Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Fine-Grained Traceability for Transparent ML Pipelines

Liping Chen, Mujie Liu, Haytham Fayek

TL;DR

FG-Trac delivers verifiable, sample-level traceability for ML pipelines by three integrated components: a Traceable Module that records fine-grained lifecycle events, a Blockchain Module that anchors logs with Merkle-tree commitments on a public ledger, and a Query Module that enables independent verification. It combines a checkpoint-based TracInCP influence estimator with tamper-evident logging to reconstruct how individual samples were used and influenced model updates without modifying model architectures. Experiments on a CNN with CIFAR-10 and a multimodal GNN in clinical settings show negligible impact on predictive performance while providing auditable, end-to-end traces and a user-facing auditing interface. This work advances accountability in high-stakes ML deployments by enabling verifiable data usage histories and reproducible provenance in both standard and high-risk domains.

Abstract

Modern machine learning systems are increasingly realised as multistage pipelines, yet existing transparency mechanisms typically operate at a model level: they describe what a system is and why it behaves as it does, but not how individual data samples are operationally recorded, tracked, and verified as they traverse the pipeline. This absence of verifiable, sample-level traceability leaves practitioners and users unable to determine whether a specific sample was used, when it was processed, or whether the corresponding records remain intact over time. We introduce FG-Trac, a model-agnostic framework that establishes verifiable, fine-grained sample-level traceability throughout machine learning pipelines. FG-Trac defines an explicit mechanism for capturing and verifying sample lifecycle events across preprocessing and training, computes contribution scores explicitly grounded in training checkpoints, and anchors these traces to tamper-evident cryptographic commitments. The framework integrates without modifying model architectures or training objectives, reconstructing complete and auditable data-usage histories with practical computational overhead. Experiments on a canonical convolutional neural network and a multimodal graph learning pipeline demonstrate that FG-Trac preserves predictive performance while enabling machine learning systems to furnish verifiable evidence of how individual samples were used and propagated during model execution.

Fine-Grained Traceability for Transparent ML Pipelines

TL;DR

FG-Trac delivers verifiable, sample-level traceability for ML pipelines by three integrated components: a Traceable Module that records fine-grained lifecycle events, a Blockchain Module that anchors logs with Merkle-tree commitments on a public ledger, and a Query Module that enables independent verification. It combines a checkpoint-based TracInCP influence estimator with tamper-evident logging to reconstruct how individual samples were used and influenced model updates without modifying model architectures. Experiments on a CNN with CIFAR-10 and a multimodal GNN in clinical settings show negligible impact on predictive performance while providing auditable, end-to-end traces and a user-facing auditing interface. This work advances accountability in high-stakes ML deployments by enabling verifiable data usage histories and reproducible provenance in both standard and high-risk domains.

Abstract

Modern machine learning systems are increasingly realised as multistage pipelines, yet existing transparency mechanisms typically operate at a model level: they describe what a system is and why it behaves as it does, but not how individual data samples are operationally recorded, tracked, and verified as they traverse the pipeline. This absence of verifiable, sample-level traceability leaves practitioners and users unable to determine whether a specific sample was used, when it was processed, or whether the corresponding records remain intact over time. We introduce FG-Trac, a model-agnostic framework that establishes verifiable, fine-grained sample-level traceability throughout machine learning pipelines. FG-Trac defines an explicit mechanism for capturing and verifying sample lifecycle events across preprocessing and training, computes contribution scores explicitly grounded in training checkpoints, and anchors these traces to tamper-evident cryptographic commitments. The framework integrates without modifying model architectures or training objectives, reconstructing complete and auditable data-usage histories with practical computational overhead. Experiments on a canonical convolutional neural network and a multimodal graph learning pipeline demonstrate that FG-Trac preserves predictive performance while enabling machine learning systems to furnish verifiable evidence of how individual samples were used and propagated during model execution.
Paper Structure (25 sections, 4 equations, 4 figures, 5 tables, 1 algorithm)

This paper contains 25 sections, 4 equations, 4 figures, 5 tables, 1 algorithm.

Table of Contents

  1. Introduction
  2. Related Work
  3. Transparency Artefacts for Machine Learning Systems
  4. Traceability in Machine Learning Systems
  5. Influence and Contribution Estimation
  6. The FG-Trac Framework
  7. Traceable Module
  8. Blockchain Module
  9. Query Module
  10. Computational Complexity
  11. Experiments
  12. Experimental Setup
  13. Datasets.
  14. Backbone models and baselines.
  15. Implementation details.
  16. ...and 10 more sections

Figures (4)

  • Figure 1: Overview of the FG-Trac framework.
  • Figure 2: Classification accuracy comparison between baseline CNN and FG-Trac-integrated CNN on CIFAR-10 after 100 epochs.
  • Figure 3: Visualisation of sample-level contribution reconstructed from a queried training sample.
  • Figure 4: User query interface automatically generated from the FG-Trac auditing module for sample-level traceability