def add_binary_feature(feature, n1, n2):
if type(n1) in (GivReg, Reg, Flag):
n1_info = coarse(n1)
else:
n1_info = fine(n1)
if type(n2) in (GivReg, Reg, Flag):
n2_info = coarse(n2)
else:
n2_info = fine(n2)
if isinstance(n1, Reg):
n1.features.add('L' + feature.format(coarse(n1), n2_info))
if isinstance(n2, IndirectOffset):
n1.features.add('L' + feature.format(coarse(n1), coarse(n2_info)))
elif isinstance(n1, IndirectOffset):
n1.features.add('L' + feature.format(coarse(n1), n2_info))
n1.features.add('L' + feature.format(fine(n1), n2_info))
if isinstance(n2, IndirectOffset):
n1.features.add('L' + feature.format(coarse(n1), coarse(n2_info)))
n1.features.add('L' + feature.format(fine(n1), coarse(n2_info)))
if isinstance(n2, Reg):
n2.features.add('R' + feature.format(n1_info, coarse(n2)))
if isinstance(n1, IndirectOffset):
n2.features.add('R' + feature.format(coarse(n1), coarse(n2)))
elif isinstance(n2, IndirectOffset):
n2.features.add('R' + feature.format(n1_info, coarse(n2)))
n2.features.add('R' + feature.format(n1_info, fine(n2)))
if isinstance(n1, IndirectOffset):
n2.features.add('R' + feature.format(coarse(n1), coarse(n2)))
n2.features.add('R' + feature.format(coarse(n1), coarse(n2)))
def initialize(self):
for direct_offset in self.binary.direct_offsets.values():
self.add_edge(direct_offset, direct_offset.ttype, 'NAME[{}]TTYPE',
coarse)
node_type = make_node_type(coarse(direct_offset), self.binary)
self.add_edge(direct_offset, node_type, 'NODE[COARSE]TYPE')
# node_type = make_node_type(coarse(direct_offset.ttype), self.binary)
# self.add_edge(
# direct_offset.ttype,
# node_type,
# 'NODE[COARSE]TYPE'
# )
if isinstance(direct_offset, StringArrayOffset):
for string_const in direct_offset.strings:
self.add_edge(direct_offset, string_const, 'STRARR[]ELEM')
keys = list(sorted(self.binary.direct_offsets.keys()))
for key1, key2 in zip(keys[:-1], keys[1:]):
direct_offset1 = self.binary.direct_offsets[key1]
direct_offset2 = self.binary.direct_offsets[key2]
diff = key2 - key1
if diff <= 0x10:
self.add_edge(direct_offset1, direct_offset2,
'LOCAL{}[DIRECT]'.format(diff))
self.add_edge(direct_offset1.ttype, direct_offset2.ttype,
'LOCAL{}[DIRECT]TTYPE'.format(diff))
self.add_edge(direct_offset1.ttype,
make_size_node(diff, self.binary),
'SIZE[DIRECT]TTYPE')
for function in self.binary.functions.functions:
if function.is_run_init:
self.add_edge(function, function.ttype, 'NAME[{}]TTYPE',
coarse)
node_type = make_node_type(coarse(function), self.binary)
self.add_edge(function, node_type, 'NODE[COARSE]TYPE')
for callee in function.callees:
self.add_edge(function, callee, 'CALL')
for indirect_offset in function.indirect_offsets.values():
for off in indirect_offset.values():
node_type = make_node_type(fine(off), self.binary)
self.add_edge(function, node_type,
'[FUNC]NODETYPE[INDIRECT]')
if not (self.binary.config.MODE == self.binary.config.TRAIN
and not function.init_run):
for i in function.string_consts:
string_const = function.binary.string_consts[i]
self.add_edge(string_const, function, 'STR[]FUNC')
for offset in function.direct_offsets:
direct_offset = function.binary.direct_offsets[offset]
self.add_edge(direct_offset, function, 'DIRECT[]FUNC')
for virtual_exp in function.virtual_exps.values():
EXP_EDGE_EXTRACTOR.visit(virtual_exp.exp,
function=function)
for indirect_offset in function.indirect_offsets.values():
for off in indirect_offset.values():
# self.add_edge(
# off,
# function,
# 'INDIRECT[{}]FUNC',
# coarse
# )
self.add_edge(off, function, 'INDIRECT[{}]FUNC', fine)
# self.add_edge(
# off,
# off.ttype,
# 'NAME[{}]TTYPE',
# coarse
# )
self.add_edge(off, off.ttype, 'NAME[{}]TTYPE', fine)
# node_type = make_node_type(coarse(off), self.binary)
# self.add_edge(
# off,
# node_type,
# 'NODE[COARSE]TYPE'
# )
self.add_edge(off, node_type, 'NODE[FINE]TYPE')
# node_type = make_node_type(fine(off.ttype), self.binary)
# self.add_edge(
# off.ttype,
# node_type,
# 'NODE[FINE]TYPE'
# )
for reg in function.regs.values():
self.add_edge(reg, function, 'REG[{}]FUNC', coarse)
if self.binary.config.MACHINE_ARCH in ('x64', 'ARM') \
and reg.var_type == constants.FUN_ARG:
self.add_edge(reg, function, 'REG[{}]FUNC', fine)
self.add_edge(reg, reg.ttype, 'NAME[{}]TTYPE', coarse)
node_type = make_node_type(coarse(reg), self.binary)
self.add_edge(reg, node_type, 'NODE[COARSE]TYPE')
# node_type = make_node_type(coarse(reg.ttype), self.binary)
# self.add_edge(
# reg.ttype,
# node_type,
# 'NODE[FINE]TYPE'
# )
for key in function.indirect_offsets:
for i in range(1,
self.binary.config.ADDRESS_BYTE_SIZE + 1):
key_1 = (key[0], key[1] + i)
if key_1 in function.indirect_offsets:
indirect_offsets = function.indirect_offsets[key]
indirect_offsets_1 = function.indirect_offsets[
key_1]
for index in indirect_offsets:
if index in indirect_offsets_1:
indirect_offset = indirect_offsets[index]
indirect_offset_1 = indirect_offsets_1[
index]
# self.add_edge(
# indirect_offset,
# indirect_offset_1,
# 'LOCAL[{}]',
# coarse
# )
self.add_edge(indirect_offset,
indirect_offset_1,
'[{}]LOCAL[{}]', fine, fine)
for key in function.regs:
key_1 = (key[0], key[1] + 1)
if key_1 in function.regs:
reg = function.regs[key]
reg_1 = function.regs[key_1]
self.add_edge(reg, reg_1, 'LOCAL[{}]', coarse)
for function in self.binary.functions.functions:
if not (self.binary.config.MODE == self.binary.config.TRAIN
and not function.init_run):
for blk in function.blks.values():
for stmt in blk.stmts:
STMT_EDGE_EXTRACTOR.visit(stmt, function=function)
def to_json(self, clear=False):
nodes = []
for node in self.nodes.values():
if type(node) == IntConst:
node_name = 'INT[{}][{}]'.format(node.width, node.value)
node_type = 'giv'
elif type(node) == StringConst:
node_name = '\"{}\"'.format(node.value)
node_type = 'giv'
elif type(node) == SwitchTable:
node_name = 'SwitchTable'
node_type = 'giv'
elif type(node) == Flag:
node_name = node.base_flag
node_type = 'giv'
elif type(node) == Insn:
node_name = node.name
node_type = 'giv'
elif type(node) == CodeOffset:
node_name = 'CodeOffset'
node_type = 'giv'
elif type(node) == VirtualElm:
node_name = node.name
node_type = 'giv'
elif type(node) == GivOffset:
node_name = str(node.offset)
node_type = 'giv'
elif type(node) == TempOffset:
node_name = fine(node)
node_type = 'giv'
elif type(node) == GivReg:
node_name = coarse(node)
node_type = 'giv'
elif type(node) == NodeType:
node_name = node.name
node_type = 'giv'
elif type(node) == OpNode:
node_name = node.name
node_type = 'giv'
elif type(node) == OtherVarNode:
node_name = node.name
node_type = 'giv'
elif type(node) == SizeNode:
node_name = 'SIZE[{}]'.format(node.size)
node_type = 'giv'
elif type(node) == UnknownNode:
node_name = 'Unknown'
node_type = 'giv'
elif type(node) == Function:
if node.is_name_given:
node_name = node.name
node_type = 'giv'
else:
node_name = node.train_name
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node) == DirectOffset:
if node.is_name_given:
node_name = node.name
node_type = 'giv'
else:
node_name = node.train_name
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node) == StringArrayOffset:
if node.is_name_given:
node_name = node.name
node_type = 'giv'
else:
node_name = node.train_name
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node) == IndirectOffset:
node_name = node.train_name
if node.n2p_type == self.binary.config.GIV:
node_type = 'giv'
else:
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node) == Reg:
node_name = node.train_name
if node.n2p_type == self.binary.config.GIV:
node_type = 'giv'
else:
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node) == Ttype:
node_name = node.train_name
if type(node.owner) == Reg:
if node.owner.n2p_type == self.binary.config.GIV:
node_type = 'giv'
else:
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
elif type(node.owner) == IndirectOffset:
if node.owner.n2p_type == self.binary.config.GIV:
node_type = 'giv'
else:
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
else:
if self.binary.config.UNK_GIV:
node_type = 'giv' if node.train_name == UNKNOWN_LABEL else 'inf'
else:
node_type = 'inf'
else:
print(node)
if clear and node_type == 'inf':
nodes.append(OrderedDict([('v', node.id), (node_type, '')]))
else:
nodes.append(
OrderedDict([('v', node.id), (node_type, node_name)]))
return nodes
def add_unary_feature(feature, node):
if isinstance(node, Reg):
node.features.add(feature.format(coarse(node)))
elif isinstance(node, IndirectOffset):
node.features.add(feature.format(coarse(node)))
node.features.add(feature.format(fine(node)))