itextpdf Redaction :Partly redacted text string is fully removed - itext

Question

Used itextpdf-5.5.9 and itext-xtra-5.5.9
I am trying to applying redaction on Partly text string but after applied redaction whole string is removed from document.Please find attached screenshot.
PdfReader reader = new PdfReader(src);
PdfCleanUpProcessor cleaner= null;
PdfStamper stamper = new PdfStamper(reader, new FileOutputStream(targetPdf));
stamper.setRotateContents(false);
List<PdfCleanUpLocation> cleanUpLocations = new ArrayList<PdfCleanUpLocation>();
Rectangle rectangle = new Rectangle(380, 640, 430, 665);
cleanUpLocations.add(new PdfCleanUpLocation(1, rectangle, BaseColor.BLACK));
cleaner = new PdfCleanUpProcessor(cleanUpLocations, stamper);
cleaner.cleanUp();
stamper.close();
reader.close();

Answer 1

The OP clarified in comments that it is indeed his expectation that redaction only removes text which is completely contained in the redaction area; text, though, whose bounding box is even partially outside that area, is expected to remain.
This expectation may be unwise as far as security of redaction is concerned because this way text a casual redactor does not see anymore due to the colored redaction area may still remain in the PDF content available to text extraction or even simple copy&paste.
If in spite of such reservations one still wants to tweak PdfCleanup to work like expected by the OP, one essentially merely has to change the PdfCleanUpRegionFilter used by the PdfCleanUpProcessor: The filter implementation used by default rejects a glyph (and so marks it for removal) if its bounding box intersects the redaction area. To fulfill the OP's expectations, this behavior has to be replaced by a check whether the bounding box is completely contained in the redaction area.
This sounds simple. Unfortunately it is not as simple as it sounds because the clean-up code is not designed for easy replacement of the region filter implementation, many relevant objects or methods are private or at best package protected.
Thus, to achieve the OP's desired behavior I simply copied all of the classes from the com.itextpdf.text.pdf package into an own package, therein added a new filter class derived from my copy of PdfCleanUpRegionFilter with the different text rejection algorithm mentioned above, and then changed the copy of PdfCleanUpProcessor to use this other filter class:
/**
* In contrast to the base class {#link PdfCleanUpRegionFilter}, this filter
* only rejects text <b>completely</b> inside the redaction zone. The original
* also rejects text located merely <b>partially</b> inside the redaction zone.
*/
public class StrictPdfCleanUpRegionFilter extends PdfCleanUpRegionFilter
{
public StrictPdfCleanUpRegionFilter(List<Rectangle> rectangles)
{
super(rectangles);
this.rectangles = rectangles;
}
/**
* Checks if the text is COMPLETELY inside render filter region.
*/
#Override
public boolean allowText(TextRenderInfo renderInfo) {
LineSegment ascent = renderInfo.getAscentLine();
LineSegment descent = renderInfo.getDescentLine();
Point2D[] glyphRect = new Point2D[] {
new Point2D.Float(ascent.getStartPoint().get(0), ascent.getStartPoint().get(1)),
new Point2D.Float(ascent.getEndPoint().get(0), ascent.getEndPoint().get(1)),
new Point2D.Float(descent.getEndPoint().get(0), descent.getEndPoint().get(1)),
new Point2D.Float(descent.getStartPoint().get(0), descent.getStartPoint().get(1)),
};
for (Rectangle rectangle : rectangles)
{
boolean glyphInRectangle = true;
for (Point2D point2d : glyphRect)
{
glyphInRectangle &= rectangle.getLeft() <= point2d.getX();
glyphInRectangle &= point2d.getX() <= rectangle.getRight();
glyphInRectangle &= rectangle.getBottom() <= point2d.getY();
glyphInRectangle &= point2d.getY() <= rectangle.getTop();
}
if (glyphInRectangle)
return false;
}
return true;
}
List<Rectangle> rectangles;
}
(StrictPdfCleanUpRegionFilter)
public class StrictPdfCleanUpProcessor {
...
private PdfCleanUpRegionFilter createFilter(List<PdfCleanUpLocation> cleanUpLocations) {
List<Rectangle> regions = new ArrayList<Rectangle>(cleanUpLocations.size());
for (PdfCleanUpLocation location : cleanUpLocations) {
regions.add(location.getRegion());
}
return new StrictPdfCleanUpRegionFilter(regions);
}
...
}
(StrictPdfCleanUpProcessor, my copy of PdfCleanUpProcessor)
All classes can be found here.
It can be used just like the original cleanup implementation, one merely has to remember using the copied classes, not the original ones:
try ( InputStream resource = getClass().getResourceAsStream("Document.pdf");
OutputStream result = new FileOutputStream(new File(OUTPUTDIR, "Document-redacted-strict.pdf")) )
{
PdfReader reader = new PdfReader(resource);
StrictPdfCleanUpProcessor cleaner= null;
PdfStamper stamper = new PdfStamper(reader, result);
stamper.setRotateContents(false);
List<mkl.testarea.itext5.pdfcleanup.PdfCleanUpLocation> cleanUpLocations = new ArrayList<>();
Rectangle rectangle = new Rectangle(380, 640, 430, 665);
cleanUpLocations.add(new mkl.testarea.itext5.pdfcleanup.PdfCleanUpLocation(1, rectangle, BaseColor.BLACK));
cleaner = new StrictPdfCleanUpProcessor(cleanUpLocations, stamper);
cleaner.cleanUp();
stamper.close();
reader.close();
}
(RedactText test method testRedactStrictForMayankPandey)
The example PDF provided by the OP
After redaction using the original classes
After redaction using the tweaked classes
A sanity check with the tweaked classes
To be sure the tweaked classes still removed any text at all, I enlarged the redaction area so that "heet", the last characters of "Document Submission Sheet", were completely contained in the redaction area:
try ( InputStream resource = getClass().getResourceAsStream("Document.pdf");
OutputStream result = new FileOutputStream(new File(OUTPUTDIR, "Document-redacted-strict-large.pdf")) )
{
PdfReader reader = new PdfReader(resource);
StrictPdfCleanUpProcessor cleaner= null;
PdfStamper stamper = new PdfStamper(reader, result);
stamper.setRotateContents(false);
List<mkl.testarea.itext5.pdfcleanup.PdfCleanUpLocation> cleanUpLocations = new ArrayList<>();
Rectangle rectangle = new Rectangle(380, 640, 430, 680);
cleanUpLocations.add(new mkl.testarea.itext5.pdfcleanup.PdfCleanUpLocation(1, rectangle, BaseColor.BLACK));
cleaner = new StrictPdfCleanUpProcessor(cleanUpLocations, stamper);
cleaner.cleanUp();
stamper.close();
reader.close();
}
(RedactText test method testRedactStrictForMayankPandeyLarge)
And indeed, copy&paste (and other text extraction methods, too) now only render
"Document Submission S".