gateway/modules/features/neutralization/serviceNeutralization/subProcessPdfInPlace.py

110 lines
3.6 KiB
Python

# Copyright (c) 2025 Patrick Motsch
# All rights reserved.
"""
PDF in-place neutralization using PyMuPDF.
Removes original text completely and inserts full UUID placeholders.
PyMuPDF uses insert_textbox which wraps long placeholders to preserve layout.
"""
import io
import logging
from typing import Dict, Optional
logger = logging.getLogger(__name__)
def neutralize_pdf_in_place(
pdf_bytes: bytes,
mapping: Dict[str, str],
) -> Optional[bytes]:
"""
Remove sensitive text and replace with UUID placeholders in-place.
Content is fully removed (not just covered) so it cannot be copied.
Args:
pdf_bytes: Original PDF file content
mapping: Dict of original_text -> placeholder (e.g. [address.uuid])
Returns:
Modified PDF bytes, or None on failure
"""
if not mapping:
return pdf_bytes
try:
import fitz # PyMuPDF
except ImportError:
logger.warning("PyMuPDF (fitz) not available for PDF in-place neutralization")
return None
try:
doc = fitz.open(stream=pdf_bytes, filetype="pdf")
except Exception as e:
logger.error(f"Failed to open PDF: {e}")
return None
sorted_items = sorted(mapping.items(), key=lambda x: -len(x[0]))
fill_color = (1, 1, 1)
text_color = (0, 0, 0)
fontname = "helv"
fontsize = 8
try:
for page_num in range(len(doc)):
page = doc[page_num]
for original_text, placeholder in sorted_items:
if not original_text or not placeholder:
continue
search_text = original_text
insert_text = placeholder
if placeholder.startswith("[policy."):
# Try label+number to get wider rect; insert UUID only (label+UUID would overflow)
for prefix in ("Police Nr. ", "Police Nr.: ", "Polizzenr. ", "Policy no. ", "Policy No. "):
candidate = prefix + original_text
try:
hits = page.search_for(candidate, quads=False)
if hits:
search_text = candidate
insert_text = placeholder # UUID only so it fits in rect
break
except Exception:
continue
try:
instances = page.search_for(search_text, quads=False)
except Exception:
instances = []
for rect in instances:
try:
fs = 5 if placeholder.startswith(("[policy.", "[address.")) else fontsize
page.add_redact_annot(
rect,
text=insert_text,
fill=fill_color,
text_color=text_color,
fontname=fontname,
fontsize=fs,
)
except Exception as e:
logger.warning(f"Redact failed for {original_text[:40]!r}: {e}")
try:
page.apply_redactions()
except Exception as e:
logger.debug(f"apply_redactions page {page_num + 1}: {e}")
buf = io.BytesIO()
doc.save(buf, garbage=4, deflate=True)
doc.close()
return buf.getvalue()
except Exception as e:
logger.error(f"PDF in-place neutralization failed: {e}", exc_info=True)
try:
doc.close()
except Exception:
pass
return None