Update README.md

README.md

@@ -1,42 +1,49 @@
 ---
+pretty_name: Streaming PHI Deidentification Benchmark
 license: mit
 language:
 - en
 tags:
-  - healthcare
-  - privacy
-  - de-identification
-  - phi
-  - hipaa
-  - reinforcement-learning
-  - streaming
-  - multimodal
-  - audit-log
-  - synthetic
-pretty_name: Synthetic Multimodal Streaming De-Identification Audit Log
+- pii
+- phi
+- de-identification
+- deidentification
+- healthcare
+- healthcare-nlp
+- clinical-text
+- medical-nlp
+- privacy
+- hipaa
+- reinforcement-learning
+- streaming
+- multimodal
+- audit-log
+- synthetic
+- benchmark
+- evaluation
+- synthetic-data
 size_categories:
 - 1K<n<10K
 ---
 
-# Synthetic Multimodal Streaming De-Identification Audit Log
+# Streaming PHI De-Identification Benchmark
 
-A synthetic audit log dataset generated by a stateful, exposure-aware de-identification controller for multimodal streaming data. Every record in this dataset is fully synthetic. No real patient data, protected health information, or identifiable individuals are present anywhere in this dataset.
+A benchmark dataset for evaluating adaptive de-identification policies on multimodal streaming healthcare data. Every record is fully synthetic. No real patient data, protected health information, or identifiable individuals are present anywhere in this dataset.
 
-## What this dataset is for
+This dataset is different from existing PHI benchmarks. Most evaluate masking quality on static, single-modality documents. This dataset captures how re-identification risk accumulates across modalities and time — the same subject appearing across clinical notes, ASR transcripts, imaging proxies, waveform data, and audio metadata over a longitudinal stream.
 
-Most de-identification research evaluates masking quality on static documents. This dataset is different. It captures per-event masking decisions across a simulated longitudinal stream — the same subject appearing across multiple modalities and time steps — which allows evaluation of how re-identification risk accumulates over time rather than within a single record.
-
-It is useful for:
+It supports:
 
 - Benchmarking adaptive vs static de-identification policies
 - Studying cross-modal PHI linkage behavior
 - Evaluating privacy-utility tradeoffs in streaming systems
 - Reproducing results from the associated research implementation
 
-## Quick start
+## Quick Start
 
-# Install the package first
-pip install phi-exposure-guard
+```python
+# Install dependencies
+# pip install phi-exposure-guard datasets pandas
 
 import pandas as pd
 import json
@@ -49,15 +56,39 @@ print(df)
 with open("audit_log.jsonl") as f:
     events = [json.loads(line) for line in f]
 
-# Example: filter adaptive policy events
+# Filter adaptive policy events
 adaptive = [e for e in events if e["policy_run"] == "adaptive"]
+
+# Filter by modality
+text_events = [e for e in events if e["modality"] == "text"]
+```
+
+Or load via Hugging Face datasets library:
+
+```python
+from datasets import load_dataset
+
+ds = load_dataset("vkatg/streaming-phi-deidentification-benchmark")
+```
+
+## Motivation
+
+Automatic de-identification of healthcare data is critical for safe secondary use in AI research and clinical analytics. Existing PHI datasets have significant limitations for streaming and multimodal evaluation:
+
+- They evaluate each document in isolation with no memory of prior events
+- They are restricted to single modalities, usually clinical text only
+- Most require data use agreements and are not openly available
+- None model cumulative re-identification risk across events and modalities
+
+This dataset was created to fill that gap — providing a reproducible, open benchmark for evaluating exposure-aware and adaptive de-identification strategies in realistic streaming pipelines.
+
 ## Privacy-Utility Tradeoff
 
 ![Privacy-Utility Tradeoff](privacy_utility_curve.png)
 
 The adaptive policy achieves full utility while keeping leakage close to the redact floor. It only escalates masking strength when cumulative exposure actually justifies it, not on every record by default.
 
-## Policy benchmark results
+## Policy Benchmark Results
 
 | Policy | Leak Total | Utility Proxy | Mean Latency (ms) | P90 Latency (ms) |
 | --- | --- | --- | --- | --- |
@@ -67,11 +98,21 @@ The adaptive policy achieves full utility while keeping leakage close to the red
 | redact | 0.51 | 0.51 | 0.157 | 0.192 |
 | adaptive | 0.56 | 1.0 | 1.16 | 1.237 |
 
-## Dataset structure
+## Comparison with Existing PHI Datasets
+
+| Dataset | Modality | Access | Real Data | Streaming | Size |
+| --- | --- | --- | --- | --- | --- |
+| i2b2 2014 | Text only | DUA required | Yes | No | 1,304 records |
+| PhysioNet deid | Text only | DUA required | Yes | No | 2,434 notes |
+| This dataset | Text, ASR, Image, Waveform, Audio | Open | No (synthetic) | Yes | 1K-10K events |
+
+The key distinction is multimodal streaming evaluation. Existing datasets evaluate masking quality on static text records in isolation. This dataset captures how re-identification risk accumulates across modalities and time steps.
+
+## Dataset Structure
 
 The primary file is `audit_log.jsonl`. Each line is one masking decision for one event.
 
-### Key fields
+### Key Fields
 
 | Field | Description |
 | --- | --- |
@@ -88,6 +129,24 @@ The primary file is `audit_log.jsonl`. Each line is one masking decision for one
 | `policy_version` | Policy version token |
 | `decision_blob` | Full structured decision record including risk components, DCPG graph state, CMO execution log, and cross-modal matches |
 
+### Sample Record
+
+```json
+{
+  "event_id": "evt_0",
+  "patient_key": "patient_1",
+  "modality": "text",
+  "policy_run": "adaptive",
+  "chosen_policy": "pseudo",
+  "reason": "risk_threshold_exceeded",
+  "risk": 0.43,
+  "localized_remask_trigger": false,
+  "latency_ms": 1.12,
+  "leaks_after": 0.0,
+  "policy_version": "v1"
+}
+```
+
 ### Modalities
 
 Each event is evaluated across five modalities representing a realistic multimodal healthcare stream:
@@ -100,15 +159,30 @@ Each event is evaluated across five modalities representing a realistic multimod
 
 ### Policies
 
-Five masking policies are included for comparison:
-
 - `raw` — no masking applied
 - `weak` — minimal token suppression
 - `pseudo` — pseudonymization
 - `redact` — full redaction
 - `adaptive` — risk-governed dynamic policy selection
 
-## Additional files
+## System Architecture
+
+The dataset was generated by a modular pipeline with the following components:
+
+| Module | Description |
+| --- | --- |
+| `context_state.py` | Per-subject exposure state, persisted via SQLite |
+| `controller.py` | Risk scoring and adaptive policy selection |
+| `dcpg.py` | Dynamic Context Persistence Graph — tracks cross-modal identity linkage |
+| `dcpg_crdt.py` | CRDT-based federated graph merging for distributed deployments |
+| `cmo_registry.py` | Composable Masking Operator registry and DAG execution |
+| `flow_controller.py` | DAG-based policy flow controller with audit provenance |
+| `rl_agent.py` | PPO reinforcement learning agent for adaptive policy control |
+| `audit_signing.py` | Cryptographic signing and FHIR export of audit records |
+| `phi_detector.py` | PHI detection and leakage measurement |
+| `masking_ops.py` | Token-level masking operations per policy |
+
+## Additional Files
 
 | File | Description |
 | --- | --- |
@@ -117,7 +191,11 @@ Five masking policies are included for comparison:
 | `audit_log_signed_adaptive.jsonl` | Cryptographically signed adaptive policy audit trail |
 | `EXPERIMENT_REPORT.md` | Full experiment report with leakage breakdown by modality |
 
-## Source code and reproduction
+## Limitations
+
+This dataset is fully synthetic and designed to model structural properties of longitudinal healthcare streams. It does not contain real clinical language, real diagnoses, or real patient records. Models trained or evaluated on this dataset should be validated on real clinical data before deployment. The dataset covers English language proxies only.
+
+## Source Code and Reproduction
 
 All data in this dataset was generated by the open-source research implementation:
 
@@ -125,7 +203,9 @@ GitHub: [azithteja91/phi-exposure-guard](https://github.com/azithteja91/phi-expo
 
 Hugging Face Space: [vkatg/amphi-rl-dpgraph](https://huggingface.co/spaces/vkatg/amphi-rl-dpgraph)
 
-To regenerate:
+Colab demo: [![Open In Colab](https://colab.research.google.com/assets/colab-badge.svg)](https://colab.research.google.com/github/azithteja91/phi-exposure-guard/blob/main/notebooks/demo_colab.ipynb)
+
+To regenerate locally:
 
 ```bash
 git clone https://github.com/azithteja91/phi-exposure-guard.git
@@ -142,6 +222,6 @@ If you use this dataset in academic or technical work, please cite via the `CITA
 
 MIT
 
-## Note on intended use
+## Note on Intended Use
 
 This dataset is intended for privacy research, streaming system design, and reproducible evaluation of de-identification strategies. It is not derived from real clinical data and should not be used as a substitute for validated compliance infrastructure.