BGP route hijacks continue to pose a risk to the security and stability of the Internet, while the adoption of mechanisms such as RPKI remains incomplete across Latin America and the Caribbean. At present, the region lacks a system capable of monitoring these incidents from a regional perspective, proactively notifying affected networks, and identifying recurring patterns. BGP-TI LAC proposes an automated regional observatory that uses public data sources to detect and classify anomalous BGP announcements in near real time, incorporating information on ownership, RPKI validation, topology, and historical behavior. The system also provides for notifications to affected networks and the corresponding CSIRTs, together with technical evidence and guidance to strengthen protection through RPKI. The project seeks to turn this evidence into an open, automated, and reusable tool that contributes to improving regional visibility and strengthening routing security across Latin America and the Caribbean.
WAN operators in Latin America and the Caribbean face challenges related to budget constraints, limited availability of specialized technical staff, and dependence on commercial solutions to automate network management. In this context, the use of cloud-based language models can also introduce recurring costs, reliance on external connectivity, and risks related to data sovereignty. This research explores the use of a locally deployed Large Action Model to interpret natural language instructions and translate them into concrete actions on WAN infrastructure. The work will be carried out on a reproducible experimentation platform and will evaluate different levels of complexity, ranging from diagnostics and configuration adjustments to traffic engineering and incident response.
The project aims to develop a decentralized, open-source ecosystem to optimize and protect Wi-Fi networks through Federated Learning techniques implemented directly on low-cost commercial routers running OpenWrt. Each router will be able to learn locally from network behavior without sharing users’ sensitive data, while contributing to the development of a global model that can improve overall network performance, energy efficiency, and security. The solution includes use cases such as optimizing energy consumption and Quality of Experience (QoE), as well as the autonomous detection and mitigation of DoS/DDoS attacks at the edge. Validation is being carried out on university Wi-Fi infrastructures, with the goal of developing accessible technology that can be applied in schools, community centers, and community networks across Latin America.
Ransomware and DDoS attacks are among the leading threats to digital security worldwide and are becoming increasingly prevalent in Latin America, where organizations and service providers also face limitations in accessing advanced protection mechanisms. The research focuses on the use of Artificial Intelligence to improve the detection of these types of attacks, combining Machine Learning, Deep Learning, and generative AI techniques. The work will also use real traffic data and seek to validate the results in environments directly connected to industry. The initiative brings together academic and industry capabilities from Mexico and Brazil, with the goal of applying research results in real-world scenarios and helping reduce disruptions, financial losses, and risks associated with DDoS and ransomware attacks.
The growing number of orders to block websites and applications is creating new challenges for Internet stability in Latin America. When such measures are implemented directly at the IP address level, they may unintentionally affect numerous legitimate services that share the same infrastructure. This research seeks to measure and understand this collateral impact through observations carried out across different Internet service providers, identifying how blocking measures are implemented, where they occur, and what effects they have on other content and services. Based on this evidence, the project aims to improve visibility into these interventions and provide technical information that can contribute to dialogue between operators and authorities, supporting more precise measures with less impact on Internet stability and resilience across the region.
BGP routing security remains a challenge for Internet service providers, particularly those with more limited technical and financial resources. Across the LACNIC region, adoption of mechanisms such as RPKI/ROV and ASPA remains uneven, while the transition toward post-quantum cryptographic technologies introduces new challenges for network infrastructure. PRTR seeks to help reduce these barriers through an open and accessible platform that enables routing security mechanisms to be tested, validated, and deployed in real-world environments. The project will combine BGP validation testing, assessments of post-quantum technologies, and operator-oriented documentation, with a particular focus on the needs of small and medium-sized ISPs in Latin America and the Caribbean.
BGP filters are essential for preventing route leaks and prefix hijacks, but misconfigured or outdated filters can unintentionally block legitimate route propagation and are often difficult to detect until a network disruption occurs. Because BGP policies are largely opaque, these errors may remain hidden while traffic continues to reach its destination through alternative paths. This project will develop techniques that allow network operators to actively test and validate BGP filters by manipulating route advertisements and controlling how prefixes propagate through upstream networks. Using mechanisms such as selective advertisements, BGP communities, and BGP poisoning, the research will seek to identify incorrect filtering not only at direct providers but also several AS hops away. The techniques will be evaluated in real-world scenarios through the PEERING Testbed, and the resulting tool will be packaged for use by network operators in their own environments. The project will also seek collaboration with South American operators, IXPs, and research institutions to expand PEERING’s presence and improve visibility of networks across the LACNIC region.
The project aims to develop an open-source software architecture to integrate L4S technology into Wi-Fi routers and GPON terminals, with the goal of reducing latency in wireless access networks. The solution will incorporate intelligent congestion management mechanisms based on Reinforcement Learning, capable of dynamically adapting to variations in Wi-Fi links. As a result, a Proof of Concept will be developed and validated, together with OpenWRT-compatible modules, seeking to provide Internet service providers with an open alternative that requires low computational resources and avoids vendor lock-in.
In this project, Multiview AI algorithms will be explored and implemented that have the ability to generate levels of uncertainty in their predictions in such a way that it is possible to quantify how reliable a given prediction is. This way, if a prediction is known to be unreliable, an expert can be notified to make the final decision. Multiview AI algorithms with reliability quantification will be tested in different scenarios relevant to ISP, IXP, and IoT systems, including anomaly detection and device attacks. Specifically, we will seek to combine specific protocol characteristics, statistics based on flows, contents, and packet-level behaviors, with the objective of building more efficient and discriminative multiview representations for supervised learning tasks.
The RedLLM-Integrator project proposes a solution to strengthen the capabilities of Security Operations Centers (SOCs) in Latin America through the progressive and ethical integration of explanatory artificial intelligence tools into their existing processes. Instead of training a new model from scratch – which implies high costs, technological dependence and advanced knowledge in AI infrastructure – the proposal focuses on leveraging already established architectures such as Retrieval-Augmented Generation (RAG) and pre-trained Large Language Models (LLMs) to support offensive simulation, identification of vulnerabilities and the generation of understandable recommendations in digital security environments.