https://hdl.handle.net/20.500.12761/2 2026-09-09T16:31:47Z https://hdl.handle.net/20.500.12761/2067 Rent-a-RAG: Embedding-Space Watermarks for Auditing Third-Party RAG Goultiaev Tolstokorov, Alexandr; Mouratidis, Kyriakos; Dogani, Javad; Laoutaris, Nikolaos Third-party retrieval-augmented generation (RAG) marketplaces create a new auditing problem: data providers may license corpora to a RAG operator, yet later have no visibility into whether their documents are being reused without compensation. Auditing this misuse is difficult because the operator is non-cooperative, answers are paraphrased by the generator, and one response may combine evidence from many providers. We propose DirBucket, a provider-side semantic watermarking and black-box auditing framework for document-level reuse in multi-provider RAG. DirBucket watermarks documents by meaning-preserving paraphrases whose embeddings are biased toward provider-bucket secret directions, enabling detection from black-box answers while preserving retrieval utility. On a challenging benchmark that reflects mixed-provider reuse under black-box access, DirBucket is the only method that consistently achieves strong target detection with no non-target activation, detecting non-compliance in every audit within 23 audited answers on our primary benchmark. The watermark survives adversarial post-answer laundering, and none of the evaluated evasion strategies simultaneously defeats detection while preserving user-perceived answer quality. Detection transfers unchanged to a second benchmark built from real clinical, cyber-threat-intelligence, and legal provider corpora. These results suggest that embedding-space watermarking can make document reuse in third-party RAG statistically auditable. 2026-10-01T00:00:00Z https://hdl.handle.net/20.500.12761/2066 LARS: Keeping Local Traffic Local with Latency-Aware Route Servers at IXPs de Andrés Hernández, David; Alhamwy, Yasin; Wagner, Daniel; Stephan, Maximilian; Wichtlhuber, Matthias; Hohlfeld, Oliver Internet Exchange Points facilitate traffic exchange through Route Servers that use BGP for path selection without considering latency. This paper presents LARS, the first system to enable latency-aware path optimization at IXP route servers. It requires no changes to BGP and no bilateral coordination among members, thus directly enabling latency-aware routing for all IXP members. LARS continuously probes the route-server RIB, annotates routes with RTT-derived BGP communities, and enables members to (i) enforce a configurable latency radius that suppresses high-latency routes and (ii) prefer lower-latency alternatives when multiple paths exist. We evaluate LARS at a very large IXP using a production-level prototype that demonstrates stable latency estimation, effective tail-latency control via selective redistribution, and potential to optimize BGP best paths. 2026-08-11T00:00:00Z https://hdl.handle.net/20.500.12761/2065 Opening Wide, Moving Fast: Polymarket Spread Dynamics Jörnell, Carl; Perez, Vicente; Saguillo, Oriol Polymarket creates a new market for every event it lists, and each contract begins with an empty order book: no inherited orders, no assigned market maker, and no opening auction. Every contract there- fore has a clear opening moment and an early period during which its trading conditions are still forming. Using roughly 160,000 contracts cre- ated between October 2025 and March 2026, we study what it costs to trade in a market’s first hours of life and how that cost behaves across different kinds of events. Over the first twenty-four hours of each market, the genesis window, we measure how wide spreads start and how quickly they narrow, how much it actually costs to trade immediately, and how much of that cost stays with the liquidity provider versus reflecting the price moving after- wards. We compare these patterns across eight event categories. Trading costs are strongly category-dependent from the very first trade: sports markets open cheaply with deep, balanced books, while other categories open much wider and take longer to settle, with most of the narrowing happening within the first hour. Centrally, we document a dissociation: ranking categories by how expensive they are to trade and by how much early trades move the price produces two orderings that do not agree. These are separate properties, not two views of the same thing 2026-09-01T00:00:00Z https://hdl.handle.net/20.500.12761/2064 Distributing Inference in the User Plane of Complex Network Topologies with DUNE Bütün, Beyza; de Andrés Hernández, David; Carré, Alexis; Gucciardo, Michele; Fiore, Marco; Bütün, Beyza The deployment of Machine Learning (ML) models in the user plane has emerged as a promising approach to enable line-rate in-network inference, thereby reducing end-to-end latency and improving the scalability of network functions such as network telemetry. Nevertheless, integrating ML models into programmable switches remains challenging due to stringent memory and computational constraints. Prior work has predominantly focused on deploying monolithic ML models into individual programmable network devices, an approach fundamentally limited by hardware resources, particularly for complex classification tasks. In this paper, we introduce DUNE, a novel framework that, for the first time, enables distributed user-plane inference across multiple programmable network devices. DUNE employs original, fully automated techniques to (i) decompose large ML models into lightweight sub-models that retain inference accuracy while reducing resource consumption, and (ii) determine the design, sequencing, and placement of these sub-models to support efficient, distributed joint packet- and flow-level inference. We deploy P4 implementations of DUNE both in an experimental testbed with industry-grade programmable switches and in an emulation environment. We leverage the former setup to demonstrate the practical viability of the solution with real-world hardware in a simple linear topology, and the latter to evaluate DUNE’s scalability to larger and more complex network topologies coexisting with routing strategies. In both cases, we evaluate the framework using real-world traffic on two challenging classification tasks: our results show that DUNE not only reduces per-switch resource usage compared to traditional monolithic ML deployments but also improves inference accuracy by up to 7.5%. 2026-01-01T00:00:00Z