arXiv subjects

Joao Barros

Publications and source records attributed to Joao Barros.

At least 19 recordsLinked to original sources

arXiv2026

Phoenix: Safe GitHub Issue Resolution via Multi-Agent LLMs

We present Phoenix, a multi-agent LLM system that resolves GitHub issues from triage through pull-request creation, combining seven layered safety controls with a baseline-aware test evaluation strategy. Phoenix decomposes the work across six specialized agents. Planner, reproducer, coder, tester, failure analyst and Pull Request (PR) agent, all coordinated by a label-based GitHub webhook state machine. Every change is checked against a baseline test run before a pull request is opened. On a 24-instance slice of SWE-bench Lite. run on the production webhook path, Phoenix oracle-resolves 75% of instances with no pass-to-pass regressions on successful runs; this curated slice is not directly comparable to full-split leaderboard results, and we discuss the limits of the comparison. A complementary pilot on 42 real issues across 14 repositories yields 100% correctness preservation (CP; mean 122s on the hard tier). Manual inspection shows that about half of the resulting pull requests are well-targeted fixes. The other half place code at incorrect paths, a planner localization limitation we are addressing with retrieval. We also report the deployment failure modes (WAF filtering, token expiry, permission boundaries, flaky CI) that motivated each safety mechanism.

cs.SE↗

arXiv2026

Surviving the Edge: Federated Learning under Networking and Resource Constraints

Motivated by the growing proliferation of federated learning (FL) in edge environments, we present the first systematic characterization of transport-layer breaking points in FL systems operating under conditions of highly constrained network and compute resources. Using a reproducible testbed with chaos engineering tools, we evaluate Flower under progressively degraded network conditions representative of resource-constrained deployments in Africa and similar environments. Our empirical investigation reveals a fundamental mismatch between FL's burst-idle communication pattern and standard TCP connection management. We identify precise operational boundaries: FL training catastrophically fails at 5-second one-way latency due to TCP handshake timeouts, above 50% packet loss due to buffer exhaustion, and with 90% client dropout rates. Through systematic analysis of connection patterns during training rounds, we demonstrate that FL's periodic model update bursts, separated by extended local training periods, violate the assumptions underlying default TCP configurations. To validate the significance of these findings, we show that adjusting just three TCP connection management parameters can significantly reduce training time under extreme latency, proving that transport-layer awareness is not merely beneficial but essential for FL deployment at the network edge. Our characterization methodology and findings provide practitioners with concrete thresholds for determining when standard FL deployments will fail and when advanced reliability techniques become necessary.

cs.NI↗

arXiv2025

Network Hexagons Under Attack: Secure Crowdsourcing of Geo-Referenced Data

A critical requirement for modern-day Intelligent Transportation Systems (ITS) is the ability to collect geo-referenced data from connected vehicles and mobile devices in a safe, secure and anonymous way. The Nexagon protocol, which builds on the IETF Locator/ID Separation Protocol (LISP) and the Hierarchical Hexagonal Clustering (H3) geo-spatial indexing system, offers a promising framework for dynamic, privacy-preserving data aggregation. Seeking to address the critical security and privacy vulnerabilities that persist in its current specification, we apply the STRIDE and LINDDUN threat modelling frameworks and prove among other that the Nexagon protocol is susceptible to user re-identification, session linkage, and sparse-region attacks. To address these challenges, we propose an enhanced security architecture that combines public key infrastructure (PKI) with ephemeral pseudonym certificates. Our solution guarantees user and device anonymity through randomized key rotation and adaptive geospatial resolution, thereby effectively mitigating re-identification and surveillance risks in sparse environments. A prototype implementation over a microservice-based overlay network validates the approach and underscores its readiness for real-world deployment. Our results show that it is possible to achieve the required level of security without increasing latency by more than 25% or reducing the throughput by more than 7%.

cs.CR↗

arXiv2014

Geometry-Based Vehicle-to-Vehicle Channel Modeling for Large-Scale Simulation

Due to the dynamic nature of vehicular traffic and the road surroundings, vehicle-to-vehicle (V2V) propagation characteristics vary greatly on both small- and large-scale. Recent measurements have shown that both large static objects (e.g., buildings and foliage) as well as mobile objects (surrounding vehicles) have a profound impact on V2V communication. At the same time, system-level Vehicular Ad Hoc Network (VANET) simulators by and large employ simple statistical propagation models, which do not account for surrounding objects explicitly. We designed GEMV$^2$ (Geometry-based Efficient propagation Model for V2V communication), which uses outlines of vehicles, buildings, and foliage to distinguish the following three types of links: line of sight (LOS), non-LOS due to vehicles, and non- LOS due to static objects. For each link, GEMV$^2$ calculates the large-scale signal variations deterministically, whereas the small- scale signal variations are calculated stochastically based on the number and size of surrounding objects. We implement GEMV$^2$ in MATLAB and show that it scales well by using it to simulate radio propagation for city-wide networks with tens of thousands of vehicles on commodity hardware. We make the source code of GEMV$^2$ freely available. Finally, we validate GEMV$^2$ against extensive measurements performed in urban, suburban, highway, and open space environment.

cs.NI↗

arXiv2013

TVR - Tall Vehicle Relaying in Vehicular Networks

Vehicle-to-Vehicle (V2V) communication is a core technology for enabling safety and non-safety applications in next generation Intelligent Transportation Systems. Due to relatively low heights of the antennas, V2V communication is often influenced by topographic features, man-made structures, and other vehicles located between the communicating vehicles. On highways, it was shown experimentally that vehicles can obstruct the line of sight (LOS) communication up to 50 percent of the time; furthermore, a single obstructing vehicle can reduce the power at the receiver by more than 20 dB. Based on both experimental measurements and simulations performed using a validated channel model, we show that the elevated position of antennas on tall vehicles improves communication performance. Tall vehicles can significantly increase the effective communication range, with an improvement of up to 50 percent in certain scenarios. Using these findings, we propose a new V2V relaying scheme called Tall Vehicle Relaying (TVR) that takes advantage of better channel characteristics provided by tall vehicles. TVR distinguishes between tall and short vehicles and, where appropriate, chooses tall vehicles as next hop relays. We investigate TVR's system-level performance through a combination of link-level experiments and system-level simulations and show that it outperforms existing techniques.

cs.NI↗

arXiv2013

Contextual Query Using Bell Tests

Tests are essential in Information Retrieval and Data Mining in order to evaluate the effectiveness of a query. An automatic measure tool intended to exhibit the meaning of words in context has been developed and linked with Quantum Theory, particularly entanglement. "Quantum like" experiments were undertaken on semantic space based on the Hyperspace Analogue Language (HAL) method. A quantum HAL model was implemented using state vectors issued from the HAL matrix and query observables, testing a wide range of windows sizes. The Bell parameter S, associating measures on two words in a document, was derived showing peaks for specific window sizes. The peaks show maximum quantum violation of the Bell inequalities and are document dependent. This new correlation measure inspired by Quantum Theory could be promising for measuring query relevance.

cs.IR↗

arXiv2012

Modeling Network Coded TCP: Analysis of Throughput and Energy Cost

We analyze the performance of TCP and TCP with network coding (TCP/NC) in lossy networks. We build upon the framework introduced by Padhye et al. and characterize the throughput behavior of classical TCP and TCP/NC as a function of erasure probability, round-trip time, maximum window size, and duration of the connection. Our analytical results show that network coding masks random erasures from TCP, thus preventing TCP's performance degradation in lossy networks. It is further seen that TCP/NC has significant throughput gains over TCP. In addition, we show that TCP/NC may lead to cost reduction for wireless network providers while maintaining a certain quality of service to their users. We measure the cost in terms of number of base stations, which is highly correlated to the energy, capital, and operational costs of a network provider. We show that increasing the available bandwidth may not necessarily lead to increase in throughput, particularly in lossy networks in which TCP does not perform well. We show that using protocols such as TCP/NC, which are more resilient to erasures, may lead to a throughput commensurate the bandwidth dedicated to each user.

cs.NI↗

arXiv2012

Trade-off between cost and goodput in wireless: Replacing transmitters with coding

We study the cost of improving the goodput, or the useful data rate, to user in a wireless network. We measure the cost in terms of number of base stations, which is highly correlated to the energy cost as well as capital and operational costs of a network provider.We show that increasing the available bandwidth, or throughput, may not necessarily lead to increase in goodput, particularly in lossy wireless networks in which TCP does not perform well. As a result, much of the resources dedicated to the user may not translate to high goodput, resulting in an inefficient use of the network resources. We show that using protocols such as TCP/NC, which are more resilient to erasures and failures in the network, may lead to a goodput commensurate the throughput dedicated to each user. By increasing goodput, users' transactions are completed faster; thus, the resources dedicated to these users can be released to serve other requests or transactions. Consequently, we show that translating efficiently throughput to goodput may bring forth better connection to users while reducing the cost for the network providers.

cs.NI↗

arXiv2011

Wireless Secrecy in Large-Scale Networks

The ability to exchange secret information is critical to many commercial, governmental, and military networks. The intrinsically secure communications graph (iS-graph) is a random graph which describes the connections that can be securely established over a large-scale network, by exploiting the physical properties of the wireless medium. This paper provides an overview of the main properties of this new class of random graphs. We first analyze the local properties of the iS-graph, namely the degree distributions and their dependence on fading, target secrecy rate, and eavesdropper collusion. To mitigate the effect of the eavesdroppers, we propose two techniques that improve secure connectivity. Then, we analyze the global properties of the iS-graph, namely percolation on the infinite plane, and full connectivity on a finite region. These results help clarify how the presence of eavesdroppers can compromise secure communication in a large-scale network.

cs.IT↗

arXiv2010

Algebraic Watchdog: Mitigating Misbehavior in Wireless Network Coding

We propose a secure scheme for wireless network coding, called the algebraic watchdog. By enabling nodes to detect malicious behaviors probabilistically and use overheard messages to police their downstream neighbors locally, the algebraic watchdog delivers a secure global self-checking network. Unlike traditional Byzantine detection protocols which are receiver-based, this protocol gives the senders an active role in checking the node downstream. The key idea is inspired by Marti et al.'s watchdog-pathrater, which attempts to detect and mitigate the effects of routing misbehavior. As an initial building block of a such system, we first focus on a two-hop network. We present a graphical model to understand the inference process nodes execute to police their downstream neighbors; as well as to compute, analyze, and approximate the probabilities of misdetection and false detection. In addition, we present an algebraic analysis of the performance using an hypothesis testing framework that provides exact formulae for probabilities of false detection and misdetection. We then extend the algebraic watchdog to a more general network setting, and propose a protocol in which we can establish trust in coded systems in a distributed manner. We develop a graphical model to detect the presence of an adversarial node downstream within a general multi-hop network. The structure of the graphical model (a trellis) lends itself to well-known algorithms, such as the Viterbi algorithm, which can compute the probabilities of misdetection and false detection. We show analytically that as long as the min-cut is not dominated by the Byzantine adversaries, upstream nodes can monitor downstream neighbors and allow reliable communication with certain probability. Finally, we present simulation results that support our analysis.

cs.CR↗

arXiv2010

Techniques for Enhanced Physical-Layer Security

Information-theoretic security--widely accepted as the strictest notion of security--relies on channel coding techniques that exploit the inherent randomness of propagation channels to strengthen the security of communications systems. Within this paradigm, we explore strategies to improve secure connectivity in a wireless network. We first consider the intrinsically secure communications graph (iS-graph), a convenient representation of the links that can be established with information-theoretic security on a large-scale network. We then propose and characterize two techniques--sectorized transmission and eavesdropper neutralization--which are shown to dramatically enhance the connectivity of the iS-graph.

cs.NI↗

arXiv2010

Modeling Network Coded TCP Throughput: A Simple Model and its Validation

We analyze the performance of TCP and TCP with network coding (TCP/NC) in lossy wireless networks. We build upon the simple framework introduced by Padhye et al. and characterize the throughput behavior of classical TCP as well as TCP/NC as a function of erasure rate, round-trip time, maximum window size, and duration of the connection. Our analytical results show that network coding masks erasures and losses from TCP, thus preventing TCP's performance degradation in lossy networks, such as wireless networks. It is further seen that TCP/NC has significant throughput gains over TCP. In addition, we simulate TCP and TCP/NC to verify our analysis of the average throughput and the window evolution. Our analysis and simulation results show very close concordance and support that TCP/NC is robust against erasures. TCP/NC is not only able to increase its window size faster but also to maintain a large window size despite losses within the network, whereas TCP experiences window closing essentially because losses are mistakenly attributed to congestion.

cs.IT↗

arXiv2010

A Multi-hop Multi-source Algebraic Watchdog

In our previous work "An Algebraic Watchdog for Wireless Network Coding", we proposed a new scheme in which nodes can detect malicious behaviors probabilistically, police their downstream neighbors locally using overheard messages; thus, provide a secure global "self-checking network". As the first building block of such a system, we focused on a two-hop network, and presented a graphical model to understand the inference process by which nodes police their downstream neighbors and to compute the probabilities of misdetection and false detection. In this paper, we extend the Algebraic Watchdog to a more general network setting, and propose a protocol in which we can establish "trust" in coded systems in a distributed manner. We develop a graphical model to detect the presence of an adversarial node downstream within a general two-hop network. The structure of the graphical model (a trellis) lends itself to well-known algorithms, such as Viterbi algorithm, that can compute the probabilities of misdetection and false detection. Using this as a building block, we generalize our scheme to multi-hop networks. We show analytically that as long as the min-cut is not dominated by the Byzantine adversaries, upstream nodes can monitor downstream neighbors and allow reliable communication with certain probability. Finally, we present preliminary simulation results that support our analysis.

cs.CR↗

arXiv2010

Secure Communication in Stochastic Wireless Networks

Information-theoretic security -- widely accepted as the strictest notion of security -- relies on channel coding techniques that exploit the inherent randomness of the propagation channels to significantly strengthen the security of digital communications systems. Motivated by recent developments in the field, this paper aims at a characterization of the fundamental secrecy limits of wireless networks. Based on a general model in which legitimate nodes and potential eavesdroppers are randomly scattered in space, the intrinsically secure communications graph (iS-graph) is defined from the point of view of information-theoretic security. Conclusive results are provided for the local connectivity of the Poisson iS-graph, in terms of node degrees and isolation probabilities. It is shown how the secure connectivity of the network varies with the wireless propagation effects, the secrecy rate threshold of each link, and the noise powers of legitimate nodes and eavesdroppers. Sectorized transmission and eavesdropper neutralization are explored as viable strategies for improving the secure connectivity. Lastly, the maximum secrecy rate between a node and each of its neighbours is characterized, and the case of colluding eavesdroppers is studied. The results help clarify how the spatial density of eavesdroppers can compromise the intrinsic security of wireless networks.

cs.IT↗

arXiv2009

Interfacing network coding with TCP: an implementation

In previous work (`Network coding meets TCP') we proposed a new protocol that interfaces network coding with TCP by means of a coding layer between TCP and IP. Unlike the usual batch-based coding schemes, the protocol uses a TCP-compatible sliding window code in combination with new rules for acknowledging bytes to TCP that take into account the network coding operations in the lower layer. The protocol was presented in a theoretical framework and considered only in conjunction with TCP Vegas. In this paper we present a real-world implementation of this protocol that addresses several important practical aspects of incorporating network coding and decoding with TCP's window management mechanism. Further, we work with the more widespread and practical TCP Reno. Our implementation significantly advances the goal of designing a deployable, general, TCP-compatible protocol that provides the benefits of network coding.

cs.NI↗

arXiv2009

An Algebraic Watchdog for Wireless Network Coding

In this paper, we propose a scheme, called the "algebraic watchdog" for wireless network coding, in which nodes can detect malicious behaviors probabilistically, police their downstream neighbors locally using overheard messages, and, thus, provide a secure global "self-checking network". Unlike traditional Byzantine detection protocols which are receiver-based, this protocol gives the senders an active role in checking the node downstream. This work is inspired by Marti et. al.'s watchdog-pathrater, which attempts to detect and mitigate the effects of routing misbehavior. As the first building block of a such system, we focus on a two-hop network. We present a graphical model to understand the inference process nodes execute to police their downstream neighbors; as well as to compute, analyze, and approximate the probabilities of misdetection and false detection. In addition, we present an algebraic analysis of the performance using an hypothesis testing framework, that provides exact formulae for probabilities of false detection and misdetection.

cs.NI↗

arXiv2009

On Counteracting Byzantine Attacks in Network Coded Peer-to-Peer Networks

Random linear network coding can be used in peer-to-peer networks to increase the efficiency of content distribution and distributed storage. However, these systems are particularly susceptible to Byzantine attacks. We quantify the impact of Byzantine attacks on the coded system by evaluating the probability that a receiver node fails to correctly recover a file. We show that even for a small probability of attack, the system fails with overwhelming probability. We then propose a novel signature scheme that allows packet-level Byzantine detection. This scheme allows one-hop containment of the contamination, and saves bandwidth by allowing nodes to detect and drop the contaminated packets. We compare the net cost of our signature scheme with various other Byzantine schemes, and show that when the probability of Byzantine attacks is high, our scheme is the most bandwidth efficient.

cs.NI↗

arXiv2009

Counteracting Byzantine Adversaries with Network Coding: An Overhead Analysis

Network coding increases throughput and is robust against failures and erasures. However, since it allows mixing of information within the network, a single corrupted packet generated by a Byzantine attacker can easily contaminate the information to multiple destinations. In this paper, we study the transmission overhead associated with three different schemes for detecting Byzantine adversaries at a node using network coding: end-to-end error correction, packet-based Byzantine detection scheme, and generation-based Byzantine detection scheme. In end-to-end error correction, it is known that we can correct up to the min-cut between the source and destinations. However, if we use Byzantine detection schemes, we can detect polluted data, drop them, and therefore, only transmit valid data. For the dropped data, the destinations perform erasure correction, which is computationally lighter than error correction. We show that, with enough attackers present in the network, Byzantine detection schemes may improve the throughput of the network since we choose to forward only reliable information. When the probability of attack is high, a packet-based detection scheme is the most bandwidth efficient; however, when the probability of attack is low, the overhead involved with signing each packet becomes costly, and the generation-based scheme may be preferred. Finally, we characterize the tradeoff between generation size and overhead of detection in bits as the probability of attack increases in the network.

cs.IT↗