Datasets:

utilvo-com
/

pdf-corruption-lab

#!/usr/bin/env node
/**
 * pdf-corruption-lab
 * Generates a small, fully synthetic corpus of damaged PDF files with a known
 * kind and severity of damage, plus a manifest.json describing each file.
 * Intended for testing and benchmarking PDF repair / recovery tools.
 *
 * Usage:  node generate.js [outputDir]
 */
const fs = require('fs');
const path = require('path');
const { PDFDocument, rgb, StandardFonts } = require('pdf-lib');

// 2x2 PNG used to make sure the base document contains an image stream.
const TEST_IMAGE_BASE64 =
  'iVBORw0KGgoAAAANSUhEUgAAAAIAAAACCAYAAABytg0kAAAAFElEQVQIW2NkwAJGDjACU2M4wAgAJu4BCHgB5hQAAAAASUVORK5CYII=';

const toLatin1 = (bytes) => new TextDecoder('latin1').decode(bytes);
const fromLatin1 = (str) => Uint8Array.from(str, (c) => c.charCodeAt(0) & 0xff);

async function buildBasePdf() {
  const doc = await PDFDocument.create();
  doc.setTitle('pdf-corruption-lab base document');
  const font = await doc.embedFont(StandardFonts.Helvetica);

  const p1 = doc.addPage([595, 842]);
  p1.drawText('pdf-corruption-lab: base document', { x: 40, y: 780, size: 20, font, color: rgb(0.1, 0.3, 0.8) });
  p1.drawText('Page 1 contains text and a vector rule.', { x: 40, y: 740, size: 11, font });
  p1.drawRectangle({ x: 40, y: 760, width: 515, height: 3, color: rgb(0, 0.6, 0.6) });

  const p2 = doc.addPage([595, 842]);
  p2.drawText('Page 2 contains an embedded raster image.', { x: 40, y: 780, size: 14, font });
  const png = await doc.embedPng(Buffer.from(TEST_IMAGE_BASE64, 'base64'));
  p2.drawImage(png, { x: 100, y: 450, width: 120, height: 120 });

  const p3 = doc.addPage([595, 842]);
  p3.drawText('Page 3 contains text only.', { x: 40, y: 780, size: 14, font });
  for (let i = 0; i < 20; i++) {
    p3.drawText(`Line ${i + 1}: the quick brown fox jumps over the lazy dog.`, { x: 40, y: 740 - i * 18, size: 10, font });
  }

  return doc.save({ useObjectStreams: false });
}

// Each scenario receives (bytes, text) and returns damaged bytes.
const scenarios = [
  {
    id: 'missing_eof',
    severity: 1,
    description: 'The trailing %%EOF marker is removed. Structure is otherwise intact.',
    apply: (_b, t) => fromLatin1(t.replace(/%%EOF\s*$/, '')),
  },
  {
    id: 'invalid_startxref',
    severity: 2,
    description: 'startxref points to a non-numeric value; the xref table itself is valid.',
    apply: (_b, t) => fromLatin1(t.replace(/startxref\s+\d+/, 'startxref\nINVALID')),
  },
  {
    id: 'scrambled_xref',
    severity: 3,
    description: 'All xref offsets are replaced with bogus values and startxref is invalid.',
    apply: (_b, t) => {
      const pos = t.lastIndexOf('xref\n') !== -1 ? t.lastIndexOf('\nxref') + 1 : t.lastIndexOf('xref');
      const head = t.substring(0, pos);
      const count = (t.match(/\n\d{10} \d{5} n/g) || []).length + 1;
      const rows = Array.from({ length: count - 1 }, (_, i) => `${String(99999 - i * 111).padStart(10, '0')} 00000 n \n`).join('');
      return fromLatin1(`${head}xref\n0 ${count}\n0000000000 65535 f \n${rows}trailer\n<< /Size ${count} >>\nstartxref\nINVALID\n%%EOF\n`);
    },
  },
  {
    id: 'bad_root_reference',
    severity: 3,
    description: 'The trailer /Root points to a non-existent object and startxref is -1.',
    apply: (_b, t) => {
      let s = t.replace(/\/Root \d+ \d+ R/, '/Root 9999 0 R');
      const pos = s.lastIndexOf('trailer');
      if (pos !== -1) s = `${s.substring(0, pos)}trailer\n<< /Size 9999 >>\nstartxref\n-1\n%%EOF\n`;
      return fromLatin1(s);
    },
  },
  {
    id: 'truncated_75',
    severity: 4,
    description: 'File cut at 75% of its length: xref, trailer and EOF are lost, last object is incomplete.',
    apply: (b) => b.slice(0, Math.floor(b.length * 0.75)),
  },
  {
    id: 'truncated_55',
    severity: 5,
    description: 'File cut at 55%: roughly half of the objects are lost; only partial content is recoverable.',
    apply: (b) => b.slice(0, Math.floor(b.length * 0.55)),
  },
];

async function main() {
  const outDir = path.resolve(process.argv[2] || path.join(__dirname, 'output'));
  fs.mkdirSync(outDir, { recursive: true });

  const base = await buildBasePdf();
  const baseText = toLatin1(base);
  fs.writeFileSync(path.join(outDir, 'base_valid.pdf'), base);

  const manifest = {
    generator: 'pdf-corruption-lab',
    base_file: 'base_valid.pdf',
    base_pages: 3,
    files: [],
  };

  for (const s of scenarios) {
    const damaged = s.apply(base, baseText);
    const name = `damaged_${s.id}.pdf`;
    fs.writeFileSync(path.join(outDir, name), damaged);
    manifest.files.push({ file: name, damage: s.id, severity: s.severity, description: s.description, bytes: damaged.length });
    console.log(`ok  ${name} (severity ${s.severity}, ${damaged.length} bytes)`);
  }

  fs.writeFileSync(path.join(outDir, 'manifest.json'), JSON.stringify(manifest, null, 2));
  console.log(`\nWrote ${manifest.files.length + 1} PDFs and manifest.json to ${outDir}`);
}

module.exports = { scenarios, toLatin1, fromLatin1 };

if (require.main === module) {
  main().catch((e) => {
    console.error(e);
    process.exit(1);
  });
}

	#!/usr/bin/env node
	/**
	* pdf-corruption-lab
	* Generates a small, fully synthetic corpus of damaged PDF files with a known
	* kind and severity of damage, plus a manifest.json describing each file.
	* Intended for testing and benchmarking PDF repair / recovery tools.
	*
	* Usage: node generate.js [outputDir]
	*/
	const fs = require('fs');
	const path = require('path');
	const { PDFDocument, rgb, StandardFonts } = require('pdf-lib');

	// 2x2 PNG used to make sure the base document contains an image stream.
	const TEST_IMAGE_BASE64 =
	'iVBORw0KGgoAAAANSUhEUgAAAAIAAAACCAYAAABytg0kAAAAFElEQVQIW2NkwAJGDjACU2M4wAgAJu4BCHgB5hQAAAAASUVORK5CYII=';

	const toLatin1 = (bytes) => new TextDecoder('latin1').decode(bytes);
	const fromLatin1 = (str) => Uint8Array.from(str, (c) => c.charCodeAt(0) & 0xff);

	async function buildBasePdf() {
	const doc = await PDFDocument.create();
	doc.setTitle('pdf-corruption-lab base document');
	const font = await doc.embedFont(StandardFonts.Helvetica);

	const p1 = doc.addPage([595, 842]);
	p1.drawText('pdf-corruption-lab: base document', { x: 40, y: 780, size: 20, font, color: rgb(0.1, 0.3, 0.8) });
	p1.drawText('Page 1 contains text and a vector rule.', { x: 40, y: 740, size: 11, font });
	p1.drawRectangle({ x: 40, y: 760, width: 515, height: 3, color: rgb(0, 0.6, 0.6) });

	const p2 = doc.addPage([595, 842]);
	p2.drawText('Page 2 contains an embedded raster image.', { x: 40, y: 780, size: 14, font });
	const png = await doc.embedPng(Buffer.from(TEST_IMAGE_BASE64, 'base64'));
	p2.drawImage(png, { x: 100, y: 450, width: 120, height: 120 });

	const p3 = doc.addPage([595, 842]);
	p3.drawText('Page 3 contains text only.', { x: 40, y: 780, size: 14, font });
	for (let i = 0; i < 20; i++) {
	p3.drawText(`Line ${i + 1}: the quick brown fox jumps over the lazy dog.`, { x: 40, y: 740 - i * 18, size: 10, font });
	}

	return doc.save({ useObjectStreams: false });
	}

	// Each scenario receives (bytes, text) and returns damaged bytes.
	const scenarios = [
	{
	id: 'missing_eof',
	severity: 1,
	description: 'The trailing %%EOF marker is removed. Structure is otherwise intact.',
	apply: (_b, t) => fromLatin1(t.replace(/%%EOF\s*$/, '')),
	},
	{
	id: 'invalid_startxref',
	severity: 2,
	description: 'startxref points to a non-numeric value; the xref table itself is valid.',
	apply: (_b, t) => fromLatin1(t.replace(/startxref\s+\d+/, 'startxref\nINVALID')),
	},
	{
	id: 'scrambled_xref',
	severity: 3,
	description: 'All xref offsets are replaced with bogus values and startxref is invalid.',
	apply: (_b, t) => {
	const pos = t.lastIndexOf('xref\n') !== -1 ? t.lastIndexOf('\nxref') + 1 : t.lastIndexOf('xref');
	const head = t.substring(0, pos);
	const count = (t.match(/\n\d{10} \d{5} n/g) \|\| []).length + 1;
	const rows = Array.from({ length: count - 1 }, (_, i) => `${String(99999 - i * 111).padStart(10, '0')} 00000 n \n`).join('');
	return fromLatin1(`${head}xref\n0 ${count}\n0000000000 65535 f \n${rows}trailer\n<< /Size ${count} >>\nstartxref\nINVALID\n%%EOF\n`);
	},
	},
	{
	id: 'bad_root_reference',
	severity: 3,
	description: 'The trailer /Root points to a non-existent object and startxref is -1.',
	apply: (_b, t) => {
	let s = t.replace(/\/Root \d+ \d+ R/, '/Root 9999 0 R');
	const pos = s.lastIndexOf('trailer');
	if (pos !== -1) s = `${s.substring(0, pos)}trailer\n<< /Size 9999 >>\nstartxref\n-1\n%%EOF\n`;
	return fromLatin1(s);
	},
	},
	{
	id: 'truncated_75',
	severity: 4,
	description: 'File cut at 75% of its length: xref, trailer and EOF are lost, last object is incomplete.',
	apply: (b) => b.slice(0, Math.floor(b.length * 0.75)),
	},
	{
	id: 'truncated_55',
	severity: 5,
	description: 'File cut at 55%: roughly half of the objects are lost; only partial content is recoverable.',
	apply: (b) => b.slice(0, Math.floor(b.length * 0.55)),
	},
	];

	async function main() {
	const outDir = path.resolve(process.argv[2] \|\| path.join(__dirname, 'output'));
	fs.mkdirSync(outDir, { recursive: true });

	const base = await buildBasePdf();
	const baseText = toLatin1(base);
	fs.writeFileSync(path.join(outDir, 'base_valid.pdf'), base);

	const manifest = {
	generator: 'pdf-corruption-lab',
	base_file: 'base_valid.pdf',
	base_pages: 3,
	files: [],
	};

	for (const s of scenarios) {
	const damaged = s.apply(base, baseText);
	const name = `damaged_${s.id}.pdf`;
	fs.writeFileSync(path.join(outDir, name), damaged);
	manifest.files.push({ file: name, damage: s.id, severity: s.severity, description: s.description, bytes: damaged.length });
	console.log(`ok ${name} (severity ${s.severity}, ${damaged.length} bytes)`);
	}

	fs.writeFileSync(path.join(outDir, 'manifest.json'), JSON.stringify(manifest, null, 2));
	console.log(`\nWrote ${manifest.files.length + 1} PDFs and manifest.json to ${outDir}`);
	}

	module.exports = { scenarios, toLatin1, fromLatin1 };

	if (require.main === module) {
	main().catch((e) => {
	console.error(e);
	process.exit(1);
	});
	}