juikim / acvpparser

ACVP Parser for invocation of cryptographic implementations using the ACVP JSON test vectors

Generic ACVP JSON Parser

This parser implements the ACVP protocol used by NIST for the automated CAVS testing (Automated Cryptographic Validation Program - ACVP).

This parser processes JSON files that are already downloaded from the ACVP server. It invokes the cryptographic implementation and generates the test response JSON data as defined by the ACVP protocol.

The entire ACVP server interaction including download of test vectors and upload of test responses must be handled with a separate tool, like the ACVP Proxy.

The ACVP Proxy is proudly supported by atsec information security corp.

Usage

The ACVP Parser operates on one file at a time. Simply invoke

acvp-parser <testvector-request.json> <testvector-response.json>

NOTE: If you want to use the ACVP Parser in unison with the ACVP Proxy, you MUST use the file name testvector-response.json to hold the test responses!

If you want to compare two test responses with each other, simply use

acvp-parser -e <expected-response.json> <testvector-response.json>

To increase verbosity, use -v one or more times.

Concept of Parser

The parser is implemented as a data-driven engine. The JSON parsing code is separated from the definitions of which parts of the JSON input stream to parse and which format the JSON input stream is expected to have.

This implies that an implementation for a specific test definition, such as the AEAD or symmetric cipher algorithm, the test definitions do not contain any code, but only the data specifying the expected JSON format and pointing to the local variables where the data is to be stored into.

In addition to the parsing of the input data, the data-driven model allows specifying the test execution convention and which local variables shall be exported to the JSON stream holding the test results.

The data model is defined with the different data structures found in parser_common.h.

As a rule of thumb: if there is any parsing related code except the invocation of process_json in the test definition specification of the different parser_*.c files (except parser_common.c), there is a programming error and a violation of the basic concept.

Architecture

The parser consists of 3 layers:

1. parser.c implements a linked list where all parsers with usable callbacks are registered. Each parser must register with a particular cipher identifier. When a JSON file is parsed, the cipher identifier is searched for and the respective parser handler is called if a match is found.

2. parser_*.c implement the register functions registering with parser.c In addition, they implement the handler function that is triggered when parser.c is finished. The idea now is that the respective parser is invoked with the JSON file it can handle. The parser now implements all logic to:

   • parse the JSON file

   • obtain all relevant data from the JSON file and store it in a parser specific data structure

   • invoke the backend implementation that interfaces with the cipher mechanism implementation

   • create the JSON response file with the correct format.

3. backend_*.c implement the backend that uses the parser data structure to invoke a specific crypto library. The backend is not needed to implement all backend functions of all parsers. If a callback from a parser is not supported, it must be marked as NULL. If a parser identifies a NULL handler, it returns with an error such that the JSON file cannot be processed.

Backend Selection

To select the appropriate backend, invoke make with the right option.

When invoking make the implemented compile options are given.

Building

Prerequisite

The ACVP parser requires the presence of the POSIX APIs.

Compiling

As discussed above, the backend must be chosen out of the list given with make.

NOTE: If the backend is chosen, the file backend_.c must exist.

Example: To compile the OpenSSL support, invoke: make openssl.

Building on Cygwin

The following steps are required to build the ACVP Proxy on Cygwin:

• Verify that the required cygwin packages are installed:

  • gcc

  • make

• Build acvp-parser.exe using with a build target as outlined above.

Backend-specific Hints: Linux Kernel

The ACVP test for the Linux kernel requires a kernel module that is provided with the directory backend_interfaces/kcapi/. You need to compile it and load it into the kernel.

The ACVP Parser tool must now be executed as root to access the kernel module's interface files at /sys/kernel/debug/kcapi-cavs.

Backend-specific Hints: Libreswan

To execute the Libreswan ACVP test, you need to compile an application that will be invoked by the ACVP parser. See backend_interfaces/libreswan/README for details on compiling the interface application.

Ensure that the inteface application is found via the PATH environment variable.

Backend-specific Hints: Strongswan

To execute the Strongswan ACVP test, you need to compile an application that will be invoked by the ACVP parser. See backend_interfaces/strongswan/README for details on compiling the interface application.

Ensure that the inteface application is found via the PATH environment variable.

Backend-specific Hints: OpenSSH

To execute the OpenSSH ACVP test, you need to compile an application that will be invoked by the ACVP parser. See backend_interfaces/openssh/README for details on compiling the interface application.

Ensure that the inteface application is found via the PATH environment variable.

Backend-specific Hints: BoringSSL

Please modify the Makefile to point to the static library to compile against.

Helper Scripts

Test Execution Scripts

Helper scripts for executing multiple test vectors for a given module are provided in the helper/ directory. Note, the script exec_lib.sh is not intended to be executed directly, but provides a library for the other scripts.

The scripts are all named exec_<modulename>.sh.

Regression Test Scripts

The ACVP Parser can also be used as a regression test system. The regression tests are executed with a helper script given for each module.

The scripts are named exec_<modulename>_regression.sh.

It is permissible to provide one or more vsIDs as command line parameters for the regression test script. In this case, only the referenced vsIDs regression tested.

Note, the regression test is deactivated for test vectors that are based on random numbers to be generated by the module.

Backend Implementation

A backend is the code that connects the parser with a particular cryptographic implementation. A backend implements the code that invokes the cryptographic implementation to be tested using the data provided with various data structures.

The backend implementation is invoked once for each individual cipher test. It does not need to carry state information other than what is required by the particular cryptographic implementation it interfaces with.

The backend is unrelated to any formatting or other CAVS/ACVP logic.

A backend implementation is achieved with the following steps:

1. Include backend_common.h from the parser. There are no other header files from the parser needed. The `backend_common.h includes the various header files of the different parsers for the cipher implementations.

2. Select which cipher implementations the backend shall handle. For each cipher implementation, there is a parser_*.h header that defines the data structure(s) used to exchange data between the parser and the backend as well as the interface functions that need to be implemented by the backend. For the following example, a SHA hash implementation shall be tested. The corresponding header file is parser_sha.h which contains the struct sha_data definition. The documentation explains which member variables are provided by the parser and which data is expected to be returned by the backend. Furthermore, the parser_sha.h header defines the struct sha_backend function pointer data structure that must be filled by the backend.

3. The backend implements the functions defined by the function pointer data structure. In case of SHA, the hash_generate function must be implemented following the definition of the function in parser_sha.h.

4. The backend must now register its implementation by defining the function pointer data structure and registering it with the register call found in parser_sha.h. The function pointer data structure instance must be marked with the ACVP_DEFINE_CONSTRUCTOR macro to ensure that it is invoked by the operating system loader during load time of the executable. The following example illustrates this registering:

   static int backend_sha_generate(struct sha_data *data, flags_t parsed_flags)
   {
   ...
           <invoke the backend's SHA function>;
   ....
   }
   
   static struct sha_backend backend_sha =
   {
           backend_sha_generate,
   };
   
   ACVP_DEFINE_CONSTRUCTOR(backend_sha_backend)
   static void backend_sha_backend(void)
   {
            register_sha_impl(&backend_sha);
   }
   

5. Ensure that the C file(s) implementing the backend is compiled when compiling the parser.

The example outlines that apart from including the backend_common.h, no further change is needed to link the backend implementation with the parser.

Note, the parser will guarantee that only one backend can register its callback function pointer data structure for a given cipher. I.e. the function register_sha_impl used in the example above MUST ONLY be used once in the entire code base of the parser that is compiled. Subsequent register operations for a particular cipher type will fail.

Author

Stephan Mueller [e-mail] Copyright (C) 2018 - 2020