Searcharxiv⌕ Search

arXiv subjects

Richard Mortier

Publications and source records attributed to Richard Mortier.

At least 19 recordsLinked to original sources
arXiv2026

Model Multiplicity for Adversarial Detection in Small Language Model Training on Edge Devices

The rise of edge-based machine learning has enabled distributed adaptation of language models across mobile and IoT devices, offering privacy preservation and real-time responsiveness. However, distributed fine-tuning of language models on untrusted or heterogeneous edge nodes introduces new vulnerabilities. Compromised or unreliable devices can inject poisoned updates, leading to stealthy model manipulation or convergence degradation. Classical defenses such as robust aggregation or temporal anomaly detection operate on a single global model and are therefore limited in detecting coordinated or persistent poisoning. This work proposes a new system-level defense based on model multiplicity. Instead of maintaining one global model, the system rotates or concurrently trains multiple small language models (e.g., DistilGPT-2), each updated by independently sampled subsets of edge nodes. These models evolve under distinct training trajectories, creating multiple independent views of the same distributed population. Divergence between models quantified through gradient similarity, loss evolution, or parameter variance serves as a signal of anomalous or adversarial behavior. When one model deviates significantly from the ensemble mean, the system flags its contributing nodes for isolation or re-weighting. We implement this framework and evaluate it on edge-scale simulations of Small Language Model (SLM) training under varying heterogeneity and attack conditions. Results show that model multiplicity enables earlier and more reliable detection of poisoning compared to classical single-model defenses such as Flanders and Robust methods. Our findings demonstrate that diversity in model evolution can serve as a practical and effective defense mechanism for secure distributed learning on resource-constrained edge devices.

cs.CR↗
arXiv2026

Robust Synchronisation for Federated Learning in The Face of Correlated Device Failure

Probabilistic Synchronous Parallel (PSP) is a technique in distributed learning systems to reduce synchronization bottlenecks by sampling a subset of participating nodes per round. In Federated Learning (FL), where edge devices are often unreliable due to factors including mobility, power constraints, and user activity, PSP helps improve system throughput. However, PSP has a key limitation: it assumes device behavior is static and different devices are independent. This can lead to unfair distributed synchronization, due to highly available nodes dominating training while those that are often unavailable rarely participate and so their data may be missed. If both data distribution and node availability are simultaneously correlated with the device, then both PSP and standard FL algorithms will suffer from persistent under-representation of certain classes or groups resulting in inefficient or ineffective learning of certain features. We introduce Availability-Weighted PSP (AW-PSP), an extension to PSP that addresses the issue of co-correlation of unfair sampling and data availability by dynamically adjusting node sampling probabilities using real-time availability predictions, historical behavior, and failure correlation metrics. A Markov-based availability predictor distinguishes transient \emph{vs} chronic failures, while a Distributed Hash Table (DHT) layer decentralizes metadata, including latency, freshness, and utility scores. We implement AW-PSP and trace-driven evaluation shows that it improves robustness to both independent and correlated failures, increases label coverage, and reduces fairness variance compared to standard PSP. AW-PSP thus provides an availability-aware, and fairness-conscious node sampling protocol for FL deployments that will scale to large numbers of nodes even in heterogeneous and failure-prone environments.

cs.DC↗
arXiv2026

Cumulative Utility Parity for Fair Federated Learning under Intermittent Client Participation

In real-world federated learning (FL) systems, client participation is intermittent, heterogeneous, and often correlated with data characteristics or resource constraints. Existing fairness approaches in FL primarily focus on equalizing loss or accuracy conditional on participation, implicitly assuming that clients have comparable opportunities to contribute over time. However, when participation itself is uneven, these objectives can lead to systematic under-representation of intermittently available clients, even if per-round performance appears fair. We propose cumulative utility parity, a fairness principle that evaluates whether clients receive comparable long-term benefit per participation opportunity, rather than per training round. To operationalize this notion, we introduce availability-normalized cumulative utility, which disentangles unavoidable physical constraints from avoidable algorithmic bias arising from scheduling and aggregation. Experiments on temporally skewed, non-IID federated benchmarks demonstrate that our approach substantially improves long-term representation parity, while maintaining near-perfect performance.

cs.LG↗
arXiv2025

What if we could hot swap our Biometrics?

What if you could really revoke your actual biometric identity, and install a new one, by live rewriting your biological self? We propose some novel mechanisms for hot swapping identity based in novel biotechnology. We discuss the potential positive use cases, and negative consequences if such technology was to become available and affordable. Biometrics are selected on the basis that they are supposed to be unfakeable, or at least not at reasonable cost. If they become easier to fake, it may be much cheaper to fake someone else's biometrics than it is for you to change your own biometrics if someone does copy yours. This potentially makes biometrics a bad trade-off for the user. At the time of writing, this threat is highly speculative, but we believe it is worth raising and considering the potential consequences.

cs.CR↗
arXiv2025

Mitigating context switching in densely packed Linux clusters with Latency-Aware Group Scheduling

Cluster orchestrators such as Kubernetes depend on accurate estimates of node capacity and job requirements. Inaccuracies in either lead to poor placement decisions and degraded cluster performance. In this paper, we show that in densely packed workloads, such as serverless applications, CPU context switching overheads can become so significant that a node's performance is severely degraded, even when the orchestrator placement is theoretically sound. In practice this issue is typically mitigated by over-provisioning the cluster, leading to wasted resources. We show that these context switching overhead arise from both an increase in the average cost of an individual context switch and a higher rate of context switching, which together amplify overhead multiplicatively when managing large numbers of concurrent cgroups, Linux's group scheduling mechanism for managing multi-threaded colocated workloads. We propose and evaluate modifications to the standard Linux kernel scheduler that mitigate these effects, achieving the same effective performance with a 28% smaller cluster size. The key insight behind our approach is to prioritise task completion over low-level per-task fairness, enabling the scheduler to drain contended CPU run queues more rapidly and thereby reduce time spent on context switching.

cs.OS↗
arXiv2024

LSKV: A Confidential Distributed Datastore to Protect Critical Data in the Cloud

Software services are increasingly migrating to the cloud, requiring trust in actors with direct access to the hardware, software and data comprising the service. A distributed datastore storing critical data sits at the core of many services; a prime example being etcd in Kubernetes. Trusted execution environments can secure this data from cloud providers during execution, but it is complex to build trustworthy data storage systems using such mechanisms. We present the design and evaluation of the Ledger-backed Secure Key-Value datastore (LSKV), a distributed datastore that provides an etcd-like API but can use trusted execution mechanisms to keep cloud providers outside the trust boundary. LSKV provides a path to transition traditional systems towards confidential execution, provides competitive performance compared to etcd, and helps clients to gain trust in intermediary services. LSKV forms a foundational core, lowering the barriers to building more trustworthy systems.

cs.DC↗
arXiv2024

Reducing Tail Latencies Through Environment- and Neighbour-aware Thread Management

Application tail latency is a key metric for many services, with high latencies being linked directly to loss of revenue. Modern deeply-nested micro-service architectures exacerbate tail latencies, increasing the likelihood of users experiencing them. In this work, we show how CPU overcommitment by OS threads leads to high tail latencies when applications are under heavy load. CPU overcommitment can arise from two operational factors: incorrectly determining the number of CPUs available when under a CPU quota, and the ignorance of neighbour applications and their CPU usage. We discuss different languages' solutions to obtaining the CPUs available, evaluating the impact, and discuss opportunities for a more unified language-independent interface to obtain the number of CPUs available. We then evaluate the impact of neighbour usage on tail latency and introduce a new neighbour-aware threadpool, the friendlypool, that dynamically avoids overcommitment. In our evaluation, the friendlypool reduces maximum worker latency by up to $6.7\times$ at the cost of decreasing throughput by up to $1.4\times$.

cs.DC↗
arXiv2024

Offline Energy-Optimal LLM Serving: Workload-Based Energy Models for LLM Inference on Heterogeneous Systems

The rapid adoption of large language models (LLMs) has led to significant advances in natural language processing and text generation. However, the energy consumed through LLM model inference remains a major challenge for sustainable AI deployment. To address this problem, we model the workload-dependent energy consumption and runtime of LLM inference tasks on heterogeneous GPU-CPU systems. By conducting an extensive characterization study of several state-of-the-art LLMs and analyzing their energy and runtime behavior across different magnitudes of input prompts and output text, we develop accurate (R^2>0.96) energy and runtime models for each LLM. We employ these models to explore an offline, energy-optimal LLM workload scheduling framework. Through a case study, we demonstrate the advantages of energy and accuracy aware scheduling compared to existing best practices.

cs.DC↗
arXiv2024

Hybrid Heterogeneous Clusters Can Lower the Energy Consumption of LLM Inference Workloads

Both the training and use of Large Language Models (LLMs) require large amounts of energy. Their increasing popularity, therefore, raises critical concerns regarding the energy efficiency and sustainability of data centers that host them. This paper addresses the challenge of reducing energy consumption in data centers running LLMs. We propose a hybrid data center model that uses a cost-based scheduling framework to dynamically allocate LLM tasks across hardware accelerators that differ in their energy efficiencies and computational capabilities. Specifically, our workload-aware strategy determines whether tasks are processed on energy-efficient processors or high-performance GPUs based on the number of input and output tokens in a query. Our analysis of a representative LLM dataset, finds that this hybrid strategy can reduce CPU+GPU energy consumption by 7.5% compared to a workload-unaware baseline.

cs.DC↗
arXiv2024

Locally Differentially Private Embedding Models in Distributed Fraud Prevention Systems

Global financial crime activity is driving demand for machine learning solutions in fraud prevention. However, prevention systems are commonly serviced to financial institutions in isolation, and few provisions exist for data sharing due to fears of unintentional leaks and adversarial attacks. Collaborative learning advances in finance are rare, and it is hard to find real-world insights derived from privacy-preserving data processing systems. In this paper, we present a collaborative deep learning framework for fraud prevention, designed from a privacy standpoint, and awarded at the recent PETs Prize Challenges. We leverage latent embedded representations of varied-length transaction sequences, along with local differential privacy, in order to construct a data release mechanism which can securely inform externally hosted fraud and anomaly detection models. We assess our contribution on two distributed data sets donated by large payment networks, and demonstrate robustness to popular inference-time attacks, along with utility-privacy trade-offs analogous to published work in alternative application domains.

cs.CR↗
arXiv2023

FedSZ: Leveraging Error-Bounded Lossy Compression for Federated Learning Communications

With the promise of federated learning (FL) to allow for geographically-distributed and highly personalized services, the efficient exchange of model updates between clients and servers becomes crucial. FL, though decentralized, often faces communication bottlenecks, especially in resource-constrained scenarios. Existing data compression techniques like gradient sparsification, quantization, and pruning offer some solutions, but may compromise model performance or necessitate expensive retraining. In this paper, we introduce FedSZ, a specialized lossy-compression algorithm designed to minimize the size of client model updates in FL. FedSZ incorporates a comprehensive compression pipeline featuring data partitioning, lossy and lossless compression of model parameters and metadata, and serialization. We evaluate FedSZ using a suite of error-bounded lossy compressors, ultimately finding SZ2 to be the most effective across various model architectures and datasets including AlexNet, MobileNetV2, ResNet50, CIFAR-10, Caltech101, and Fashion-MNIST. Our study reveals that a relative error bound 1E-2 achieves an optimal tradeoff, compressing model states between 5.55-12.61x while maintaining inference accuracy within <0.5% of uncompressed results. Additionally, the runtime overhead of FedSZ is <4.7% or between of the wall-clock communication-round time, a worthwhile trade-off for reducing network transfer times by an order of magnitude for networks bandwidths <500Mbps. Intriguingly, we also find that the error introduced by FedSZ could potentially serve as a source of differentially private noise, opening up new avenues for privacy-preserving FL.

cs.DC↗
arXiv2023

Mutating etcd Towards Edge Suitability

In the edge environment servers are no longer being co-located away from clients, instead they are being co-located with clients away from other servers, focusing on reliable and performant operation. Orchestration platforms, such as Kubernetes, are a key system being transitioned to the edge but they remain unsuited to the environment, stemming primarily from their critical key-value stores. In this work we derive requirements from the edge environment showing that, fundamentally, the design of distributed key-value datastores, such as etcd, is unsuited to meet them. Using these requirements, we explore the design space for distributed key-value datastores and implement two successive mutations of etcd for different points: mergeable-etcd and dismerge, trading linearizability for causal consistency based on CRDTs. mergeable-etcd retains the linear revision history but encounters inherent shortcomings, whilst dismerge embraces the causal model. Both stores are local-first, maintaining reliable performance under network partitions and variability, drastically surpassing etcd's performance, whilst maintaining competitive performance in reliable settings.

cs.DC↗
arXiv2023

Mitigating IoT Botnet DDos Attacks through MUD and eBPF based Traffic Filtering

As the prevalence of Internet-of-Things (IoT) devices becomes more and more dominant, so too do the associated management and security challenges. One such challenge is the exploitation of vulnerable devices for recruitment into botnets, which can be used to carry out Distributed Denial-of-Service (DDoS) attacks. The recent Manufacturer Usage Description (MUD) standard has been proposed as a way to mitigate this problem, by allowing manufacturers to define communication patterns that are permitted for their IoT devices, with enforcement at the gateway home router. In this paper, we present a novel integrated system implementation that uses a MUD manager (osMUD) to parse an extended set of MUD rules, which also allow for rate-limiting of traffic and for setting appropriate thresholds. Additionally, we present two new backends for MUD rule enforcement, one based on eBPF and the other based on the Linux standard iptables. The reported evaluation results show that these techniques are feasible and effective in protecting against attacks and in terms of their impact on legitimate traffic and on the home gateway.

cs.NI↗
arXiv2022

CDBB West Cambridge Digital Twin: Lessons Learned

The report describes the digital architecture developed for the West Cambridge Digital Twin, particularly focussed on real-time sensor data collection and analysis with a privacy framework allowing occupants of the buildings to be first-class participants in the system. The implementation has some notable characteristics. In particular 'push' technology is used throughout such that information streams from the incoming asynchronous individual sensor events through to the end-user web pages with the absolute minimum latency, including real-time generated simple and complex events derived from the the underlying sensor data and the updating of visualisations such as an in-building heatmap. We believe the ability of the entire system to respond in the timescale of individual sensor messages to be unique. JSON structures are used to represent all data types including sensor readings, sensor types, building objects, organisations and people, with the idea that JSON-LD may represent a more suitable way than XML/RDF for managing relations between those objects (such as the 'occupies' relationship of people to offices, or the 'type' relationship of sensors to sensor types).

eess.SY↗
arXiv2022

Distributed data analytics

Machine Learning (ML) techniques have begun to dominate data analytics applications and services. Recommendation systems are a key component of online service providers. The financial industry has adopted ML to harness large volumes of data in areas such as fraud detection, risk-management, and compliance. Deep Learning is the technology behind voice-based personal assistants, etc. Deployment of ML technologies onto cloud computing infrastructures has benefited numerous aspects of our daily life. The advertising and associated online industries in particular have fuelled a rapid rise the in deployment of personal data collection and analytics tools. Traditionally, behavioural analytics relies on collecting vast amounts of data in centralised cloud infrastructure before using it to train machine learning models that allow user behaviour and preferences to be inferred. A contrasting approach, distributed data analytics, where code and models for training and inference are distributed to the places where data is collected, has been boosted by two recent, ongoing developments: increased processing power and memory capacity available in user devices at the edge of the network, such as smartphones and home assistants; and increased sensitivity to the highly intrusive nature of many of these devices and services and the attendant demands for improved privacy. Indeed, the potential for increased privacy is not the only benefit of distributing data analytics to the edges of the network: reducing the movement of large volumes of data can also improve energy efficiency, helping to ameliorate the ever increasing carbon footprint of our digital infrastructure, enabling much lower latency for service interactions than is possible when services are cloud-hosted. These approaches often introduce challenges in privacy, utility, and efficiency trade-offs, while having to ensure fruitful user engagement.

cs.DC↗
arXiv2022

Human-centred home network security

This chapter draws from across the foregoing chapters discussing many core HDI approaches and disciplinary perspectives to consider the specific application of HDI in home network security. While much work has considered the challenges of securing in home IoT devices and their communications, especially for those with limited power or computational capacity, scant attention has been paid by the research community to home network security, and its acceptability and usability, from the viewpoint of ordinary citizens. It will be clear that we need a radical transformation in our approach to designing domestic networking infrastructure to guard against widespread cyber-attacks that threaten to counter the benefits of the IoT. Our aim has to be to defend against enemies inside the walls, to protect critical functionality in the home against rogue devices and prevent the proliferation of disruptive wide-scale IoT DDOS attacks that are already occurring [1].

cs.CR↗
arXiv2022

Relaxed Paxos: Quorum Intersection Revisited (Again)

Distributed consensus, the ability to reach agreement in the face of failures, is a fundamental primitive for constructing reliable distributed systems. The Paxos algorithm is synonymous with consensus and widely utilized in production. Paxos uses two phases: phase one and phase two, each requiring a quorum of acceptors, to reach consensus during a round of the protocol. Traditionally, Paxos requires that all quorums, regardless of phase or round, intersect and majorities are often used for this purpose. Flexible Paxos proved that it is only necessary for phase one quorum of a given round to intersect with the phase two quorums of all previous rounds. In this paper, we re-examine how Paxos approaches the problem of consensus. We look again at quorum intersection in Flexible Paxos and observe that quorum intersection can be safely weakened further. Most notably, we observe that if a proposer learns that a value was proposed in some previous round then its phase one no longer needs to intersect with the phase two quorums from that round or from any previous rounds. Furthermore, in order to provide an intuitive explanation of our results, we propose a novel abstraction for reasoning about Paxos which utilizes write-once registers.

cs.DC↗