Search

In Mauritania, many learners struggle with French, one of the official languages of instruction at school. This language barrier hinders student progress with the curriculum, increasing the chances of leaving school. This risk disproportionately affects adolescent girls due to discriminatory social norms undermining their continued education, even in urban areas of the capital, Nouakchott. To address these challenges, UNICEF Mauritania, in collaboration with the Akelius Foundation and national partners, has initiated a blended course to improve French proficiency for girls in a disadvantaged neighbourhood of Nouakchott. The blended course, which takes place at a community-based facility during afternoons, integrates regular teaching with the offline use of the Akelius Digital Learning app. This research brief presents evidence on how the blended course supported vulnerable girls' learning and other socio-emotional skills. It highlights lessons learnt for practitioners in implementing digital learning in resource-constrained environments, including the reliance on offline solutions and the co-design of digital content with the teachers.

ModSecurity is widely recognized as the standard open-source Web Application Firewall (WAF), maintained by the OWASP Foundation. It detects malicious requests by matching them against the Core Rule Set (CRS), identifying well-known attack patterns. Each rule is manually assigned a weight based on the severity of the corresponding attack, and a request is blocked if the sum of the weights of matched rules exceeds a given threshold. However, we argue that this strategy is largely ineffective against web attacks, as detection is only based on heuristics and not customized on the application to protect. In this work, we overcome this issue by proposing a machine-learning model that uses the CRS rules as input features. Through training, ModSec-Learn is able to tune the contribution of each CRS rule to predictions, thus adapting the severity level to the web applications to protect. Our experiments show that ModSec-Learn achieves a significantly better trade-off between detection and false positive rates. Finally, we analyze how sparse regularization can reduce the number of rules that are relevant at inference time, by discarding more than 30% of the CRS rules. We release our open-source code and the dataset at https://github.com/pralab/modsec-learn and https://github.com/pralab/http-traffic-dataset, respectively., Comment: arXiv admin note: text overlap with arXiv:2308.04964

Heart rate variability (HRV) series reflects the dynamical variation of heartbeat-to-heartbeat intervals in time and is one of the outputs of the cardiovascular system. Over the years, this system has been recognized for generating nonlinear and complex heartbeat dynamics, with the latter referring to a high sensitivity to small -- theoretically infinitesimal -- input changes. While early research associated chaotic behavior with the cardiovascular system, evidence of stochastic inputs to the system, i.e., a physiological noise, invalidated those conclusions. To date, a comprehensive characterization of the cardiovascular system dynamics, accounting for dynamical noise input, has not been undertaken. In this study, we propose a novel methodological framework for evaluating the presence of regular or chaotic dynamics in noisy dynamical systems. The method relies on the estimation of asymptotic growth rate of noisy mean square displacement series in a two-dimensional phase space. We validated the proposed method using synthetic series comprising well-known regular and chaotic maps. We applied the method to real HRV series from healthy subjects, as well as patients with atrial fibrillation and congestive heart failure, during unstructured long-term activity. Results indicate that HRV series are consistently generated by a regular system driven by dynamical noise.

Purpose: This study examined a criterion-based sample of high school librarians who met AASL's definition of effective school library practice to uncover how they translated their professional practices during COVID-19. Methods: Three instruments were used to compare pre-COVID-19 school librarian practice with its evolution during two specific points during the pandemic. High School Librarian Survey 1 (Pre-COVID) (S1) provided a baseline view into the pre-COVID-19 practice of purposively sampled participants; open-ended Survey 2 (S2) and a follow-up focus group (FG) uncovered their practice at two specific points during the crisis and explored how librarians' priorities shifted. Results: The instruments which collected librarian perceptions, revealed that while librarians supported and contributed to their learning communities in essential ways, they were also forced to compromise their priorities. Librarians furthered their leadership roles, supported learning communities, exhibited resilience and integrity, and worked to assure equitable access to learning tools and reading materials. Yet, these efforts went largely unreported. Limited contact with teacher partners and classrooms compromised librarians' engagement with information literacy and inquiry instruction. Implications: Recommendations for future research include exploring how recognized gaps in information literacy learning and instruction may have been recovered since the pandemic and how school librarians' roles may have shifted and been impacted by the pandemic. Resources: Research instruments are collected on this LibGuide. https://comminfo.libguides.com/FirstYearsFrames/ResearchInstruments

Many Web Application Firewalls (WAFs) leverage the OWASP Core Rule Set (CRS) to block incoming malicious requests. The CRS consists of different sets of rules designed by domain experts to detect well-known web attack patterns. Both the set of rules to be used and the weights used to combine them are manually defined, yielding four different default configurations of the CRS. In this work, we focus on the detection of SQL injection (SQLi) attacks, and show that the manual configurations of the CRS typically yield a suboptimal trade-off between detection and false alarm rates. Furthermore, we show that these configurations are not robust to adversarial SQLi attacks, i.e., carefully-crafted attacks that iteratively refine the malicious SQLi payload by querying the target WAF to bypass detection. To overcome these limitations, we propose (i) using machine learning to automate the selection of the set of rules to be combined along with their weights, i.e., customizing the CRS configuration based on the monitored web services; and (ii) leveraging adversarial training to significantly improve its robustness to adversarial SQLi manipulations. Our experiments, conducted using the well-known open-source ModSecurity WAF equipped with the CRS rules, show that our approach, named ModSec-AdvLearn, can (i) increase the detection rate up to 30%, while retaining negligible false alarm rates and discarding up to 50% of the CRS rules; and (ii) improve robustness against adversarial SQLi attacks up to 85%, marking a significant stride toward designing more effective and robust WAFs. We release our open-source code at https://github.com/pralab/modsec-advlearn.

Learning remains largely out of reach for many of the most vulnerable children around the world. In low- and middle-income countries, an estimated 56% of children cannot read a simple text by the age of 10. This share is projected to rise to 70% after the pandemic. The school closures imposed by the COVID-19 outbreak, coupled with an enduring tendency in low-income countries to allocate a limited share of the national education budget to the most vulnerable, are further widening inequalities in the global learning crisis landscape. The Let Us Learn (LUL) initiative implements innovative education programmes to improve learning for the most vulnerable children in five countries with high levels of out-of-school children: Afghanistan, Bangladesh, Liberia, Madagascar and Nepal. This report documents the outcomes, lessons learned and recommendations based on the experience of the initiative across four types of learning programmes spanning the education lifecycle: (1) pre-primary education; (2) accelerated learning pathways; (3) programmes to reduce barriers to access and stay in formal school; and (4) vocational training.

In Afghanistan, 93% of children cannot read a simple text by the age of 10. Education is not available to everyone, especially for girls and children in remote areas. A form of community-based education, called Accelerated Learning Centers (ALCs), can help close the distance barrier and meet the needs of out-of-school children and girls. In May 2021, an assessment of foundational literacy and numeracy skills of ALC students and nearby government school students was conducted. Results show that children at ALCs are learning at similar levels or better compared with children who attend government schools. This report provides insight into practices to improve education in rural areas in Afghanistan.

One tool that many families own, across the globe, is a basic mobile phone. The use of low-cost basic mobile phones for educational purposes in humanitarian settings is critical where access to connectivity and higher cost devices is limited. The portability of mobile phones, combined with their communication features, offers multiple uses to support education in emergency (EiE). This report is the second of a two-part series that aims to provide practitioners with guidance and resources on, and lessons learned for, using mobile technology in EiE settings. The series draws from a review of the existing literature with feedback from and interviews with EiE practitioners. The first piece of this series -- "On Call: The Use of Mobile Phones for Learning in Emergencies" (ED624512) -- explores how basic mobile phones can be used to support EiE learning programmes and teacher training. It emphasizes key practices that education practitioners have adopted to improve the equity and safety of mobile phone-based programmes. The present report, the second in this series, instead provides guidance on planning and delivering learning assessments through mobile technology.

The COVID-19 pandemic disrupted education systems around the world, resulting in an unprecedented emergency with school closures affecting 1.6 billion learners (Azevedo et al., 2021). This brought to light the need for resilient education systems that can provide education during emergencies and school closures (UNICEF, 2021). The challenge of reaching and educating children when schooling is interrupted is not new for education in emergencies (EiE) practitioners who develop and implement programmes in complex crises with limited resources. One tool that many families own is a mobile phone. Low-cost basic mobile phones can be used in humanitarian settings to support remote learning and are critical where access to connectivity and higher-cost devices such as laptops are limited. The portability of mobile phones, combined with their communication features, offers multiple uses in EiE. This report explores the use of mobile phones in EiE settings by combining a review of the existing literature with feedback from, and interviews with EiE practitioners on two critical questions: (1) How can basic mobile phones be used to support EiE learning programmes and teacher training?; and (2) What are the key practices undertaken by education practitioners to improve the equity and safety of mobile phone-based education programmes? Implementing an education programme would be incomplete without a learning assessment. Moreover, can mobile phones themselves be used to measure learning in EiE settings? While this question lies beyond the scope of this report, it is investigated in the second report in the series: "On Call: Using Mobile Technologies to Measure Learning in Emergencies" (ED624513). The complementary report outlines the key implementation steps and uses of mobile devices to support learning assessment needs in emergencies.

Thermodynamic properties of fluids confined in nanopores differ from those observed in the bulk. To investigate the effect of nanoconfinement on water compressibility, we performed water sorption experiments on two nanoporous glass samples while concomitantly measuring the speed of longitudinal and shear ultrasonic waves in these samples. These measurements yield the longitudinal and shear moduli of the water laden nanoporous glass as a function of relative humidity that we utilized in the Gassmann theory to infer the bulk modulus of the confined water. This analysis shows that the bulk modulus (inverse of compressibility) of confined water is noticeably higher than that of the bulk water at the same temperature. Moreover, the modulus exhibits a linear dependence on the Laplace pressure. The results for water, which is a polar fluid, agree with previous experimental and numerical data reported for non-polar fluids. This similarity suggests that irrespective of intermolecular forces, confined fluids are stiffer than bulk fluids. Accounting for fluid stiffening in nanopores may be important for accurate interpretation of wave propagation measurements in fluid-filled nanoporous media, including in petrophysics, catalysis, and other applications, such as in porous materials characterization.

The implementation of remote learning?in Latin America and the Caribbean during the COVID-19 school closures?confirmed that the divide?in access to electricity and?technology?remained?a?major hurdle?for governments across the region?to serve all children.?School closures risk widening?existing learning gaps as?private schools were?more prepared to use technology?for remote learning and children from wealthier households?received more support at home?while schools were closed.?As?countries in the region reopen their schools,?it is vital that governments incorporate key?lessons learned?to?improve?the resilience?and equity?of the education systems. This report?presents evidence on remote learning during the COVID-19?school closures in Latin America and the Caribbean?to help guide decision-makers to build more effective, sustainable and resilient education systems for current and future crises.

Countries in West and Central Africa strived to implement national responses to continue learning activities during school closures. These responses relied on a mix of channels, including online platforms, broadcast media, mobile phones and printed learning packs. Several barriers, however, still prevented many children and adolescents in the region from taking advantage of these opportunities, resulting in learning loss in a region where almost 50 per cent of children do not achieve minimum reading skills at the end of the primary cycle. This report builds on existing evidence to highlight key lessons learned in continuing education for all at times of mass school closures and provides actionable recommendations to build resilience into national education systems in view of potential future crises.

The brain computer interface (BCI) is a nonstimulatory direct and occasionally bidirectional communication link between the brain and a computer or an external device. Classically, EEG-based BCI algorithms have relied on models such as support vector machines and linear discriminant analysis or multiclass common spatial patterns. During the last decade, however, more sophisticated machine learning architectures, such as convolutional neural networks, recurrent neural networks, long short-term memory networks and gated recurrent unit networks, have been extensively used to enhance discriminability in multiclass BCI tasks. Additionally, preprocessing and denoising of EEG signals has always been key in the successful decoding of brain activity, and the determination of an optimal and standardized EEG preprocessing activity is an active area of research. In this paper, we present a deep neural network architecture specifically engineered to a) provide state-of-the-art performance in multiclass motor imagery classification and b) remain robust to preprocessing to enable real-time processing of raw data as it streams from EEG and BCI equipment. It is based on the intertwined use of time-distributed fully connected (tdFC) and space-distributed 1D temporal convolutional layers (sdConv) and explicitly addresses the possibility that interaction of spatial and temporal features of the EEG signal occurs at all levels of complexity. Numerical experiments demonstrate that our architecture provides superior performance compared baselines based on a combination of 3D convolutions and recurrent neural networks in a six-class motor imagery network, with a subjectwise accuracy that reaches 99%. Importantly, these results remain unchanged when minimal or extensive preprocessing is applied, possibly paving the way for a more transversal and real-time use of deep learning architectures in EEG classification.

Malagasy adolescents face severe challenges in accessing and completing basic education. Among those students who complete the primary cycle, one in four does not transition into lower secondary school. Economic constraints among vulnerable households coupled with low-quality education result in widespread dropout and poor learning outcomes. Acknowledging these multidimensional barriers, UNICEF Madagascar leveraged funds from the Let Us Learn (LUL) programme to implement a two-pronged strategy to support Malagasy children in accessing and continuing lower secondary school. The Catch-up Classes provide out-of-school adolescents with a learning pathway to build the foundational literacy and numeracy skills they need to resume studying in formal school. Conditional cash transfers target families with children who are at risk of abandoning school after completing the primary cycle. This brief builds on programme monitoring data, impact evaluations and qualitative insights from the field to highlight lessons learnt and actionable recommendations for accessing and continuing vulnerable children's secondary education.

The COVID-19 pandemic has disrupted every aspect of society. In mid-April 2020, 192 countries had closed their schools, putting 9 out of 10 enrolled children out of school. These closures disproportionately affected marginalized children, worsening existing inequities across education systems worldwide. This brief draws on the experience of five UNICEF education country programmes supported by the Let Us Learn (LUL) initiative, to document tangible lessons in adapting education programmes to support the most marginalized children during school and learning centre closures. The evidence in this brief stems from a series of semi-structured interviews with Education and Child Protection specialists, as well as a document review of available COVID-19 response studies, in the five LUL-supported UNICEF Country Offices.

Italy was the first country in Europe to implement a nationwide lockdown. Children and their families lived in nearly complete isolation for almost two months. Students missed 65 days of school compared to an average of 27 missed days among high-income countries worldwide. This prolonged break is of concern, as even short breaks in schooling can cause significant loss of learning for children and lead to educational inequalities over time. At least 3 million Italian students may not have been reached by remote learning due to a lack of internet connectivity or devices at home. This report explores children's and parents' experiences of remote learning during the lockdown in Italy, drawing on data collected from 11 European countries (and coordinated by the European Commission's Joint Research Center). It explores how children's access and use of digital technologies changed during the pandemic; highlights how existing inequalities might undermine remote learning opportunities, even among those with internet access; and provides insights on how to support children's remote learning in the future. [This report was produced with OssCom.]