Skip to content

Repository files navigation

AI/ML WiFi Sensing Hub

An open research hub mapping AI/ML WiFi sensing papers, datasets, code, reproducibility, clinical-safety metrics, and security gaps, starting with healthcare-relevant sensing.

Focus First Category Security Clinical Disclaimer


Overview

The AI/ML WiFi Sensing Hub is an open research mapping initiative focused on organizing scientific evidence around WiFi sensing systems powered by artificial intelligence and machine learning.

The hub maps research papers, datasets, code releases, reproducibility notes, clinical-safety translation metrics, security relevance, and open research gaps across WiFi-based sensing applications. The first focus area is healthcare-relevant sensing, including contactless monitoring, activity recognition, fall detection, respiration estimation, vital-sign-related sensing, and care-aware environments.

The long-term goal is to make the AI/ML WiFi sensing research landscape easier to understand, compare, reproduce, evaluate, and extend.


Repository Structure

Path Purpose
backbone/ Curated backbone evidence layer for conceptual papers, reproducibility resources, clinical-safety translation references, and adversarial-impact papers
clinical_safety_metrics/ Clinically motivated safety-translation metrics for fall detection and future HR/RR sensing evaluation
papers/ Paper review templates and literature organization
datasets/ Dataset review templates, access notes, and dataset tracking
code_releases/ Code repository review templates and reproducibility notes
security_robustness/ Security, privacy, robustness, and adversarial-risk mapping
collaboration/ Research groups, labs, authors, and collaboration candidates
evidence_map/ Structured CSV evidence maps for papers, datasets, code, security gaps, clinical-safety translation, and collaboration leads
ROADMAP.md Development roadmap for the hub
CONTRIBUTING.md Contribution guide for suggesting papers, datasets, code, and collaboration ideas
CITATION.cff Citation metadata for this repository

Backbone Evidence Layer

The hub includes a curated Backbone Evidence Layer that organizes the most important references into four categories:

Category Purpose
Conceptual Backbone Papers Proof-of-concept papers that justify the research problem, threat model, security gap, safety framing, or defense theory
Reproducibility Backbone Resources Datasets, code repositories, benchmark papers, and experimental resources that support verification, reproduction, or claim checking
Clinical-Safety Translation Backbone Clinical and patient-safety references that justify HR, RR, falls, missed events, false alarms, and time-to-alarm as meaningful safety metrics
Adversarial Impact Backbone Papers showing that adversarial, perturbation, spoofing, physical-layer, black-box, or transfer attacks degrade WiFi CSI sensing or closely related wireless sensing outputs

View the backbone layer here:

Backbone Evidence Layer


Clinical-Safety Metrics Layer

This repository includes a dedicated clinical_safety_metrics/ layer for translating AI/ML WiFi sensing errors into safety-relevant outcomes.

The initial focus is fall detection under clean and adversarial conditions.

The central translation pathway is:

WiFi CSI fall detection output
→ ML prediction error
→ event-level clinical-safety metric
→ adversarial safety degradation

This layer addresses a key gap in current WiFi CSI and adversarial sensing research: many papers report accuracy, F1-score, or attack success rate, but do not explain whether model degradation causes more missed falls, more false alarms, longer detection latency, greater alarm burden, or increased long-lie risk.

The fall-detection framework prioritizes:

  • event-level recall / sensitivity,
  • missed fall rate,
  • false alarms per day or user-day,
  • precision / positive predictive value,
  • detection latency,
  • delayed detection rate,
  • long-lie risk proxy,
  • alarm burden, and
  • cost-sensitive safety scoring.

Secondary ML metrics such as accuracy, F1-score, specificity, balanced accuracy, MCC, AUC-PR, AUC-ROC, and confusion matrix remain useful for benchmarking, but they are treated as supporting metrics rather than the primary safety interpretation.

This folder currently contains:

File Purpose
README.md Overview of the clinical-safety metrics layer
fall_detection_framework.md Thesis-ready fall-detection safety-translation framework
metric_table.csv Structured table of primary, secondary, and adversarial-degradation metrics
evidence_matrix_fall_detection.csv Literature evidence matrix supporting the framework

This framework is clinically motivated and safety-oriented, but it is not a formal clinical standard, clinical validation, medical-device evaluation protocol, or regulatory claim.


Research Motivation

WiFi sensing has shown promise for detecting and interpreting human activity, motion, respiration, presence, and environment-level changes without requiring wearable devices or cameras. AI/ML methods are increasingly used to classify patterns, estimate signals, and support higher-level sensing applications from WiFi channel information.

However, the field still faces important challenges:

  • Dataset availability and comparability
  • Code availability and reproducibility
  • Model robustness under noise, domain shift, and environmental change
  • Security risks at the wireless and physical layers
  • Privacy and spoofing concerns
  • Lack of standardized evaluation across tasks and environments
  • Limited connection between sensing failures and real-world application risks
  • Limited translation from technical degradation into clinical-safety metrics such as missed falls, HR/RR error, false alarms, and time-to-alarm

This hub is designed to organize that evidence in a structured, transparent, and collaboration-friendly way.


Current Focus Area: Healthcare-Relevant WiFi Sensing

The first category in this hub focuses on healthcare-relevant and care-aware WiFi sensing, including:

  • Fall detection
  • Activity recognition
  • Respiration monitoring
  • Heart-rate and respiratory-rate sensing
  • Vital-sign-related sensing
  • Aging-in-place research
  • Assisted-living environments
  • Non-invasive monitoring studies
  • Safety-aware sensing evaluation

This project does not claim clinical validation or medical-device readiness. The focus is research evidence mapping, reproducibility, security analysis, clinical-safety translation, and trustworthy AI/ML sensing methods.


Future Expansion Areas

While healthcare-relevant sensing is the first category, this hub is designed to grow into other AI/ML WiFi sensing areas, such as:

Future Category Example Topics
Smart Environments Occupancy, room-level activity, behavior patterns
Human Activity Recognition Gesture, movement, posture, daily activities
Security and Privacy Spoofing, adversarial attacks, physical-layer risks, privacy leakage
Robotics and Interaction Device-free interaction, localization, environment awareness
Industrial and IoT Sensing Presence detection, monitoring, automation, anomaly detection
Reproducible WiFi Sensing Public datasets, baselines, code releases, benchmark design

Project Direction

Direction Purpose
Scientific Evidence Mapping Curate and classify papers, datasets, methods, and open gaps in AI/ML WiFi sensing
Healthcare-Relevant Sensing Track evidence related to contactless monitoring, fall detection, respiration, HR/RR sensing, and care-aware environments
Trustworthy AI/ML Sensing Track robustness, reproducibility, and model-evaluation limitations across existing work
Clinical-Safety Translation Map technical sensing failures to safety-facing outcomes such as missed falls, false alarms, HR/RR threshold errors, and time-to-alarm
Wireless Security Analysis Identify adversarial, physical-layer, spoofing, perturbation, and privacy-related risks
Open Collaboration Create a structured place for researchers, labs, and practitioners to suggest related work, datasets, and code

Visual Research Scope

flowchart LR
    A["AI/ML WiFi Sensing Hub"] --> B["Scientific Literature"]
    A --> C["Datasets"]
    A --> D["Code & Reproducibility"]
    A --> E["Security & Robustness"]
    A --> F["Clinical-Safety Translation"]
    A --> G["Application Categories"]
    A --> H["Open Collaboration"]

    B --> B1["WiFi CSI Sensing"]
    B --> B2["AI/ML Methods"]
    B --> B3["Signal Processing"]
    B --> B4["Benchmark Papers"]

    C --> C1["Public Dataset Links"]
    C --> C2["Access Status"]
    C --> C3["License Notes"]
    C --> C4["Experiment Suitability"]

    D --> D1["GitHub Repositories"]
    D --> D2["Reproducibility Notes"]
    D --> D3["Missing / Broken Code"]
    D --> D4["Baseline Methods"]

    E --> E1["Adversarial ML"]
    E --> E2["Physical-Layer Attacks"]
    E --> E3["CSI Perturbation"]
    E --> E4["Robustness Evaluation"]
    E --> E5["Privacy and Spoofing Risks"]

    F --> F1["Missed Falls"]
    F --> F2["HR/RR Error"]
    F --> F3["False Alarms"]
    F --> F4["Time-to-Alarm"]

    G --> G1["Healthcare-Relevant Sensing"]
    G --> G2["Smart Environments"]
    G --> G3["Human Activity Recognition"]
    G --> G4["Future Categories"]

    H --> H1["Authors"]
    H --> H2["Labs"]
    H --> H3["Dataset Owners"]
    H --> H4["AI/ML WiFi Sensing Researchers"]
Loading

Evidence Workflow

flowchart TD
    A["New Paper, Dataset, Code Repository, Clinical Reference, or Safety Metric"] --> B["Evidence Intake"]
    B --> C{"Relevant to AI/ML WiFi Sensing Hub?"}

    C -- "No" --> Z["Archived / Out of Scope"]
    C -- "Yes" --> D["Scientific Summary"]

    D --> E["Dataset & Code Availability Check"]
    D --> F["Clinical-Safety Metric Mapping"]
    D --> G["Security and Robustness Mapping"]

    E --> H["Reproducibility Evidence"]
    F --> I["Clinical-Safety Translation Evidence"]
    G --> J["Adversarial / Security Evidence"]

    I --> I1["Event-Level Safety Metrics"]
    I --> I2["Clean-to-Attacked Safety Degradation"]
    I --> I3["Fall / HR / RR Metric Translation"]

    H --> K["Added to Evidence Hub"]
    I --> K
    J --> K

    K --> L{"Potential for Collaboration?"}
    L -- "Yes" --> M["Research / Lab / Community Outreach Candidate"]
    L -- "No" --> N["Internal Evidence Reference"]
Loading

Evidence Categories

Category Purpose
Backbone Evidence Highest-value references organized into conceptual, reproducibility, clinical-safety, and adversarial-impact layers
Core Research Evidence Papers that strongly shape the technical direction of AI/ML WiFi sensing
Healthcare-Relevant WiFi Sensing Work on contactless sensing for activity, fall detection, respiration, HR/RR, vital signs, and care-aware environments
Clinical-Safety Translation Evidence References that justify missed falls, false alarms, HR/RR threshold errors, response delay, and time-to-alarm as meaningful safety metrics
Event-Level Safety Metrics Metrics that translate model predictions into event-level outcomes such as missed falls, false alarms per day, detection latency, and long-lie risk
Adversarial Safety Degradation Clean-to-attacked comparison of clinical-safety metrics under adversarial or degraded WiFi sensing conditions
Security and Robustness Evidence Papers on adversarial attacks, defenses, spoofing, perturbation, privacy, and physical-layer risks
Dataset Evidence Public datasets, access status, licensing, modalities, endpoints, and suitability for experiments
Code and Reproducibility Evidence Available repositories, reproducibility notes, baseline implementations, and missing-code gaps
Future Application Categories Research connected to smart environments, human activity recognition, IoT sensing, robotics, and other WiFi sensing areas
Collaboration Leads Authors, labs, datasets, and groups whose work may connect to this hub

Evidence Map Files

The structured evidence tables are stored in evidence_map/:

File Purpose
papers_index.csv Structured paper list
datasets_index.csv Dataset access, modality, license, and experiment-suitability tracking
code_releases_index.csv Code availability, setup clarity, and reproducibility notes
security_gap_map.csv Security, privacy, adversarial ML, and robustness evidence
collaboration_candidates.csv Authors, labs, dataset owners, and communities connected to the hub
backbone_conceptual_papers.csv Structured table for conceptual backbone references
backbone_reproducibility_resources.csv Structured table for reproducibility backbone resources
backbone_clinical_safety_translation.csv Structured table for clinical-safety translation references
backbone_adversarial_impact.csv Structured table for adversarial-impact backbone papers

Clinical-Safety Metric Files

The dedicated clinical-safety metric framework files are stored in clinical_safety_metrics/:

File Purpose
README.md Overview of the clinical-safety metrics layer
fall_detection_framework.md Thesis-ready framework for evaluating fall detection under clean and adversarial conditions
metric_table.csv Structured metric definitions, formulas, required data, clean use, attacked use, and priority
evidence_matrix_fall_detection.csv Literature matrix showing how fall-detection, WiFi CSI, clinical-safety, and adversarial papers report or omit safety metrics

Review Templates

This repository includes reusable review templates:

Template Purpose
papers/paper_review_template.md Review and summarize papers
datasets/dataset_review_template.md Evaluate datasets and access conditions
code_releases/code_release_review_template.md Review code repositories and reproducibility
security_robustness/security_mapping_template.md Map security, privacy, and robustness evidence
collaboration/collaboration_candidate_template.md Evaluate authors, labs, groups, and collaboration candidates

Intended Audience

This project is designed for researchers and practitioners working across:

  • AI/ML for WiFi sensing
  • WiFi CSI and wireless sensing
  • Healthcare-relevant sensing
  • Cybersecurity and adversarial machine learning
  • Reproducible research
  • Aging technology and assisted-living research
  • Contactless monitoring systems
  • Smart environments and IoT sensing
  • Clinical-safety translation for AI-based sensing systems
  • Trustworthy AI for sensing applications

How to Suggest Resources

Researchers, students, and practitioners can suggest resources through GitHub Issues using the structured forms:

  • Suggest a paper
  • Suggest a dataset
  • Suggest a code repository
  • Suggest a collaboration idea or research group

Please see CONTRIBUTING.md for scope, review criteria, and contribution guidance.


Collaboration Invitation

Suggestions are welcome for:

  • Relevant papers
  • Public datasets
  • Code repositories
  • Reproducibility notes
  • Missing but important related work
  • Security or robustness gaps
  • Clinical-safety translation references
  • Event-level safety metric references
  • Adversarial safety-degradation evaluation methods
  • Research groups or labs working in related areas
  • New application categories for AI/ML WiFi sensing

GitHub issue templates provide a structured way to suggest papers, datasets, code resources, and collaboration ideas.


Roadmap

See ROADMAP.md for the current development roadmap.

Near-term priorities include:

  1. Continue verifying the backbone evidence layer.
  2. Review and maintain structured backbone CSV files in evidence_map/.
  3. Expand the clinical-safety metrics layer beyond fall detection to HR/RR sensing.
  4. Add reproducible examples for clean vs attacked safety-metric calculation.
  5. Identify public datasets and code repositories suitable for fall, HR, and RR sensing experiments.
  6. Mark dataset/code availability, licensing, and reproducibility status.
  7. Map security, robustness, and privacy relevance for each high-value source.
  8. Identify research groups or labs connected to high-value papers and datasets.

Citation

If you use or reference this hub, please cite it using the metadata in CITATION.cff.

GitHub also provides a Cite this repository option in the repository sidebar.


Research Disclaimer

This project is a research evidence hub. It does not claim clinical validation, medical-device readiness, real patient deployment, regulatory approval, or diagnostic capability.

The clinical-safety metrics layer is a clinically motivated research evaluation framework. It does not define a formal clinical standard, clinical validation protocol, medical-device evaluation protocol, or regulatory pathway.

The focus is scientific evidence mapping, reproducibility, security analysis, clinical-safety translation, and trustworthy AI/ML sensing methods for WiFi-based sensing applications.


Current Status

This hub is under active development as a structured research resource for AI/ML WiFi sensing, starting with healthcare-relevant and care-aware sensing applications.

The repository now includes a dedicated clinical-safety metrics layer for fall detection, connecting WiFi CSI sensing outputs, ML prediction errors, event-level safety metrics, and adversarial safety degradation.

About

Open research hub mapping AI/ML WiFi sensing papers, datasets, code, reproducibility, and security gaps, starting with healthcare-relevant sensing.

Topics

Resources

Contributing

Stars

2 stars

Watchers

0 watching

Forks

Contributors