cloudtrail.py

"""Functions to get CloudTrail records from disk"""
import datetime
import gzip
import json
import logging
import os
import re

import pytz
from toolz import pipe
from toolz.curried import filter as filterz

from trailscraper.boto_service_definitions import operation_definition
from trailscraper.iam import Statement, Action

ALL_RECORDS_FILTERED = "No records matching your criteria found! Did you use the right filters? " \
                       "Did you download the right logfiles? "\
                       "It might take about 15 minutes for events to turn up in CloudTrail logs."


class Record:
    """Represents a CloudTrail record"""

    # pylint: disable=too-many-arguments
    def __init__(self, event_source, event_name,
                 resource_arns=None, assumed_role_arn=None, event_time=None, raw_source=None):
        self.event_source = event_source
        self.event_name = event_name
        self.raw_source = raw_source
        self.event_time = event_time
        self.resource_arns = resource_arns or ["*"]
        self.assumed_role_arn = assumed_role_arn

    def __repr__(self):
        return f"Record(event_source={self.event_source} event_name={self.event_name} " \
               f"event_time={self.event_time} resource_arns={self.resource_arns})"

    def __eq__(self, other):
        if isinstance(other, self.__class__):
            return self.event_source == other.event_source and \
                   self.event_name == other.event_name and \
                   self.event_time == other.event_time and \
                   self.resource_arns == other.resource_arns and \
                   self.assumed_role_arn == other.assumed_role_arn

        return False

    def __hash__(self):
        return hash((self.event_source,
                     self.event_name,
                     self.event_time,
                     tuple(self.resource_arns),
                     self.assumed_role_arn))

    def __ne__(self, other):
        return not self.__eq__(other)

    def _source_to_iam_prefix(self):
        special_cases = {
            'monitoring.amazonaws.com': 'cloudwatch',
            'appstream2.amazonaws.com': 'appstream',
            'models.lex.amazonaws.com': 'lex',
            'runtime.lex.amazonaws.com': 'lex',
            'mturk-requester.amazonaws.com': 'mechanicalturk',
            'streams.dynamodb.amazonaws.com': 'dynamodb',
            'tagging.amazonaws.com': 'tag',
        }

        default_case = self.event_source.split('.')[0]

        return special_cases.get(self.event_source, default_case)

    def _event_name_to_iam_action(self):
        special_cases = {
            's3.amazonaws.com': {
                'CompleteMultipartUpload': 'PutObject',
                'CopyObject': 'PutObject',
                'CreateMultipartUpload': 'PutObject',
                'DeleteBucketAnalyticsConfiguration': 'PutAnalyticsConfiguration',
                'DeleteBucketEncryption': 'PutEncryptionConfiguration',
                'DeleteBucketInventoryConfiguration': 'PutInventoryConfiguration',
                'DeleteBucketLifecycle': 'PutLifecycleConfiguration',
                'DeleteBucketMetricsConfiguration': 'PutMetricsConfiguration',
                'DeleteBucketReplication': 'DeleteReplicationConfiguration',
                'DeleteBucketTagging': 'PutBucketTagging',
                'DeleteObjects': 'DeleteObject',
                'GetBucketAccelerateConfiguration': 'GetAccelerateConfiguration',
                'GetBucketAnalyticsConfiguration': 'GetAnalyticsConfiguration',
                'GetBucketEncryption': 'GetEncryptionConfiguration',
                'GetBucketInventoryConfiguration': 'GetInventoryConfiguration',
                'GetBucketLifecycle': 'GetLifecycleConfiguration',
                'GetBucketLifecycleConfiguration': 'GetLifecycleConfiguration',
                'GetBucketMetricsConfiguration': 'GetMetricsConfiguration',
                'GetBucketNotificationConfiguration': 'GetBucketNotification',
                'GetBucketReplication': 'GetReplicationConfiguration',
                'HeadBucket': 'ListBucket',
                'HeadObject': 'GetObject',
                'ListBucketAnalyticsConfigurations': 'GetAnalyticsConfiguration',
                'ListBucketInventoryConfigurations': 'GetInventoryConfiguration',
                'ListBucketMetricsConfigurations': 'GetMetricsConfiguration',
                'ListBuckets': 'ListAllMyBuckets',
                'ListMultipartUploads': 'ListBucketMultipartUploads',
                'ListObjectVersions': 'ListBucketVersions',
                'ListObjects': 'ListBucket',
                'ListObjectsV2': 'ListBucket',
                'ListParts': 'ListMultipartUploadParts',
                'PutBucketAccelerateConfiguration': 'PutAccelerateConfiguration',
                'PutBucketAnalyticsConfiguration': 'PutAnalyticsConfiguration',
                'PutBucketEncryption': 'PutEncryptionConfiguration',
                'PutBucketInventoryConfiguration': 'PutInventoryConfiguration',
                'PutBucketLifecycle': 'PutLifecycleConfiguration',
                'PutBucketLifecycleConfiguration': 'PutLifecycleConfiguration',
                'PutBucketMetricsConfiguration': 'PutMetricsConfiguration',
                'PutBucketNotificationConfiguration': 'PutBucketNotification',
                'PutBucketReplication': 'DeleteReplicationConfiguration',
                'UploadPart': 'PutObject',
                'UploadPartCopy': 'PutObject',
            },
            'kms.amazonaws.com': {
                'ReEncrypt': 'ReEncrypt*'  # not precise. See #27 for more details.
            }
        }

        def _regex_sub(expr, subs):
            regex = re.compile(expr)
            return lambda s: regex.sub(subs, s)

        def _special_case_mappings(event_name):
            return special_cases \
                .get(self.event_source, {}) \
                .get(event_name, event_name)

        return pipe(self.event_name,
                    _special_case_mappings,
                    _regex_sub(r"DeleteBucketCors", "PutBucketCORS"),
                    _regex_sub(r"([a-zA-Z]+)[0-9v_]+$", r"\1", ),
                    _regex_sub(r"Cors$", "CORS"))

    def _to_api_gateway_statement(self):
        op_def = operation_definition("apigateway", self.event_name)

        http_method = op_def['http']['method']
        request_uri = op_def['http']['requestUri']

        resource_path = re.compile(r"{[a-zA-Z_]+}").sub("*", request_uri)

        region = "*"  # use proper region from requestParameters

        return Statement(
            Effect="Allow",
            Action=[Action("apigateway", http_method)],
            Resource=[f"arn:aws:apigateway:{region}::{resource_path}"]
        )

    def to_statement(self):
        """Converts record into a matching IAM Policy Statement"""
        if self.event_source == "sts.amazonaws.com" and self.event_name == "GetCallerIdentity":
            return None

        if self.event_source == "apigateway.amazonaws.com":
            return self._to_api_gateway_statement()

        return Statement(
            Effect="Allow",
            Action=[Action(self._source_to_iam_prefix(), self._event_name_to_iam_action())],
            Resource=sorted(self.resource_arns)
        )


class LogFile:
    """Represents a single CloudTrail Log File"""

    def __init__(self, path):
        self._path = path

    def timestamp(self):
        """Returns the timestamp the log file was delivered"""
        dstr = self.filename().split('_')[3]
        # using manual substring instead of strptime for performance reasons
        # inspired by https://stackoverflow.com/a/14166888
        return datetime.datetime(*map(int, [dstr[:4], dstr[4:6], dstr[6:8], dstr[9:11], dstr[11:13]])) \
            .replace(tzinfo=pytz.utc)

    def filename(self):
        """Name of the logfile (without path)"""
        return os.path.split(self._path)[-1]

    def has_valid_filename(self):
        """Returns if the log file represented has a valid filename"""
        pattern = re.compile(r"[0-9]+_CloudTrail_[a-z0-9-]+_[0-9TZ]+_[a-zA-Z0-9]+\.json\.gz")
        return pattern.match(self.filename())

    def records(self):
        """Returns CloudTrail Records in this log file"""
        logging.debug("Loading %s", self._path)

        try:
            with gzip.open(self._path, 'rt') as unzipped:
                json_data = json.load(unzipped)
                records = json_data['Records']
                return parse_records(records)
        except (IOError, OSError) as error:
            logging.warning("Could not load %s: %s", self._path, error)
            return []

    def contains_events_for_timeframe(self, from_date, to_date):
        """Returns true if this logfile likely contains events in the relevant timeframe"""
        return from_date <= self.timestamp() <= to_date + datetime.timedelta(hours=1)


def _resource_arns(json_record):
    resources = json_record.get('resources', [])
    arns = [resource['ARN'] for resource in resources if 'ARN' in resource]
    return arns


def _assumed_role_arn(json_record):
    user_identity = json_record['userIdentity']
    if 'type' in user_identity \
            and user_identity['type'] == 'AssumedRole' \
            and 'sessionContext' in user_identity:
        return user_identity['sessionContext']['sessionIssuer']['arn']
    return None


def _parse_record(json_record):
    try:
        return Record(json_record['eventSource'],
                      json_record['eventName'],
                      event_time=datetime.datetime.strptime(json_record['eventTime'],
                                                            "%Y-%m-%dT%H:%M:%SZ").replace(tzinfo=pytz.utc),
                      resource_arns=_resource_arns(json_record),
                      assumed_role_arn=_assumed_role_arn(json_record),
                      raw_source=json_record)
    except KeyError as error:
        logging.warning("Could not parse %s: %s", json_record, error)
        return None


def parse_records(json_records):
    """Convert JSON Records into Record objects"""
    parsed_records = [_parse_record(record) for record in json_records]
    return [r for r in parsed_records if r is not None]

def _by_timeframe(from_date, to_date):
    return lambda record: record.event_time is None or \
                          (from_date <= record.event_time <= to_date)


def _by_role_arns(arns_to_filter_for):
    if arns_to_filter_for is None:
        arns_to_filter_for = []

    return lambda record: (record.assumed_role_arn in arns_to_filter_for) or (len(arns_to_filter_for) == 0)


def filter_records(records,
                   arns_to_filter_for=None,
                   from_date=datetime.datetime(1970, 1, 1, tzinfo=pytz.utc),
                   to_date=datetime.datetime.now(tz=pytz.utc)):
    """Filter records so they match the given condition"""
    result = list(pipe(records, filterz(_by_timeframe(from_date, to_date)), filterz(_by_role_arns(arns_to_filter_for))))
    if not result and records:
        logging.warning(ALL_RECORDS_FILTERED)

    return result