crocs-muni / eacirc Goto Github PK

Automatic problem solver based on circuit-like representation and genetic programming

C++ 96.10% CMake 3.90%

randomness-testing cryptography research-tool hash-functions stream-ciphers block-ciphers

eacirc's Introduction

EACirc is a framework for automatic problem solving. It can be utilized as randomness testing tool similar to statistical batteries (NIST STS, Dieaharder, TestU01), for instance for analysis of cryptographic function outputs.

It uses supervised learning techniques based on metaheuristics to construct adapted distinguisher of two input data streams. The distinguisher can be represented as hardware-like circuits or algebraic polynomial.

The Framework

This repository contains EACirc core and code for data stream generation (mainly eSTREAM and SHA-3 candidates). Further tools are:

Randomness Testing Toolkit (RTT),
tools for GRID computations
Oneclick a tool for BOINC computation (deprecated)

For more information and details see project wiki pages.

Quickstart

Requirements are gcc5 (clang3.6), git and cmake 3.4.

Sample build on Metacentrum

module add gcc-5.3.0 cmake-3.6.1
git clone --recursive [email protected]:crocs-muni/eacirc.git src
mkdir src/build
cd src/build
ccmake ..
make

Authors

The framework is developed at the Centre for Research on Cryptography and Security (formerly Laboratory of Security and Applied Cryptography), Masaryk University, Brno, Czech Republic.

Petr Švenda 2008-now (project lead, initial implementation)
Jiří Novotný 2014-now (build system, CUDA, main developer)
Michal Hajas 2015-now (Java bytecode emulator)
Dušan Klinec 2012-now (polynomial distinguisher)
Karel Kubíček 2014-now (TEA, metaheuristics)
Ľubomír Obrátil 2014-now (RTT, Oneclick)
Marek Sýs 2013-now (statistics evaluation, polynomials)
Martin Ukrop 2012-now (framework model, refactoring, SHA-3 & CAESAR candidates testing, supporting tools)

Former participation:

Milan Čermák 2012-2013 (CUDA)
Ondrej Dubovec 2011-2012 (SHA-3 candidates testing)
Matěj Prišťák 2011-2012 (object model and refactoring, XML support, eStream candidates testing)
Zdenek Říha 2013-2016 (Java bytecode emulator)
Tobiáš Smolka 2011-2012 (BOINC related support)

eacirc's People

Contributors

Stargazers

Watchers

Forkers

gruve-p vieteg gitcollect anhtuan98 epsilonsolutions praveenmunagapati dalavancloud filipgontko-zz

eacirc's Issues

comparing audiofiles

random - noise - mp3 noise - metal

Implement other hash functions and ciphers

SHA-1, SHA-2, MD5, RC-4

circuit printing function

refactor circuit printing fuction, separate formats into own methods

determine state components

what are state components, that need to be saved for the computation to continue?

automatic program versioning

Make program version (ideally current git commit hash or its prefix) build into the program binary -> when run, log file can automatically contain version -> better experiment replication

state documentation

add state file and population file description to wiki

evaluators refactoring

SHA3 project documentation

evaluators documentation

Summarize existing evaluators, polish code

integrate Memory circuit code into eacirc-dev

project data space + backup

allocate server space for project data
make easy, presentable form of current results
backup and summarize all computation to date

command line arguments

option for verbose output (initializing generators, ... => will simplify configuration errors)
option for writing program log (like verbose output and error messages) to file instead of stdout (for batch processing)
arguments can be set in arbitrary order

dependency of AAM on set change frequency

behaviour if config.xml not found

The app should terminate and state that confix.xml was not found.
(now it terminates in EncryptorDecryptor's assert(false) as if stream number was incorrectly set)

Refactor PrintCircuitMemory()

Four different outputs are provided: XML, DOT, C, TXT. Refactor with Template Method pattern?

unify iv/key/plaintext generation in eStream test vectors

iv & key now generated when initializing encryptor decryptor
plaintext generated when generating test vectors

=> unify and make all tests in test set independent

binary output format to xml?

Should binary output format (population, genome) be saved in xml?

increases integrity
better readability
protection from incorrect processing

remake scripts to better language

ideally something scriptable (Python?)

memory initialization in encryptorDecryptor

unify memory initialization in encryptorDecryptor

command line arguments

Program does not check validity of arguments and continues working ignoring them.
=> if typo in argument, application does something else as user thinks

plus, error message should be displayed, when e.g. files with test data were not found

project specific evaluators

allow project specific evaluators - design object model for them

write tests for recommencing computation

tests:

generating system seed and loading from this state should be the same
making 20+20 generations should equal making 40 generations

use cxxtest
http://cxxtest.com/

improve eStream candidates table

design EACirc icon

get other results for SHA-3 candidates

compute distinguishers for other SHA-3 candidates not used in Bc Thesis

review licences

review licence for EACirc and for used components, add licence files

unify GAlib usage

project now uses 2 verions of GAlib

"ga" (older version, originally for windows only)
"gaunix" (newer version, both unix and windows)

Project should be unified to use "gaunix" ("ga" should be deleted)
Note: beware of differences between "ga" and "gaunix"

statistics

Which stats do we want?

graph files?
special averages?
make directives for specific stats?

review GAlib callbacks

-> initialiser will generate random connectors in selector layer as well
-> mutator will distinguish function mutation probability and connector mutation probability
-> initialiser can generate fixed layers (each function once), to be combined with fucntion mutation probability of 0

loading state

programme will be able to load saved state and continue computatuion

refactor evaluateCircuit

refactor and polish circuit interpreter (with planned enhancements in mind )

unify fitnessProgress stats

create new set of node functions

-> change circuit emulator
-> change checker
-> change circuit printing (text, code, dot)

new set of functions:
nop, nand, xor, or, and, nor, not, >>, <<, >>cycle, <<cycle

file distinguisher

Vytvorit projekt, ktory hlada distinguisher medzi datami z 2 vstupnych suborov
-> vyskusat rozoznavat chorvatske/nemecke nahodne data
-> vyskusat odlisit vystupy viacerych sifier
-> pouzitie mimo, napr. odlisovat binarne konfiguraky roznych programov? styl hudby z binarnej interpretacie mp3? Kadeco by sa dalo vyskusat.

analytical design

make analytical design separating EACirc framework from individual projects (eStream, SHA3, ...) - done

get the design diagram into documentation

mutator probabilites differentiation

allow for different mutator probabilities for functions and connectors

review info written to hr-test-vector file

And add description to documentation.

integrate CUDA branch into main stream

TSC data for 1-8 rounds

compute TSC 1-8, all with STS-NIST and Dieharder

eStream ciphers LEX, Trivium, CryptMT not working

LEX: compiles ok, seems to run on Ubuntu/GCC but does not on Win7/MinGW
Trivium: compiles ok, does not run on either Ubuntu or Win7
CryptMT: supposed not to work, seems ok on Ubuntu/GCC configuration

create SHA-3 candidates table

possible input lengths?
number of rounds?
compilation state?

variable genome size

Genome size will not be a constant, will be dynamically counted according to circuit properties (won't be loaded via config.xml)

main program cycle documentation

create picture representing the main program cycle

when are test vectors generated
when are things reseeded
when is state and population saved
...