pdf2john: fix missing /Length for V4 PDFs using crypt filters

run/lib/PDF.pm

@@ -1166,6 +1166,28 @@ sub DecryptInit($$$)
             }
             # set "_aesv2" or "_aesv3" flag in %$encrypt hash if AES encryption was used
             $$encrypt{'_' . lc($1)} = 1 if $cryptMeth =~ /^\/(AESV2|AESV3)$/;
+
+            # V4 PDFs may omit the top-level /Length; derive it from the
+            # applicable crypt filter.  Per the PDF spec, StmF/StrF default
+            # to Identity (no encryption), so resolve the actual filter name
+            # rather than assuming StdCF.
+            # CF /Length is in bytes; top-level /Length is in bits.
+            unless (defined $$encrypt{Length}) {
+                my $cf_len;
+                for my $key ('StmF', 'StrF') {
+                    my $name = $$encrypt{$key} || '';
+                    $name =~ s{^/}{};
+                    next unless $name and $name ne 'Identity';
+                    if (ref $$encrypt{CF}{$name} eq 'HASH') {
+                        $cf_len = $$encrypt{CF}{$name}{Length};
+                        last if defined $cf_len;
+                    }
+                }
+                $$encrypt{Length} = defined $cf_len ? $cf_len * 8
+                    : $cryptMeth eq '/AESV2' ? 128
+                    : $cryptMeth eq '/AESV3' ? 256
+                    : 40;
+            }
         }
         if ($ver == 5) {
             # validate OE and UE entries

run/pdf2john.py

@@ -71,10 +71,38 @@ def __init__(self, file_name: str, strict: bool = False):
                 raise RuntimeError("File not encrypted")
 
             self.algorithm: int = self.encrypt_dict.get("/V")
-            self.length: int = self.encrypt_dict.get("/Length", 40)
             self.permissions: int = self.encrypt_dict["/P"]
             self.revision: int = self.encrypt_dict["/R"]
 
+            if self.algorithm == 4:
+                # For V4 the key length lives in the applicable crypt filter,
+                # not in the top-level /Length (which may be absent or written
+                # in bytes instead of bits by some generators).
+                # /AESV2 and /AESV3 imply fixed sizes; for /V2 (RC4) we read
+                # /Length directly, but must handle the PDF 1.7 (bits) vs
+                # PDF 2.0 (bytes) ambiguity: any value < 40 must be bytes
+                # because 40-bit is the minimum valid key length.
+                cf = self.encrypt_dict.get("/CF") or {}
+                cfm = cf_length = None
+                for fkey in ("/StmF", "/StrF"):
+                    fname = self.encrypt_dict.get(fkey)
+                    if fname and fname != "/Identity":
+                        fdict = cf.get(fname) or {}
+                        cfm = fdict.get("/CFM")
+                        cf_length = fdict.get("/Length")
+                        break
+                if cfm == "/AESV2":
+                    self.length: int = 128
+                elif cfm == "/AESV3":
+                    self.length: int = 256
+                elif cf_length is not None:
+                    self.length: int = cf_length * 8 if cf_length < 40 \
+                                       else cf_length
+                else:
+                    self.length: int = 40
+            else:
+                self.length: int = self.encrypt_dict.get("/Length") or 40
+
     @property
     def document_id(self) -> bytes:
         return self.pdf.document_id[0]