blob: b16838d567b4ff64c0eda202c062aa720f697dd6 [file] [log] [blame]
# Copyright 2014-2015, Tresys Technology, LLC
#
# This file is part of SETools.
#
# SETools is free software: you can redistribute it and/or modify
# it under the terms of the GNU Lesser General Public License as
# published by the Free Software Foundation, either version 2.1 of
# the License, or (at your option) any later version.
#
# SETools is distributed in the hope that it will be useful,
# but WITHOUT ANY WARRANTY; without even the implied warranty of
# MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the
# GNU Lesser General Public License for more details.
#
# You should have received a copy of the GNU Lesser General Public
# License along with SETools. If not, see
# <http://www.gnu.org/licenses/>.
#
# pylint: disable=unsubscriptable-object
import itertools
import logging
from collections import defaultdict, namedtuple
import networkx as nx
from networkx.exception import NetworkXError, NetworkXNoPath
from .descriptors import EdgeAttrDict, EdgeAttrList
__all__ = ['DomainTransitionAnalysis']
# Return values for the analysis
# are in the following tuple formats:
step_output = namedtuple("step", ["source",
"target",
"transition",
"entrypoints",
"setexec",
"dyntransition",
"setcurrent"])
entrypoint_output = namedtuple("entrypoints", ["name",
"entrypoint",
"execute",
"type_transition"])
class DomainTransitionAnalysis(object):
"""Domain transition analysis."""
def __init__(self, policy, reverse=False, exclude=None):
"""
Parameter:
policy The policy to analyze.
"""
self.log = logging.getLogger(__name__)
self.policy = policy
self.exclude = exclude
self.reverse = reverse
self.rebuildgraph = True
self.rebuildsubgraph = True
self.G = nx.DiGraph()
self.subG = None
@property
def reverse(self):
return self._reverse
@reverse.setter
def reverse(self, direction):
self._reverse = bool(direction)
self.rebuildsubgraph = True
@property
def exclude(self):
return self._exclude
@exclude.setter
def exclude(self, types):
if types:
self._exclude = [self.policy.lookup_type(t) for t in types]
else:
self._exclude = []
self.rebuildsubgraph = True
def shortest_path(self, source, target):
"""
Generator which yields one shortest domain transition path
between the source and target types (there may be more).
Parameters:
source The source type.
target The target type.
Yield: generator(steps)
steps A generator that returns the tuple of
source, target, and rules for each
domain transition.
"""
s = self.policy.lookup_type(source)
t = self.policy.lookup_type(target)
if self.rebuildsubgraph:
self._build_subgraph()
self.log.info("Generating one domain transition path from {0} to {1}...".format(s, t))
try:
yield self.__generate_steps(nx.shortest_path(self.subG, s, t))
except (NetworkXNoPath, NetworkXError):
# NetworkXError: the type is valid but not in graph, e.g. excluded
# NetworkXNoPath: no paths or the target type is
# not in the graph
pass
def all_paths(self, source, target, maxlen=2):
"""
Generator which yields all domain transition paths between
the source and target up to the specified maximum path
length.
Parameters:
source The source type.
target The target type.
maxlen Maximum length of paths.
Yield: generator(steps)
steps A generator that returns the tuple of
source, target, and rules for each
domain transition.
"""
if maxlen < 1:
raise ValueError("Maximum path length must be positive.")
s = self.policy.lookup_type(source)
t = self.policy.lookup_type(target)
if self.rebuildsubgraph:
self._build_subgraph()
self.log.info("Generating all domain transition paths from {0} to {1}, max length {2}...".
format(s, t, maxlen))
try:
for path in nx.all_simple_paths(self.subG, s, t, maxlen):
yield self.__generate_steps(path)
except (NetworkXNoPath, NetworkXError):
# NetworkXError: the type is valid but not in graph, e.g. excluded
# NetworkXNoPath: no paths or the target type is
# not in the graph
pass
def all_shortest_paths(self, source, target):
"""
Generator which yields all shortest domain transition paths
between the source and target types.
Parameters:
source The source type.
target The target type.
Yield: generator(steps)
steps A generator that returns the tuple of
source, target, and rules for each
domain transition.
"""
s = self.policy.lookup_type(source)
t = self.policy.lookup_type(target)
if self.rebuildsubgraph:
self._build_subgraph()
self.log.info("Generating all shortest domain transition paths from {0} to {1}...".
format(s, t))
try:
for path in nx.all_shortest_paths(self.subG, s, t):
yield self.__generate_steps(path)
except (NetworkXNoPath, NetworkXError, KeyError):
# NetworkXError: the type is valid but not in graph, e.g. excluded
# NetworkXNoPath: no paths or the target type is
# not in the graph
# KeyError: work around NetworkX bug
# when the source node is not in the graph
pass
def transitions(self, type_):
"""
Generator which yields all domain transitions out of a
specified source type.
Parameters:
type_ The starting type.
Yield: generator(steps)
steps A generator that returns the tuple of
source, target, and rules for each
domain transition.
"""
s = self.policy.lookup_type(type_)
if self.rebuildsubgraph:
self._build_subgraph()
self.log.info("Generating all domain transitions {1} {0}".
format(s, "in to" if self.reverse else "out from"))
try:
for source, target in self.subG.out_edges_iter(s):
edge = Edge(self.subG, source, target)
if self.reverse:
real_source, real_target = target, source
else:
real_source, real_target = source, target
yield step_output(real_source, real_target,
edge.transition,
self.__generate_entrypoints(edge),
edge.setexec,
edge.dyntransition,
edge.setcurrent)
except NetworkXError:
# NetworkXError: the type is valid but not in graph, e.g. excluded
pass
def get_stats(self): # pragma: no cover
"""
Get the domain transition graph statistics.
Return: str
"""
if self.rebuildgraph:
self._build_graph()
return nx.info(self.G)
#
# Internal functions follow
#
@staticmethod
def __generate_entrypoints(edge):
"""
Creates a list of entrypoint, execute, and
type_transition rules for each entrypoint.
Parameter:
data The dictionary of entrypoints.
Return: list of tuple(type, entry, exec, trans)
type The entrypoint type.
entry The list of entrypoint rules.
exec The list of execute rules.
trans The list of type_transition rules.
"""
return [entrypoint_output(e, edge.entrypoint[e], edge.execute[e], edge.type_transition[e])
for e in edge.entrypoint]
def __generate_steps(self, path):
"""
Generator which yields the source, target, and associated rules
for each domain transition.
Parameter:
path A list of graph node names representing an information flow path.
Yield: tuple(source, target, transition, entrypoints,
setexec, dyntransition, setcurrent)
source The source type for this step of the domain transition.
target The target type for this step of the domain transition.
transition The list of transition rules.
entrypoints Generator which yields entrypoint-related rules.
setexec The list of setexec rules.
dyntranstion The list of dynamic transition rules.
setcurrent The list of setcurrent rules.
"""
for s in range(1, len(path)):
source = path[s - 1]
target = path[s]
edge = Edge(self.subG, source, target)
# Yield the actual source and target.
# The above perspective is reversed
# if the graph has been reversed.
if self.reverse:
real_source, real_target = target, source
else:
real_source, real_target = source, target
yield step_output(real_source, real_target,
edge.transition,
self.__generate_entrypoints(edge),
edge.setexec,
edge.dyntransition,
edge.setcurrent)
#
# Graph building functions
#
# Domain transition requirements:
#
# Standard transitions a->b:
# allow a b:process transition;
# allow a b_exec:file execute;
# allow b b_exec:file entrypoint;
#
# and at least one of:
# allow a self:process setexec;
# type_transition a b_exec:process b;
#
# Dynamic transition x->y:
# allow x y:process dyntransition;
# allow x self:process setcurrent;
#
# Algorithm summary:
# 1. iterate over all rules
# 1. skip non allow/type_transition rules
# 2. if process transition or dyntransition, create edge,
# initialize rule lists, add the (dyn)transition rule
# 3. if process setexec or setcurrent, add to appropriate dict
# keyed on the subject
# 4. if file exec, entrypoint, or type_transition:process,
# add to appropriate dict keyed on subject,object.
# 2. Iterate over all graph edges:
# 1. if there is a transition rule (else add to invalid
# transition list):
# 1. use set intersection to find matching exec
# and entrypoint rules. If none, add to invalid
# transition list.
# 2. for each valid entrypoint, add rules to the
# edge's lists if there is either a
# type_transition for it or the source process
# has setexec permissions.
# 3. If there are neither type_transitions nor
# setexec permissions, add to the invalid
# transition list
# 2. if there is a dyntransition rule (else add to invalid
# dyntrans list):
# 1. If the source has a setcurrent rule, add it
# to the edge's list, else add to invalid
# dyntransition list.
# 3. Iterate over all graph edges:
# 1. if the edge has an invalid trans and dyntrans, delete
# the edge.
# 2. if the edge has an invalid trans, clear the related
# lists on the edge.
# 3. if the edge has an invalid dyntrans, clear the related
# lists on the edge.
#
def _build_graph(self):
self.G.clear()
self.G.name = "Domain transition graph for {0}.".format(self.policy)
self.log.info("Building domain transition graph from {0}...".format(self.policy))
# hash tables keyed on domain type
setexec = defaultdict(list)
setcurrent = defaultdict(list)
# hash tables keyed on (domain, entrypoint file type)
# the parameter for defaultdict has to be callable
# hence the lambda for the nested defaultdict
execute = defaultdict(lambda: defaultdict(list))
entrypoint = defaultdict(lambda: defaultdict(list))
# hash table keyed on (domain, entrypoint, target domain)
type_trans = defaultdict(lambda: defaultdict(lambda: defaultdict(list)))
for rule in self.policy.terules():
if rule.ruletype == "allow":
if rule.tclass not in ["process", "file"]:
continue
perms = rule.perms
if rule.tclass == "process":
if "transition" in perms:
for s, t in itertools.product(rule.source.expand(), rule.target.expand()):
# only add edges if they actually
# transition to a new type
if s != t:
edge = Edge(self.G, s, t, create=True)
edge.transition.append(rule)
if "dyntransition" in perms:
for s, t in itertools.product(rule.source.expand(), rule.target.expand()):
# only add edges if they actually
# transition to a new type
if s != t:
e = Edge(self.G, s, t, create=True)
e.dyntransition.append(rule)
if "setexec" in perms:
for s in rule.source.expand():
setexec[s].append(rule)
if "setcurrent" in perms:
for s in rule.source.expand():
setcurrent[s].append(rule)
else:
if "execute" in perms:
for s, t in itertools.product(
rule.source.expand(),
rule.target.expand()):
execute[s][t].append(rule)
if "entrypoint" in perms:
for s, t in itertools.product(rule.source.expand(), rule.target.expand()):
entrypoint[s][t].append(rule)
elif rule.ruletype == "type_transition":
if rule.tclass != "process":
continue
d = rule.default
for s, t in itertools.product(rule.source.expand(), rule.target.expand()):
type_trans[s][t][d].append(rule)
invalid_edge = []
clear_transition = []
clear_dyntransition = []
for s, t in self.G.edges_iter():
edge = Edge(self.G, s, t)
invalid_trans = False
invalid_dyntrans = False
if edge.transition:
# get matching domain exec w/entrypoint type
entry = set(entrypoint[t].keys())
exe = set(execute[s].keys())
match = entry.intersection(exe)
if not match:
# there are no valid entrypoints
invalid_trans = True
else:
# TODO try to improve the
# efficiency in this loop
for m in match:
if s in setexec or type_trans[s][m]:
# add key for each entrypoint
edge.entrypoint[m] += entrypoint[t][m]
edge.execute[m] += execute[s][m]
if type_trans[s][m][t]:
edge.type_transition[m] += type_trans[s][m][t]
if s in setexec:
edge.setexec.extend(setexec[s])
if not edge.setexec and not edge.type_transition:
invalid_trans = True
else:
invalid_trans = True
if edge.dyntransition:
if s in setcurrent:
edge.setcurrent.extend(setcurrent[s])
else:
invalid_dyntrans = True
else:
invalid_dyntrans = True
# cannot change the edges while iterating over them,
# so keep appropriate lists
if invalid_trans and invalid_dyntrans:
invalid_edge.append(edge)
elif invalid_trans:
clear_transition.append(edge)
elif invalid_dyntrans:
clear_dyntransition.append(edge)
# Remove invalid transitions
self.G.remove_edges_from(invalid_edge)
for edge in clear_transition:
# if only the regular transition is invalid,
# clear the relevant lists
del edge.transition
del edge.execute
del edge.entrypoint
del edge.type_transition
del edge.setexec
for edge in clear_dyntransition:
# if only the dynamic transition is invalid,
# clear the relevant lists
del edge.dyntransition
del edge.setcurrent
self.rebuildgraph = False
self.rebuildsubgraph = True
self.log.info("Completed building domain transition graph.")
self.log.debug("Graph stats: nodes: {0}, edges: {1}.".format(
nx.number_of_nodes(self.G),
nx.number_of_edges(self.G)))
def __remove_excluded_entrypoints(self):
invalid_edges = []
for source, target in self.subG.edges_iter():
edge = Edge(self.subG, source, target)
entrypoints = set(edge.entrypoint)
entrypoints.intersection_update(self.exclude)
if not entrypoints:
# short circuit if there are no
# excluded entrypoint types on
# this edge.
continue
for e in entrypoints:
# clear the entrypoint data
del edge.entrypoint[e]
del edge.execute[e]
try:
del edge.type_transition[e]
except KeyError: # setexec
pass
# cannot delete the edges while iterating over them
if not edge.entrypoint and not edge.dyntransition:
invalid_edges.append(edge)
self.subG.remove_edges_from(invalid_edges)
def _build_subgraph(self):
if self.rebuildgraph:
self._build_graph()
self.log.info("Building domain transition subgraph.")
self.log.debug("Excluding {0}".format(self.exclude))
self.log.debug("Reverse {0}".format(self.reverse))
# reverse graph for reverse DTA
if self.reverse:
self.subG = self.G.reverse(copy=True)
else:
self.subG = self.G.copy()
if self.exclude:
# delete excluded domains from subgraph
self.subG.remove_nodes_from(self.exclude)
# delete excluded entrypoints from subgraph
self.__remove_excluded_entrypoints()
self.rebuildsubgraph = False
self.log.info("Completed building domain transition subgraph.")
self.log.debug("Subgraph stats: nodes: {0}, edges: {1}.".format(
nx.number_of_nodes(self.subG),
nx.number_of_edges(self.subG)))
class Edge(object):
"""
A graph edge. Also used for returning domain transition steps.
Parameters:
graph The NetworkX graph.
source The source type of the edge.
target The target tyep of the edge.
Keyword Parameters:
create (T/F) create the edge if it does not exist.
The default is False.
"""
transition = EdgeAttrList('transition')
setexec = EdgeAttrList('setexec')
dyntransition = EdgeAttrList('dyntransition')
setcurrent = EdgeAttrList('setcurrent')
entrypoint = EdgeAttrDict('entrypoint')
execute = EdgeAttrDict('execute')
type_transition = EdgeAttrDict('type_transition')
def __init__(self, graph, source, target, create=False):
self.G = graph
self.source = source
self.target = target
if not self.G.has_edge(source, target):
if not create:
raise ValueError("Edge does not exist in graph")
else:
self.G.add_edge(source, target)
self.transition = None
self.entrypoint = None
self.execute = None
self.type_transition = None
self.setexec = None
self.dyntransition = None
self.setcurrent = None
def __getitem__(self, key):
# This is implemented so this object can be used in NetworkX
# functions that operate on (source, target) tuples
if isinstance(key, slice):
return [self._index_to_item(i) for i in range(* key.indices(2))]
else:
return self._index_to_item(key)
def _index_to_item(self, index):
"""Return source or target based on index."""
if index == 0:
return self.source
elif index == 1:
return self.target
else:
raise IndexError("Invalid index (edges only have 2 items): {0}".format(index))