iText 9: Java PDF Parsing Powerhouse — Encryption, Signatures, and Extraction

Introduction to iText

iText is a low-level PDF SDK for Java and .NET. The community edition (iText Core/Community) is licensed under AGPL; commercial use in closed-source products requires a paid license. Core capabilities include generation (building pages with paragraphs, tables, images, headers/footers), parsing (extracting text, images, layers), manipulation (split, merge, stamp, fill forms), and compliance (PDF/A archiving, PDF/UA accessibility, digital signatures). Compared to Apache PDFBox, iText leans more toward production-grade typesetting and standards compliance, and unlike online converters it can be embedded directly in services — making it a standard choice for invoice recognition, contract archiving, and e-signature workflows in the Java ecosystem.

What's New in iText 9

iText 9.0 was released late 2024; the current version is 9.7.1. The major version breaks some 8.x compatibility, but migration is less painful than the 5-to-7 jump. Key changes developers will notice:

1. Encryption Catches Up to PDF 2.0 Standards

Version 9.0 adds support for two new ISO technical specifications:

ISO/TS 32003 : AES-GCM encryption in PDF 2.0 — faster and more secure than the legacy AES-CBC.

ISO/TS 32004 : PDF MAC — adds an integrity layer so any tampering with an encrypted document is immediately detectable.

In short: previously you could only lock a document; now you can also prove it hasn't been altered.

2. Digital Signatures: From Signing to Verification

iText 8 deepened signing capabilities; 9.0 completes the loop with a verification module. You can now verify a single signature, handle signatures inside encrypted documents, and traverse revision history and certificate chains — all without switching tools.

3. PDF/A and PDF/UA Creation Made Simpler

Accessibility and archiving have long been iText strengths. 9.0 streamlines creation and conformance checking for both standards. When signing a PDF/UA document, missing fonts or missing alternative text in the signature appearance now throw a conformance exception immediately, instead of leaving you to guess among a pile of property errors. A small but practical addition: an API to list which optional content groups (OCGs / layers) are actually used on a given page — handy for engineering drawings and layered design PDFs.

4. Minor Releases Keep Adding Value

9.5.0 : Brotli compression stream (future PDF spec), post-quantum signature algorithm experiments

9.7.0 : Native WebP image support, dynamic page margins, footnote layout, stronger decompression-bomb protection

9.7.1 : Fixed Jackson dependency security issue; upgrade from 9.7.0 recommended

Adding Dependencies in Five Minutes

The simplest Maven setup pulls the itext-core BOM plus the official Bouncy Castle adapter (required for crypto and signing):

<properties>
  <itext.version>9.7.1</itext.version>
</properties>
<dependencies>
  <dependency>
    <groupId>com.itextpdf</groupId>
    <artifactId>itext-core</artifactId>
    <version>${itext.version}</version>
    <type>pom</type>
  </dependency>
  <dependency>
    <groupId>com.itextpdf</groupId>
    <artifactId>bouncy-castle-adapter</artifactId>
    <version>${itext.version}</version>
  </dependency>
</dependencies>

For lean projects that only need text extraction (no signing), you can depend on just kernel and layout:

<dependency>
  <groupId>com.itextpdf</groupId>
  <artifactId>kernel</artifactId>
  <version>9.7.1</version>
</dependency>
<dependency>
  <groupId>com.itextpdf</groupId>
  <artifactId>layout</artifactId>
  <version>9.7.1</version>
</dependency>

Chinese PDFs require explicit font files (e.g., Source Han Sans, SimSun); iText does not synthesize CJK glyphs.

Three Most Common Parsing Tasks

Daily parsing work concentrates on three operations.

1. Extract All Visible Text from the Entire Document

Entry-level code and the go-to for checking whether a file contains selectable text:

import com.itextpdf.kernel.pdf.PdfDocument;
import com.itextpdf.kernel.pdf.PdfReader;
import com.itextpdf.kernel.pdf.canvas.parser.PdfTextExtractor;
import java.io.IOException;

public class ExtractAllText {
  public static String extract(String pdfPath) throws IOException {
    try (PdfDocument pdf = new PdfDocument(new PdfReader(pdfPath))) {
      StringBuilder all = new StringBuilder();
      int pages = pdf.getNumberOfPages();
      for (int i = 1; i <= pages; i++) {
        String pageText = PdfTextExtractor.getTextFromPage(pdf.getPage(i));
        all.append("----- Page ").append(i).append(" -----
");
        all.append(pageText).append('
');
      }
      return all.toString();
    }
  }
  public static void main(String[] args) throws IOException {
    String text = extract("invoice.pdf");
    System.out.println(text);
  }
}

PdfTextExtractor

defaults to a location-based strategy, which yields reading order closer to human perception than raw content-stream order. Scanned or image-only PDFs return empty strings — that's an OCR job, requiring pdfOCR or another recognition service.

2. Control Line Breaks and Reading Order with a Custom Strategy

When the default isn't enough, use LocationTextExtractionStrategy directly. In 9.x its delimiter is customizable, and the multi-page regex-based strategy ( RegexBasedLocationExtractionStrategy) has been fixed for cleaner results:

import com.itextpdf.kernel.pdf.PdfDocument;
import com.itextpdf.kernel.pdf.PdfReader;
import com.itextpdf.kernel.pdf.canvas.parser.PdfCanvasProcessor;
import com.itextpdf.kernel.pdf.canvas.parser.listener.LocationTextExtractionStrategy;
import java.io.IOException;

public class ExtractWithStrategy {
  public static String extractFirstPage(String pdfPath) throws IOException {
    try (PdfDocument pdf = new PdfDocument(new PdfReader(pdfPath))) {
      LocationTextExtractionStrategy strategy = new LocationTextExtractionStrategy();
      PdfCanvasProcessor processor = new PdfCanvasProcessor(strategy);
      processor.processPageContent(pdf.getFirstPage());
      return strategy.getResultantText();
    }
  }
}

For invoices, the typical flow: extract full-page text, then regex-match anchors like invoice number and total amount. Once the layout is stable, this is far cheaper than invoking a large model.

3. Export Embedded Images from Pages

Common in receipt recognition and ID archiving — extract images first, then feed to OCR:

import com.itextpdf.kernel.pdf.PdfDocument;
import com.itextpdf.kernel.pdf.PdfPage;
import com.itextpdf.kernel.pdf.PdfReader;
import com.itextpdf.kernel.pdf.xobject.PdfImageXObject;
import com.itextpdf.kernel.pdf.PdfName;
import com.itextpdf.kernel.pdf.PdfDictionary;
import com.itextpdf.kernel.pdf.PdfStream;
import java.io.IOException;
import java.nio.file.Files;
import java.nio.file.Path;
import java.nio.file.Paths;

public class ExtractImages {
  public static void dump(String pdfPath, String outputDir) throws IOException {
    Path dir = Paths.get(outputDir);
    Files.createDirectories(dir);
    try (PdfDocument pdf = new PdfDocument(new PdfReader(pdfPath))) {
      int seq = 1;
      for (int i = 1; i <= pdf.getNumberOfPages(); i++) {
        PdfPage page = pdf.getPage(i);
        PdfDictionary resources = page.getResources().getPdfObject();
        PdfDictionary xobjects = resources.getAsDictionary(PdfName.XObject);
        if (xobjects == null) continue;
        for (PdfName name : xobjects.keySet()) {
          PdfStream stream = xobjects.getAsStream(name);
          if (stream == null || !PdfName.Image.equals(stream.getAsName(PdfName.Subtype))) continue;
          PdfImageXObject image = new PdfImageXObject(stream);
          String ext = image.identifyImageFileExtension();
          Path dest = dir.resolve("page-" + i + "-img-" + seq + "." + ext);
          Files.write(dest, image.getImageBytes());
          seq++;
        }
      }
    }
  }
}

Check the extension after export — JPEG and PNG are most common; since 9.7 WebP can also appear in PDFs.

Parsing Pipeline Overview

Chaining the snippets above yields a short pipeline. iText does not "understand" business meaning; it only restores page content into raw material you can further process. One habit to remember: open with PdfReader, wrap in PdfDocument, and always close — the examples use try-with-resources; keep that pattern in production to avoid handle leaks during batch jobs on large files. If a file may be corrupt, create the PdfReader in non-strict mode; 9.x gives clearer diagnostics when xref table reconstruction fails.

Generating PDFs

Parsing is often followed by generating a receipt PDF. Minimal example:

import com.itextpdf.kernel.pdf.PdfDocument;
import com.itextpdf.kernel.pdf.PdfWriter;
import com.itextpdf.layout.Document;
import com.itextpdf.layout.element.Paragraph;
import com.itextpdf.kernel.font.PdfFont;
import com.itextpdf.kernel.font.PdfFontFactory;
import com.itextpdf.io.font.constants.StandardFonts;
import java.io.IOException;

public class CreateSimplePdf {
  public static void create(String dest) throws IOException {
    PdfFont font = PdfFontFactory.createFont(StandardFonts.HELVETICA);
    try (PdfWriter writer = new PdfWriter(dest);
         PdfDocument pdf = new PdfDocument(writer);
         Document doc = new Document(pdf)) {
      doc.setFont(font);
      doc.add(new Paragraph("Hello, iText 9"));
      doc.add(new Paragraph("This PDF was created by Java."));
    }
  }
}

For Chinese, replace StandardFonts.HELVETICA with a disk font path, e.g.:

PdfFont font = PdfFontFactory.createFont(
  "C:/Windows/Fonts/simsun.ttc,0",
  "Identity-H",
  PdfFontFactory.EmbeddingStrategy.PREFER_EMBEDDED
);

Identity-H

denotes Unicode horizontal writing; PREFER_EMBEDDED ensures the font is embedded so the PDF opens correctly on machines without that font installed.

Common Pitfalls

Scanned pages are not text. PdfTextExtractor returns empty; verify in Acrobat whether text is selectable first.

Chinese requires explicit fonts. Omitting them causes missing glyphs, tofu blocks, or exceptions.

Encrypted documents need passwords. Use new PdfReader(path, new ReaderProperties().setPassword(...)); empty password, user password, and owner password are distinct.

AGPL is not "free for any use". Internal tools and open-source projects are fine; closed-source commercial distribution requires a commercial license.

Upgrading from 8 to 9: read the Breaking Changes first. Conformance, signature, and layer APIs have been reorganized; following the official migration guide is faster than fixing compile errors one by one.