Skip to content

Latest commit

 

History

History
474 lines (363 loc) · 21 KB

README.md

File metadata and controls

474 lines (363 loc) · 21 KB
Raw

DDer / MDer

Overview

This repository contains a generic ASN.1/DER parsing library, written in C#, that has the following features:

  • Arbitrary ASN.1/DER object can be decoded into object instances that can be explored programmatically.
  • Such instances can also be built programmatically, and encoded.
  • Decoding supports binary DER as well as Base64 and PEM.
  • A text-based structure syntax can be used to describe values. The syntax is supported as output (convert an object to a human-readable text format), as input (convert a human-writable format into an ASN.1 object that can be encoded), and as a parsing template (exploring a decoded ASN.1 object, to check the presence of expected structures and elements, and to gather values from elements).

DDer.exe is a simple command-line tool that converts encoded ASN.1 object (DER or BER) to the structured text-based syntax that is reminiscent of Lisp (i.e. it uses parentheses). MDer.exe performs the reverse operation. These tools are meant to allow easy analysis of DER-encoded objects (in particular X.509 certificates) and creation of synthetic objects from a text editor or a scripting environment. MDer.exe can additionally use parameters to insert sub-objects and elements at various places. The internal classes DDer and MDer provide programmatic access to these fonctionalities, in particular with powerful parametrization features to allow easier building and parsing of objects as used in some complicated ASN.1-based protocols (e.g. exploring X.509 certificate contents).

ZInt is a self-contained big integer implementation, in pure C#. It is implemented in the ZInt.cs file. This file defines the ZInt struct type, which can be used basically like plain integers thanks to operator overloads. It offers functionalities similar to that of System.Numerics.BigInteger, with the following differences:

  • ZInt is much faster on small values: when the value would fit in a signed 32-bit type (int), then ZInt use entails no dynamic (GC-based) memory allocation. This makes ZInt more appropriate for computations over integers that are usually small but may occasionally exceed the range of int.

  • ZInt also works on old C#/.NET (pre-4.0) that do not support System.Numerics.BigInteger.

  • ZInt has a few extra number-theoretic functions. In particular, it can return the Bézout coefficients when computing a GCD; it also includes primality tests, and random generation.

ZInt is useful for some mathematics-related tasks, e.g. involving some kinds of cryptography, with the following important caveat: it is not constant-time, and thus MUST NOT be used to manipulate private data. ZInt is meant to support research tasks and prototyping.

License

License is MIT-like: you acknowledge that the code is provided without any guarantee of anything, and that I am not liable for anything which follows from using it. Subject to these conditions, you can do whatever you want with the code. See the LICENSE file in the source code for the legal wording.

Installation

The source code is obtained from GitHub; use the "Download ZIP" to obtain a fresh snapshot, or use git to clone the repository. In the source tree, you will find the simple build scripts, build.cmd (for Windows) and build.sh (for Linux and OS X).

The Windows script invokes the command-line compiler (csc.exe) that is found in the v2.0.50727 .NET framework. This framework is installed by default on Windows 7. More recent versions of Windows do not have the .NET 2.0 framework, but a more recent version (4.x or later). Though these framework versions are not completely compatible with each other, TestSSLServer uses only features that work identically on both, so you can compile TestSSLServer with either .NET version. The resulting TestSSLServer.exe is stand-alone and needs no further "installation"; you simply copy the file where you want it to be, and run it from a console (cmd.exe) with the appropriate arguments.

The Linux / OS X script tries to invoke the Mono C# compiler under the names mono-csc (which works on Ubuntu) and dmcs (which works on OS X). On Ubuntu, install the mono-devel package; it should pull as dependencies the runtime and the compiler. On OS X, fetch a package from the Mono project and install it; it should provide the mono command-line tool to run compiled asemblies, and dmcs to invoke the C# compiler.

Usage

On Windows, the compiled DDer.exe and MDer.exe files can be launched directly. On Linux and OS X, use mono DDer.exe and mono MDer.exe.

DDer.exe expects one or several file names; in each file, a single ASN.1 object will be decoded. Files may contain the raw object (in binary), a Base64-encoded object, or a PEM object (Base64-encoding with the -----BEGIN XXX----- and -----END XXX----- headers). Object type is automatically detected. Decoded format is written on standard output, so use a shell redirection to store it in a file. If no file name is provided, then standard input is used.

A file name given as "-" (a single minus character) designates standard input. Since DDer.exe reads each input source as a whole, the "-" source can be given only once on the command-line.

The -n command-line option forces DDer.exe to produce numerical output for all OIDs. Without that option, DDer.exe will recognize some standard OIDs and provide their name (e.g. if the OID is 2.5.29.15, DDer will print its symbolic name id-ce-keyUsage).

The -i pref command-line option makes DDer.exe use the string provided as the pref parameter for each indentation level; normally, pref will consist of one or several spaces. Default indentation prefix is four spaces. If pref is set to none, then no indentation is used, and newlines are suppressed (the output will fit on a single line of text).

MDer.exe expects two arguments: first one is the name of the source file (text encoding of the value), second one is the name of the output file to produce. Output is always binary DER (no Base64). If the name of the input file is "-" (a single minus character), then the source text is read from standard input. Similarly, if the output file name is "-", then the encoded DER object will be written on standard output (as binary).

Extra arguments to MDer.exe are extra values that can be inserted in the output object by using parameter references such as %0, %1... See the MDer class for details. The command-line tool can only use string parameters; the MDer class gives programmatic access to parameterized object building which can use constructed objects.

Syntax

We describe here the base text syntax, as is produced by DDer.exe and expected by MDer.exe. This does not cover use of parameters for programmatic building or parsing; see the MDer class for details.

Text format consists in tokens, with the following rules:

  • Input is supposed to be UTF-8. An UTF-16 input with a starting BOM should work.

  • Whitespace consists in all ASCII control characters (so this includes CR, LF, and tabulations), ASCII space, and U+00A0 unbreakable space. Whitespace may appear anywhere between two tokens, and is insignificant except insofar as it separates tokens. Whitespace characters within string literals is significant.

  • A semicolon ";" starts a comment, that extends to the end of the current line. Comments are equivalent to whitespace.

  • An opening brace "{" starts a comment. The comment extends to the matching closing brace "}". Such comments may nest and may extend over several lines. Comments are equivalent to whitespace.

    Within a brace-comment, MDer counts opening and closing braces so that comment nesting works; it still properly ignores braces that appear as part of semicolon-comments and literal strings. Brace comments can thus be used to "comment out" large chunks of data.

  • A word is a sequence of characters taken among: uppercase ASCII letters, lowercase ASCII letters, digits, "$", "_", "-", "+", ".", and ",". A word stops at the first character which is not part of these sets. Words, in general, are symbolic identifiers; such identifiers are not case-sensitive.

  • The characters "(" (opening parenthesis), ")" (closing parenthesis), "[" (opening bracket) and "]" (closing bracket) are all tokens on their own.

  • String literals begin with a double-quote character, and end on the next double-quote character. Within a string literal, the backslash "\" introduces an escape sequence:

    • "\n" encodes a LF (U+000A).
    • "\r" encodes a CR (U+000D).
    • "\t" encodes a tabulation (U+0009).
    • "\uXXXX", where "XXXX" are exactly four hexadecimal characters, encodes an arbitrary Unicode code point (for code points beyond the first plane, use two sequences for a surrogate pair).
    • "\" followed by any other character encodes that specific character. In particular, "\\" encodes a backslash, and "\"" encodes a double-quote character that does not end the string literal.

  • In some cases, input is parsed as hexadecimal bytes, up to the next closing parenthesis. Whitespace and colon characters (":") are ignored within hexadecimal bytes; other characters shall be hexadecimal digits (ASCII digits, ASCII letters from A to F, ASCII letters from a to f). The total number of hexadecimal digits shall be even. An empty value is allowed (no hexadecimal digit at all).

    When hexadecimal bytes are expected, a sub-object can be provided instead. The sub-object begins with its own opening parenthesis, and follows the object format. In that case, the DER-encoded object is used for the byte values. This supports encoding formats that use BIT STRING or OCTET STRING whose values are themselves ASN.1 objects (e.g. public keys in SubjectPublicKeyIdentifier structures).

An ASN.1 object has the following generic format:

( [ class tag ] name value )

with the following rules:

  • The "[class tag]" sequence is optional; the presence of the initial opening bracket indicates its presence. This sequence encodes an ASN.1 tag value that overrides the normal (universal) tag of the object.

    • The "class" is a word, among universal, application, context and private. This is the tag class. It can be omitted, in which case the tag has class context.

    • The "tag" is a word which is parsed as an integer (signed 32-bit). Tag value must be nonnegative.

  • The "name" is a symbolic identifier for the object type. This indicates the parsing rules for the value; it also sets the object tag, unless overridden with an explicit "[class tag]" sequence.

The following object types are defined:

  • bool and boolean: a BOOLEAN object. The value is a word, which must be one of true, on, yes or 1 (for TRUE), or one of false, off, no or 0 (for FALSE).

  • int and integer: an INTEGER object. The value is a word, which is then parsed as an integer:

    • a starting minus sign is used for a negative integer;
    • the integer is normally decoded in decimal, but an explicit 0x header can be used for hexadecimal, or 0b for binary;
    • arbitrarily long integers are supported (no 32-bit or 64-bit limit).

  • bits: a BIT STRING object. The value should be a word, parsed as an integer with value 0 to 7, followed by hexadecimal bytes (or a sub-object). The initial integer indicates the number of ignored bits in the last value byte; the hexadecimal bytes encode the BIT STRING contents themselves.

  • blob or bytes: an OCTET STRING object. The value consists in hexadecimal bytes (or a sub-object).

  • null: a NULL object. Value is empty.

  • oid: an OBJECT IDENTIFIER. The value is a word, which must be either the OID in decimal-dotted format (e.g. 2.5.29.15) or a symbolic identifier for one of the standard OID (e.g. id-ce-keyUsage).

  • numeric or numericstring: a NumericString. The value is a string literal.

  • printable or printablestring: a PrintableString. The value is a string literal.

  • ia5 or ia5string: an IA5String. The value is a string literal.

  • teletex or teletexstring: a TeletexString. The value is a string literal.

    How exactly Teletex strings are supposed to be encoded is a mystery shrouded in many layers of ill-documented committee meetings. DDer and MDer follow the commonly encountered tradition of using latin-1 (ISO-8859-1).

  • utf8, utf-8 or utf8string: an UTF8String. The value is a string literal.

    Upon decoding, DDer recognizes a BOM (leading U+FEFF) and removes it. Upon encoding, no BOM is produced.

  • utf16, utf-16, bmp or bmpstring: a BMPString. The value is a string literal.

    While a BMPString is nominally limited to the first Unicode plane, DDer and MDer use UTF-16 (surrogate pairs) for upper planes. When decoding, DDer recognizes a BOM (leading U+FEFF) and removes it; decoding uses big-endian convention, unless a leading little-endian BOM is present. When encoding, MDer uses big-endian with no BOM.

  • utf32, utf-32, universal or universalstring: a UniversalString. The value is a string literal.

    When decoding, DDer recognizes a BOM (leading U+FEFF) and removes it; decoding uses big-endian convention, unless a leading little-endian BOM is present. When encoding, MDer uses big-endian with no BOM. While UniversalString values are 32-bit numbers, code points must have a value in the 0 to 0x10FFFF range.

  • utc or utctime: a UTCTime. The value is a string literal.

    When encoding, MDer does not check time values; the value is expected to be in a proper format. When decoding, DDer interprets time values along a proleptic Gregorian calendar.

  • gentime or generalizedtime: a GeneralizedTime. The value is a string literal.

    When encoding, MDer does not check time values; the value is expected to be in a proper format. When decoding, DDer interprets time values along a proleptic Gregorian calendar.

  • set: a SET. The value is a concatenation of 0, 1 or more sub-objects.

    Objects will be encoded in the order they are provided (thus, this will be strict DER only if objects are already in DER-mandated tag order).

  • sequence: a SEQUENCE. The value is a concatenation of 0, 1 or more sub-objects, encoding in the order they are provided.

  • setof: a SET OF object. The value is a concatenation of 0, 1 or more sub-objects.

    When encoding a SET OF, MDer will sort the value elements in lexicographic ascending order of their respective DER-encoded values, as mandated by DER.

Notes

Sub-Objects for Blobs

When decoding a BIT STRING or OCTET STRING, DDer will check if the value is itself a valid DER-encoded object. If it is, then the "sub-object" syntax is used; otherwise, hexadecimal bytes are used. It can thus happen that a binary value turns out to be interpreted as a sub-object; e.g., if a 20-byte key identifier in a certificate (nominally a hash output) starts with bytes 0x04 0x12, then it will "look like" a DER-encoded OCTET STRING and will be decoded as such. In practice, this occurs very rarely.

DDer takes care not to use a sub-object if MDer would not reencode the value exactly. Such things may happen because many encoding variants are accepted (e.g. endianness in character strings, minimality of integer values, BER indefinite lengths...) but not transcribed in the text output of DDer. If the output would not be reencoded exactly, then hexadecimal output is produced instead of a sub-object.

Checks on Strings

DDer and MDer check that string contents match their respective types. Therefore, if you want to produce (for testing reasons) an "invalid" string, you should use another type with a tag override. For instance, this:

([universal 19] ia5 "foo&bar")

produces a value with the tag for PrintableString (universal 19), but the contents include a "&" character which is not allowed in a PrintableString. When decoding it back, DDer will complain ("unexpected character U+0026 in string of type 19").

Tentative String Decoding

If an OCTET STRING or a BIT STRING contains bytes that appear to be "plain ASCII" (all byte values are 9, 10, 13, or between 32 and 126, inclusive), then DDer assumes that the bytes may be an encoded ASCII string, and prints out the string contents as a string literal enclosed in a brace-delimited comment. This is done in addition to the normal hexadecimal printout for such values.

This is used for instance with GeneralName structures, which are commonly encountered in certificates for encoding URL (the format for such an URL is an IA5String with a contextual tag override of value 6, so the fact that it is an IA5String is not known to DDer).

GeneralString and TeletexString

The ASN.1 GeneralString and TeletexString types are handled as ISO-8859-1, aka "Latin-1", i.e. as sequences of characters with a one-byte-per-character mapping. This does not conform to the de jure ASN.1 standard, but aligns with common practice in protocols that use such types. In any case, this allows reading and writing arbitrary string contents through \x escape sequences in string literals.

Date and Time Processing

When decoding a time string (UTCTime or GeneralizedTime), DDer produces both the string contents as a string literal, and a human readable date and time in a brace-delimited comment. Note that a DateTime object is used internally, with the following consequences:

  • The date is converted to UTC. Any time zone offset in the string is processed and applied.
  • In GeneralizedTime values, only up to seven fractional digits are read (i.e. 'ticks' of 100ns); other digits are ignored.
  • Year 0 is not supported (DateTime starts at year 1).
  • The calendar is a "proleptic Gregorian calendar", which means that the Gregorian rules, theoretically valid only from October 15th, 1589 AD onwards, are retroactively applied to previous dates.
  • Leap seconds are ignored: if the second count is 60, it is internally converted to 59. No attempt is made to check whether the leap second really occurred, or even might have occurred at that date, as per UTC rules.

All this only matters for the human-readable printout, which is a comment. The string literal still contains the complete string, as it was received.

Memory Allocation Behaviour

When decoding, the underlying ASN.1 library copies the whole input buffer exactly once, then keeps references within that buffer. This should make it immune to, or at least robust against, malformed inputs with extra-large announced lengths in headers: if a 20-byte object begins with a header that claims the value length to be one gigabyte, DDer will not allocate a one-gigabyte array.

Accepted Variants

DDer accepts as input a number of variants which are not strict DER, but are still unambiguous. These variants include the following:

  • Non-minimal tag and length encoding: encodings that use more bytes than the strict minimum are accepted.

  • Indefinite lengths (BER) are accepted. Note that DDer does not perform reassembly of parts for primitive values that were split into chunks.

  • Non-zero values for the value byte of a BOOLEAN are considered to encode TRUE (in strict DER, only 0xFF may be used for TRUE).

  • Non-minimal integer encodings: extra leading 0x00 (for nonnegative integers) or 0xFF (for negative integers) are tolerated.

  • Ignored bits in the last byte of a BIT STRING may have non-zero value (strict DER mandates that these bits are set to zero).

  • Non-minimal OID element encodings: each numerical element uses the "7E" encoding (big-endian, 7 bits per byte) but extra leading bytes are allowed.

  • UTF8String, BMPString and UniversalString may use an explicit BOM, which is removed upon decoding.

  • Little-endian decoding is accepted for BMPString and UniversalString if a BOM is included, and indicates such an encoding (strict DER mandates big-endian with no BOM).

  • When decoding UTF8String and UniversalString, surrogates are accepted, and pairs will be internally reunited.

Numerical Limits

DDer and MDer operate under the following constraints:

  • The complete ASN.1 object must fit in RAM. Processing is not streamed. Moreover, 32-bit signed integers are used for lengths, so total object length may not exceed 2 gigabytes.

  • Tag value must fit on a 32-bit signed integer.

  • Individual OID components can have arbitrary size; be aware that many other existing ASN.1 libraries cannot handle components beyond 32 bits or so, therefore larger components should be avoided for better interoperability.

  • Dates must have year 1 to 9999. Year 0 is not supported.

Values which exceed these limits are detected and trigger exceptions.

Author

Question and comments can be sent to: Thomas Pornin <[email protected]>