ESC

LE Huyen-Trang

PhD Candidate in Computer Networks, LyRIDS (ECE) & CEDRIC (Cnam), Paris

My research is on trustworthy networked systems: networks whose timing can be proven, and data infrastructures whose privacy can be proven. On the networking side, I use deterministic network calculus to certify worst-case latency in Time-Sensitive Networks. On the data side, I build blockchain platforms with zero-knowledge proofs and decentralised storage so that health data records of different specifications, from clinical EHRs to genomic data, can be shared, audited and analysed without exposing patients.

I am a PhD candidate at Cnam and ECE Engineering School in Paris (2026–2029), in the LyRIDS and CEDRIC laboratories. My thesis uses deterministic network calculus to keep worst-case latency guarantees certified while Time-Sensitive Networks are reconfigured at runtime.

Before my PhD, I completed an M2 in Computer Networks at Université d’Évry Paris-Saclay and an engineering degree in Computer Science at HUST, Hanoi. I have worked on blockchain and zero-knowledge proofs for health and genomic data at IBISC, differential privacy for location data at the Universidade do Porto (M1 internship), and deep reinforcement learning for edge-cloud systems at the BKCS lab.

Research areas

View more →
  • AI for Networked Systems

    Deep reinforcement learningMulti-objective optimisationLanguage agentsIntent-based networking

    I use learning where analytical models are too costly: multi-objective deep reinforcement learning for edge-cloud server selection (more than 60% routing improvement over classical heuristics), and language agents that turn clinicians' natural-language intents into orchestration of medical IoT devices over MCP on an SDN hospital network (LLMThings, HOPPERS).

    Technologies: Python, TensorFlow, OpenAI Gym, NumPy, CrewAI, LangChain, LangGraph

  • Differential Privacy for Location Data

    Geo-indistinguishabilityGPS obfuscationLocation privacy-preserving mechanismsMultimodal transportationPrivacy–utility trade-off

    Location-based services collect GPS traces from which homes, workplaces and transport modes can be inferred. During my M1 internship at the University of Porto, I built a context-aware pipeline that first splits real-world trajectories by transport mode, then obfuscates each GPS point with calibrated noise using four geo-indistinguishability mechanisms (Planar Laplace, Adaptive, Clustering and Velocity-Aware Geo-Ind), and finally scores privacy and utility, such as average quality loss in metres. Adapting the mechanism to the context pays off: Velocity-Aware Geo-Ind often stays below 50 m of distortion across all modes, against about 120 m for the Planar Laplace baseline.

    Technologies: Python, Privkit, NumPy, pandas, Matplotlib

  • Blockchain for Health Data Records Privacy

    BlockchainSmart contractsDecentralised storageHealth data recordsAccess controlTraceability

    Health data records come in many specifications: clinical electronic health records, genomic sequences and variants, and biomedical measurements, each with its own size, sensitivity and sharing rules. I design blockchain platforms that keep these records private while letting several parties use them without a central custodian. My M2 internship at IBISC focused on multi-party genomic data sharing; in the Shaker Genopole and D4Gen hackathons I applied the same approach to clinical EHRs on Sui, with access policies and audit trails in Move smart contracts and the encrypted records stored off-chain on Walrus, Sui's decentralised blob storage, referenced on-chain so that every access stays traceable.

    Technologies: Sui, Move, Walrus

  • Zero-Knowledge Proofs

    Zero-knowledge proofsZero-knowledge smart contractsSelective disclosureCommitmentsNullifiersThreshold proofs

    Zero-knowledge proofs let a party prove a statement about private data without revealing the data. I compare ZKP constructions for multi-party genomic data sharing, and use them for selective disclosure of health records (prove a clinical attribute, hide the rest), privacy-preserving aggregated biomedical analysis, and anonymous threshold reporting in SafeProof.

    Technologies: Sui, Move, JavaScript

  • Time-Sensitive Networking

    Deterministic networkingTraffic shapingRuntime reconfigurationIndustrial networks

    I model TSN switches and their shapers (credit-based, time-aware, asynchronous) as service curves, then study how stream admission, rerouting and schedule updates change the worst-case delay of every flow. The goal is a reconfiguration procedure that is certified before it is applied, so latency guarantees hold during and after the change.

    Technologies: Python, Java, NetworkCalculus.org DNC

  • Network Calculus

    Deterministic network calculusArrival curvesService curvesMin-plus algebraWorst-case latencyFormal bounds

    Deterministic network calculus bounds the worst-case delay and backlog of a flow from arrival curves (what a source may send) and service curves (what a node guarantees), using min-plus algebra. Unlike measurement or simulation, which only observe a lower bound on the worst case, these bounds are proofs. I use them as the certification layer of my thesis and maintain open course material on the topic.

    Technologies: Python, Java, NetworkCalculus.org DNC, Quarto

  • IoT and Software-Defined Networking

    Medical IoTSoftware-defined networkingNetwork programmabilityIntent-based orchestrationSmart hospitals

    I design secure infrastructures that collect data from medical IoT devices over software-defined hospital networks, where a central controller programs forwarding and isolation policies. On top of this, language agents translate clinicians' intents into device and network actions through standardised MCP interfaces (HOPPERS, LLMThings). Prototypes are emulated in Mininet-WiFi before any deployment.

    Technologies: Python, FastAPI, React, Next.js, ESP32, ZigBee, Mininet-WiFi, OpenFlow, ONOS

Selected publications

View more →
  1. In preparation for Financial Cryptography and Data Security (FC’27), 2026
  2. Junior Conference on Data Science and Engineering (JDSE 2026), 2026
    Best Presentation Award

Research

View more →

Projects

View more →
LE Huyen-Trang

LE Huyen-Trang

PhD Candidate in Computer Networks

LyRIDS (ECE) & CEDRIC (Cnam), Paris

  • (2029) PhD Networks, Cnam & ECE
  • (2026) M2 Networks, Univ. Évry Paris Saclay
  • (2024) Eng. Computer Science, HUST
More about me →
November 2026
Started my PhD at Cnam and ECE (LyRIDS / CEDRIC laboratories, doctoral school SMI) on certified dynamic reconfiguration of TSN networks
September 2026
Best Presentation Award at JDSE 2026 for LLMThings, shared with Anaïs Sadoudi
September 2026
Attended ACM womENcourage™ 2026 in Sophia Antipolis with an ACM conference scholarship, built SafeProof with our team at the hackathon, and presented a poster on ZKP-based genomic data sharing

Teaching

View more →

Network courses coming soon.

Learning materials

View more →

Contact

Maps →