The rapid growth of the Internet has raised concerns about the carbon emissions linked to data transmission. Autonomous Systems (ASes), which make inter-domain routing decisions, influence how traffic moves through the network and, indirectly, where emissions are generated. Although carbon-aware routing is gaining interest, limited visibility into emission sources remains a key challenge. In this paper, we present GRASS (Green Ranking of Autonomous SystemS), a framework for estimating the CO2intensity of ASes based on their geographic distribution and the carbon efficiency of regional electricity grids. GRASS combines geolocation, routing, and carbon data to infer probabilistic location profiles and compute a CO2intensity score for each AS. Using GRASS, we analyze emission patterns across AS popularity tiers, and customer cone sizes, and assess the carbon intensity of AS-to-AS links based on associated organizations. Our results show that a small number of structurally central ASes and links account for a large fraction of total Internet-related emissions. We propose a novel approach to assessing the CO2intensity of ASes and their interconnections, offering insights that enable targeted interventions for greener network operations and routing policies.

Remote peering (RP) enables networks to reach Internet Exchange Points (IXPs) without physical presence in the same datacenter/location, offering cost-effective interconnection. However, in the context of anycast, where routing efficiency relies on the assumption that traffic is routed to the geographically nearest instance, RP can undermine this principle by introducing hidden performance costs and inflating paths. This paper presents a hybrid methodology, combining data-plane measurements with control-plane insights, to detect and geolocate RP links at scale using IP-level geolocation, AS-level mapping, and IXP facility data. Using 3 million traceroutes targeting 13,735 anycast /24-prefixes, we quantify the impact of RP on anycast routing.Our analysis shows that whilst RP is widespread, it often connects ASes geographically close to the IXP facility. However, for cases where it is “remote” we find it frequently results in long detours, elevated latency, and sub-optimal anycast site selection. Finally, we identify key ASes and IXP regions that play a central role in driving RP-related detours and locality violations in anycast routing. Our findings demonstrate at Internet scale that RP undermines anycast locality by increasing the likelihood of detours and inflated paths, underscoring the need for improved visibility into interconnection practices that impact latency-sensitive services.

The Border Gateway Protocol (BGP), a cornerstone of global Internet routing, remains vulnerable to various attacks due to its lack of integrated security features and mechanisms to verify the authenticity and integrity of routing information. Although BGPsec was introduced as a standardized solution to address these concerns, it has not seen real-world deployment even after eight years, primarily due to operational complexity and deployment challenges. We present the design and implementation of xBGPsec, a protocol that improves BGP security by attaching digital signatures to BGP updates using optional transitive attributes. These signatures cover both the AS path and dedicated BGP attributes, and are generated and verified by a dedicated external validator, allowing enhanced security without disrupting existing routing operations. To demonstrate the feasibility of this approach, we built a dedicated testbed integrating xBGPsec-enabled routers and a supporting cryptographic infrastructure. This proof-of-concept provides the foundation for future empirical evaluation and offers initial insights for network operators and researchers exploring scalable, interoperable BGP security solutions.

In 2003, Wang and Gao [67] presented an algorithm to infer and characterize routing policies as this knowledge could be valuable in predicting and debugging routing paths. They used their algorithm to measure the phenomenon of selectively announced prefixes, in which, ASes would announce their prefixes to specific providers to manipulate incoming traffic. Since 2003, the Internet has evolved from a hierarchical graph, to a flat and dense structure. Despite 20 years of extensive research since that seminal work, the impact of these topological changes on routing policies is still blurred.

The Border Gateway Protocol (BGP) is a policy-based protocol, which enables Autonomous Systems (ASes) to independently define their routing policies with little or no global coordination. AS-level topology and AS-level paths inference have been long-standing problems for the past two decades, yet, an important question remains open: "which elements of Internet routing affect the AS-path inference accuracy and how much do they contribute to the error?". In this work, we: (1) identify the confounding factors behind Internet routing modeling, and (2) quantify their contribution on the inference error. Our results indicate that by solving the first-hop inference problem, we can increase the exact-path score from 33.6% to 84.1%, and, by taking geolocation into consideration, we can refine the accuracy up to 94.6%.

Joint caching and recommendation has been recently proposed as a new paradigm for increasing the efficiency of mobile edge caching. Early findings demonstrate significant gains for the network performance. However, previous works evaluated the proposed schemes exclusively on simulation environments. Hence, it still remains uncertain whether the claimed benefits would change in real settings. In this paper, we propose a methodology that enables to evaluate joint network and recommendation schemes in real content services by only using publicly available information. We apply our methodology to the YouTube service, and conduct extensive measurements to investigate the potential performance gains. Our results show that significant gains can be achieved in practice; e.g., 8 to 10 times increase in the cache hit ratio from cache-aware recommendations. Finally, we build an experimental testbed and conduct experiments with real users; we make available our code and datasets to facilitate further research. To our best knowledge, this is the first realistic evaluation (over a real service, with real measurements and user experiments) of the joint caching and recommendations paradigm. Our findings provide experimental evidence for the feasibility and benefits of this paradigm, validate assumptions of previous works, and provide insights that can drive future research.