textcoll.sh

#!/usr/bin/env bash

export BINDIR=$(dirname $0)/../bin
export HELPER=$BINDIR/md5_diffpathhelper
export TEXTCOLL=$BINDIR/md5_textcoll

export DIFFPATH=$BINDIR/../src/md5textcoll/path2.txt
MDIFF="--diffm5 11"

prefixfile=$1

if [ -z $prefixfile ]; then
	prefixfile=dummy.prefix.bin
fi
if [ ! -f $prefixfile ]; then
	touch $prefixfile
fi

ALPHABET="!#&'()*+,-./0123456789:;<=?@ABCDEFGHIJKLMNOPQRSTUVWXYZ[\]^_abcdefghijklmnopqrstuvwxyz{|}~"
#ALPHABET="abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789.,_-~=+:;|?@#^&*"
#ALPHABET="abcdefghijklmnopqrstuvwxyzABCDEFGHIJKLMNOPQRSTUVWXYZ0123456789"

# First block: byte21 has a +4 difference: you must ensure that this difference is possible in the alphabet, e.g., have both A and E(=A+4)
FIRSTBLOCKBYTES='--byte21 :> --byte22 ABCDEF0123456789 --byte23 < --byte24 cdefghijklmnoqrstuvwxyzCDEFGHIJKLMNOQRSTUVWXYZ'

# Second block: 
# - keep the alphabet of bytes 0-7 large: otherwise there could be no solutions
# - keep the alphabet of bytes 56-63 large: to make the search fast
# - if you want to set many bytes of the 2nd block then you should customize the 2nd block search in src/md5textcoll/block2.cpp
SECONDBLOCKBYTES=''

# First convert the differential path from text to binary for the attack
if [ ! -f textcoll_path.bin.gz ]; then
	$HELPER $MDIFF --pathfromtext --inputfile1 $DIFFPATH --outputfile1 textcoll_path.bin.gz || exit 1
fi



# TEXTCOLL Stage 1: preparation to find solutions over Q7-Q24

# REDUCING MEMORY USAGE: If textcoll crashes due to out of memory then you can reduce memory usage by decreasing the exponent of the maxLUT constant:
# `static const size_t maxLUT = 1ULL<<29;` in src/md5textcoll/block1.cpp

if [ ! -f Q7Q24.bin.gz ]; then
	( $TEXTCOLL $MDIFF $FIRSTBLOCKBYTES --prepare --pathfile textcoll_path.bin.gz --alphabet $ALPHABET --prefixfile ${prefixfile} | tee prepare.log ) || exit 1
	echo "Solutions stored in Q7Q24.bin.gz."
	echo "It's possible to try to control more bytes by editing this script and deleting Q7Q24.bin.gz."
fi



# TEXTCOLL Stage 2: first block attack, but only if there is NO first block found yet
collfile=`ls textcoll1_block1_[0-9]*.txt | head -n1`

if [ "$collfile" = "" ]; then
	echo "Starting search for first near-collision block in 10 seconds..."
	sleep 10
	
	( $TEXTCOLL $MDIFF $FIRSTBLOCKBYTES --firstblock --pathfile textcoll_path.bin.gz --alphabet $ALPHABET --prefixfile ${prefixfile} | tee firstblock.log ) || exit 1
	collfile=`ls textcoll1_*.txt | head -n1`
fi

echo "Found first near-collision block in file: $collfile"
cat ${prefixfile} ${collfile} > partial_solution.txt



# TEXTCOLL Stage 3: do second block attack 
# Basically denBoer-Bosselaers pseudo-collision attack, where trying 2^46 random messages is sufficient
echo "Starting search for second near-collision block in 10 seconds..."
sleep 10

$TEXTCOLL $MDIFF $SECONDBLOCKBYTES --secondblock --alphabet $ALPHABET --prefixfile partial_solution.txt || exit 1




# SUCCESS!! so now write full solutions

# second block file
coll2file=`ls textcoll1_block2_[0-9]*.txt | head -n1`

# other first block file
collfile2=`echo $collfile | sed "s/textcoll1/textcoll2/"`

cat ${prefixfile} ${collfile} ${coll2file} > final_collision1.txt
cat ${prefixfile} ${collfile2} ${coll2file} > final_collision2.txt

md5sum final_collision*
sha1sum final_collision*

echo -e "\n"
echo "========= final_collision1.txt =========="
cat final_collision1.txt
echo -e "\n"
echo "========= final_collision2.txt =========="
cat final_collision2.txt
echo -e "\n"