I do apply a signature to a pdf document via delayed signing(SignDeferred) using IText. The process contains the following steps:
The whole process works and i end with a pdf document where the signature is set and is valid.
The original pdf is a PDF-A1a but the resulting pdf is not a valid PDF-A1a anymore. I am aware that there is a documentation about IText PDF-A support (https://kb.itextpdf.com/home/it7kb/ebooks/itext-7-jump-start-tutorial-for-java/chapter-7-creating-pdf-ua-and-pdf-a-documents), but this seems to not apply since i don't change the content of the document.
My question: How can i apply a signature using deferred signing and keep the PDF-A1a in the resulting document?
Note: If i do apply a signature directly (Without SignDeferred), the resulting pdf is still a PDF-A1a, but i do have to use SignDeferred Note: I do use https://www.pdfen.com/pdf-a-validator for checking pdf-A
The following is a complete code sample sample with everything required in one file. It only requires the references to itext and BouncyCastle and the path to a self signed certificate
using iText.Kernel.Pdf;
using iText.Signatures;
using Org.BouncyCastle.Pkcs;
using Org.BouncyCastle.Security;
using System;
using System.Collections.Generic;
using System.IO;
namespace DeferredSigningTestConsole
{
class Program
{
static string SignatureAttributeName = "DeferredSignature";
static string CertificatePath = @"C:\temp\PDFA\PdfATestCert.2pfx.pfx";
static string CertificatePassword = "test";
static void Main(string[] args)
{
var signedPdf = SignPdf(System.IO.File.ReadAllBytes(@"C:\temp\PDFA\PDF_A1a.pdf"));
System.IO.File.WriteAllBytes(@"C:\temp\PDFA\signed.pdf", signedPdf);
}
public static byte[] SignPdf(byte[] pdfToSign)
{
byte[] hash = null;
byte[] tmpPdf = null;
//Step #1 >> prepare pdf for signing (Allocate space for the signature and calculate hash)
using (MemoryStream input = new MemoryStream(pdfToSign))
{
using (var reader = new PdfReader(input))
{
StampingProperties sp = new StampingProperties();
sp.UseAppendMode();
using (MemoryStream baos = new MemoryStream())
{
var signer = new PdfSigner(reader, baos, sp);
signer.SetCertificationLevel(PdfSigner.NOT_CERTIFIED);
signer.SetFieldName(SignatureAttributeName);
DigestCalcBlankSigner external = new DigestCalcBlankSigner(PdfName.Adobe_PPKLite, PdfName.Adbe_pkcs7_detached);
signer.SignExternalContainer(external, 121743);
hash = external.GetDocBytesHash();
tmpPdf = baos.ToArray();
}
}
//Step #2 >> Create the signature based on the document hash
byte[] signature = GetSignatureFromHash(hash);
//Step #3 >> Apply the signature to the document
ReadySignatureSigner extSigContainer = new ReadySignatureSigner(signature);
using (MemoryStream preparedPdfStream = new MemoryStream(tmpPdf))
{
using (var pdfReader = new PdfReader(preparedPdfStream))
{
using (PdfDocument docToSign = new PdfDocument(pdfReader))
{
using (MemoryStream outStream = new MemoryStream())
{
PdfSigner.SignDeferred(docToSign, SignatureAttributeName, outStream, extSigContainer);
return outStream.ToArray();
}
}
}
}
}
}
public static byte[] GetSignatureFromHash(byte[] hash)
{
FileStream fs = new FileStream(CertificatePath, FileMode.Open);
Pkcs12Store store = new Pkcs12Store(fs, CertificatePassword.ToCharArray());
String alias = "";
foreach (string al in store.Aliases)
if (store.IsKeyEntry(al) && store.GetKey(al).Key.IsPrivate)
{
alias = al;
break;
}
AsymmetricKeyEntry pk = store.GetKey(alias);
X509CertificateEntry[] chain = store.GetCertificateChain(alias);
List<Org.BouncyCastle.X509.X509Certificate> c = new List<Org.BouncyCastle.X509.X509Certificate>();
foreach (X509CertificateEntry en in chain)
{
c.Add(en.Certificate);
}
PrivateKeySignature signature = new PrivateKeySignature(pk.Key, "SHA256");
String hashAlgorithm = signature.GetHashAlgorithm();
PdfPKCS7 sgn = new PdfPKCS7(null, c.ToArray(), hashAlgorithm, false);
DateTime signingTime = DateTime.Now;
byte[] sh = sgn.GetAuthenticatedAttributeBytes(hash, null, null, PdfSigner.CryptoStandard.CMS);
byte[] extSignature = signature.Sign(sh);
sgn.SetExternalDigest(extSignature, null, signature.GetEncryptionAlgorithm());
return sgn.GetEncodedPKCS7(hash, null, null, null, PdfSigner.CryptoStandard.CMS);
}
}
internal class DigestCalcBlankSigner : IExternalSignatureContainer
{
private readonly PdfName _filter;
private readonly PdfName _subFilter;
private byte[] _docBytesHash;
internal DigestCalcBlankSigner(PdfName filter, PdfName subFilter)
{
_filter = filter;
_subFilter = subFilter;
}
internal virtual byte[] GetDocBytesHash()
{
return _docBytesHash;
}
public virtual byte[] Sign(Stream docBytes)
{
_docBytesHash = CalcDocBytesHash(docBytes);
//If we retun the signature bytes, GetAuthenticatedAttributeBytes will throw an exception
//Not clear how this should be done
return new byte[0];
}
public virtual void ModifySigningDictionary(PdfDictionary signDic)
{
signDic.Put(PdfName.Filter, _filter);
signDic.Put(PdfName.SubFilter, _subFilter);
}
internal static byte[] CalcDocBytesHash(Stream docBytes)
{
byte[] docBytesHash = null;
docBytesHash = DigestAlgorithms.Digest(docBytes, DigestUtilities.GetDigest(DigestAlgorithms.SHA256));
return docBytesHash;
}
}
internal class ReadySignatureSigner : IExternalSignatureContainer
{
private byte[] cmsSignatureContents;
internal ReadySignatureSigner(byte[] cmsSignatureContents)
{
this.cmsSignatureContents = cmsSignatureContents;
}
public virtual byte[] Sign(Stream docBytes)
{
return cmsSignatureContents;
}
public virtual void ModifySigningDictionary(PdfDictionary signDic)
{
}
}
}
It seems the reason why the signed pdf is no longer a valid PDF-A1a is the estimated size for the signature. I have used a value of about 120kb for the signature.
//doesn't work
signer.SignExternalContainer(external, 121743);
//does work
signer.SignExternalContainer(external, 65000);
The concept of this is documented in the ebook "Digital Signatures for PDF documents" from iText.
It seems in order to get a valid pdf-A1a the maximum size is limited to 65kb.
I now will have to test whether this works when i add a visual representation (signature image) since this was the reason i did choose such a large estimated size.
Edit: I did some more testing and i am now able to produce valid pdf-A document with signatures: The pdf are now valid pdf-A with the changed estimated size:
When adding a visual representation (image), pdf-A1a and pdf-A1b are no longer valid.
A transparent soft mask is present. Beginning with PDF 1.4 transparency is supported. Some PDF-based ISO standards prohibit the use of transparency.
But this is another problem which i now try to figure out.