html

Neural Machine Translation (NMT) has reached a level of maturity to be recognized as the premier method for the translation between different languages and aroused interest in different research areas, including software engineering. A key step to validate the robustness of the NMT models consists in evaluating the performance of the models on adversarial inputs, i.e., inputs obtained from the original ones by adding small amounts of perturbation. However, when dealing with the specific task of the code generation (i.e., the generation of code starting from a description in natural language), it has not yet been defined an approach to validate the robustness of the NMT models. In this work, we address the problem by identifying a set of perturbations and metrics tailored for the robustness assessment of such models. We present a preliminary experimental evaluation, showing what type of perturbations affect the model the most and deriving useful insights for future directions., Comment: Paper accepted for publication in the proceedings of The 1st Intl. Workshop on Natural Language-based Software Engineering (NLBSE) to be held with ICSE 2022

Writing software exploits is an important practice for offensive security analysts to investigate and prevent attacks. In particular, shellcodes are especially time-consuming and a technical challenge, as they are written in assembly language. In this work, we address the task of automatically generating shellcodes, starting purely from descriptions in natural language, by proposing an approach based on Neural Machine Translation (NMT). We then present an empirical study using a novel dataset (Shellcode_IA32), which consists of 3,200 assembly code snippets of real Linux/x86 shellcodes from public databases, annotated using natural language. Moreover, we propose novel metrics to evaluate the accuracy of NMT at generating shellcodes. The empirical analysis shows that NMT can generate assembly code snippets from the natural language with high accuracy and that in many cases can generate entire shellcodes with no errors., Comment: 33 pages, 5 figures, 9 tables. To be published in Automated Software Engineering journal

We study the clustering of Gravitational Wave (GW) merger events and Supernovae IA (SN), as cosmic tracers in Luminosity Distance Space. We modify the publicly available CAMB code to numerically evaluate auto- and cross- power spectra for the different sources, including Luminosity Distance Space distortion effects generated by peculiar velocities and lensing convergence. We perform a multitracer Fisher analysis to forecast expected constraints on cosmological and GW bias coefficients, using outputs from hydrodinamical N-body simulations to determine the bias fiducial model and considering future observations from the Vera Rubin Observatory and Einstein Telescope (ET), both single and in a 3 detector network configuration. We find that adding SN to the GW merger dataset considerably improves the forecast, mostly by breaking significant parameter degeneracies, with final constraints comparable to those obtainable from a Euclid-like survey. GW merger bias is forecasted to be detectable with good significance even in the single ET case., Comment: Version accepted for publication on JCAP, 25 pages, 10 figures

Writing exploits for security assessment is a challenging task. The writer needs to master programming and obfuscation techniques to develop a successful exploit. To make the task easier, we propose an approach (EVIL) to automatically generate exploits in assembly/Python language from descriptions in natural language. The approach leverages Neural Machine Translation (NMT) techniques and a dataset that we developed for this work. We present an extensive experimental study to evaluate the feasibility of EVIL, using both automatic and manual analysis, and both at generating individual statements and entire exploits. The generated code achieved high accuracy in terms of syntactic and semantic correctness., Comment: Paper accepted at the 32nd International Symposium on Software Reliability Engineering (ISSRE 2021)