2017-10-28 19:00:18 +02:00
|
|
|
#!/usr/bin/env python3
|
|
|
|
|
2017-10-29 11:52:59 +01:00
|
|
|
"""
|
2019-03-31 20:51:34 +02:00
|
|
|
Example analyzing script for saved exports (by main.py, as JSON).
|
|
|
|
This file belongs to https://github.com/bitkeks/python-netflow-v9-softflowd.
|
2017-10-29 11:52:59 +01:00
|
|
|
|
2020-01-20 16:56:41 +01:00
|
|
|
Copyright 2017-2020 Dominik Pataky <dev@bitkeks.eu>
|
2017-10-29 11:52:59 +01:00
|
|
|
Licensed under MIT License. See LICENSE.
|
|
|
|
"""
|
2017-10-28 19:00:18 +02:00
|
|
|
|
2019-10-05 09:06:06 +02:00
|
|
|
import argparse
|
|
|
|
from collections import namedtuple
|
|
|
|
import contextlib
|
2017-10-28 19:00:18 +02:00
|
|
|
from datetime import datetime
|
2019-10-05 09:06:06 +02:00
|
|
|
import functools
|
2019-11-03 11:59:28 +01:00
|
|
|
import gzip
|
2017-10-28 19:00:18 +02:00
|
|
|
import ipaddress
|
|
|
|
import json
|
2019-11-03 11:59:28 +01:00
|
|
|
import logging
|
|
|
|
import os.path
|
2017-10-29 11:52:59 +01:00
|
|
|
import socket
|
2019-10-05 09:06:06 +02:00
|
|
|
import sys
|
2018-02-20 12:09:10 +01:00
|
|
|
|
|
|
|
|
2019-10-05 09:06:06 +02:00
|
|
|
Pair = namedtuple('Pair', ['src', 'dest'])
|
2019-11-03 11:59:28 +01:00
|
|
|
logger = logging.getLogger(__name__)
|
2017-10-29 11:52:59 +01:00
|
|
|
|
|
|
|
|
2019-11-03 11:59:28 +01:00
|
|
|
@functools.lru_cache(maxsize=None)
|
2019-10-05 09:06:06 +02:00
|
|
|
def resolve_hostname(ip):
|
|
|
|
return socket.getfqdn(ip)
|
2017-10-29 11:52:59 +01:00
|
|
|
|
2019-10-17 05:23:51 +02:00
|
|
|
|
|
|
|
def fallback(d, keys):
|
|
|
|
for k in keys:
|
2019-10-31 17:55:48 +01:00
|
|
|
if k in d:
|
2019-10-17 05:23:51 +02:00
|
|
|
return d[k]
|
|
|
|
raise KeyError(", ".join(keys))
|
|
|
|
|
|
|
|
|
2017-10-29 11:52:59 +01:00
|
|
|
class Connection:
|
|
|
|
"""Connection model for two flows.
|
|
|
|
The direction of the data flow can be seen by looking at the size.
|
|
|
|
|
|
|
|
'src' describes the peer which sends more data towards the other. This
|
2019-10-05 09:06:06 +02:00
|
|
|
does NOT have to mean that 'src' was the initiator of the connection.
|
2017-10-29 11:52:59 +01:00
|
|
|
"""
|
|
|
|
def __init__(self, flow1, flow2):
|
2019-10-05 09:06:06 +02:00
|
|
|
if not flow1 or not flow2:
|
|
|
|
raise Exception("A connection requires two flows")
|
|
|
|
|
2019-10-17 05:23:51 +02:00
|
|
|
# Assume the size that sent the most data is the source
|
2019-10-31 17:35:15 +01:00
|
|
|
# TODO: this might not always be right, maybe use earlier timestamp?
|
2019-10-17 05:23:51 +02:00
|
|
|
size1 = fallback(flow1, ['IN_BYTES', 'IN_OCTETS'])
|
|
|
|
size2 = fallback(flow2, ['IN_BYTES', 'IN_OCTETS'])
|
|
|
|
if size1 >= size2:
|
2017-10-29 11:52:59 +01:00
|
|
|
src = flow1
|
|
|
|
dest = flow2
|
|
|
|
else:
|
|
|
|
src = flow2
|
|
|
|
dest = flow1
|
|
|
|
|
2019-11-03 15:58:40 +01:00
|
|
|
# TODO: this next approach uses the lower port as the service identifier
|
|
|
|
# port1 = fallback(flow1, ['L4_SRC_PORT', 'SRC_PORT'])
|
|
|
|
# port2 = fallback(flow2, ['L4_SRC_PORT', 'SRC_PORT'])
|
|
|
|
#
|
|
|
|
# src = flow1
|
|
|
|
# dest = flow2
|
|
|
|
# if port1 > port2:
|
|
|
|
# src = flow2
|
|
|
|
# dest = flow1
|
|
|
|
|
|
|
|
self.src_flow = src
|
|
|
|
self.dest_flow = dest
|
2019-10-05 09:06:06 +02:00
|
|
|
ips = self.get_ips(src)
|
2017-10-29 11:52:59 +01:00
|
|
|
self.src = ips.src
|
|
|
|
self.dest = ips.dest
|
2019-10-17 05:23:51 +02:00
|
|
|
self.src_port = fallback(src, ['L4_SRC_PORT', 'SRC_PORT'])
|
2019-10-31 17:35:15 +01:00
|
|
|
self.dest_port = fallback(dest, ['L4_DST_PORT', 'DST_PORT'])
|
2019-10-17 05:23:51 +02:00
|
|
|
self.size = fallback(src, ['IN_BYTES', 'IN_OCTETS'])
|
2017-10-29 11:52:59 +01:00
|
|
|
|
2017-10-29 19:38:33 +01:00
|
|
|
# Duration is given in milliseconds
|
|
|
|
self.duration = src['LAST_SWITCHED'] - src['FIRST_SWITCHED']
|
|
|
|
if self.duration < 0:
|
|
|
|
# 32 bit int has its limits. Handling overflow here
|
2019-10-05 09:06:06 +02:00
|
|
|
# TODO: Should be handled in the collection phase
|
2017-10-29 19:38:33 +01:00
|
|
|
self.duration = (2**32 - src['FIRST_SWITCHED']) + src['LAST_SWITCHED']
|
|
|
|
|
2017-10-29 11:52:59 +01:00
|
|
|
def __repr__(self):
|
|
|
|
return "<Connection from {} to {}, size {}>".format(
|
|
|
|
self.src, self.dest, self.human_size)
|
|
|
|
|
2019-10-05 09:06:06 +02:00
|
|
|
@staticmethod
|
|
|
|
def get_ips(flow):
|
|
|
|
# IPv4
|
2019-11-03 13:30:50 +01:00
|
|
|
if flow.get('IP_PROTOCOL_VERSION') == 4 or \
|
|
|
|
'IPV4_SRC_ADDR' in flow or 'IPV4_DST_ADDR' in flow:
|
2019-10-05 09:06:06 +02:00
|
|
|
return Pair(
|
|
|
|
ipaddress.ip_address(flow['IPV4_SRC_ADDR']),
|
|
|
|
ipaddress.ip_address(flow['IPV4_DST_ADDR'])
|
|
|
|
)
|
|
|
|
|
|
|
|
# IPv6
|
|
|
|
return Pair(
|
|
|
|
ipaddress.ip_address(flow['IPV6_SRC_ADDR']),
|
|
|
|
ipaddress.ip_address(flow['IPV6_DST_ADDR'])
|
|
|
|
)
|
|
|
|
|
2017-10-29 11:52:59 +01:00
|
|
|
@property
|
|
|
|
def human_size(self):
|
|
|
|
# Calculate a human readable size of the traffic
|
|
|
|
if self.size < 1024:
|
2017-10-29 19:38:33 +01:00
|
|
|
return "%dB" % self.size
|
2017-10-29 11:52:59 +01:00
|
|
|
elif self.size / 1024. < 1024:
|
|
|
|
return "%.2fK" % (self.size / 1024.)
|
|
|
|
elif self.size / 1024.**2 < 1024:
|
|
|
|
return "%.2fM" % (self.size / 1024.**2)
|
|
|
|
else:
|
|
|
|
return "%.2fG" % (self.size / 1024.**3)
|
|
|
|
|
2017-10-29 19:38:33 +01:00
|
|
|
@property
|
|
|
|
def human_duration(self):
|
|
|
|
duration = self.duration // 1000 # uptime in milliseconds, floor it
|
|
|
|
if duration < 60:
|
|
|
|
# seconds
|
|
|
|
return "%d sec" % duration
|
|
|
|
if duration / 60 > 60:
|
|
|
|
# hours
|
|
|
|
return "%d:%02d.%02d hours" % (duration / 60**2, duration % 60**2 / 60, duration % 60)
|
|
|
|
# minutes
|
|
|
|
return "%02d:%02d min" % (duration / 60, duration % 60)
|
|
|
|
|
2017-10-29 11:52:59 +01:00
|
|
|
@property
|
|
|
|
def hostnames(self):
|
|
|
|
# Resolve the IPs of this flows to their hostname
|
2019-10-05 09:06:06 +02:00
|
|
|
src_hostname = resolve_hostname(self.src.compressed)
|
|
|
|
dest_hostname = resolve_hostname(self.dest.compressed)
|
2017-10-29 19:38:33 +01:00
|
|
|
return Pair(src_hostname, dest_hostname)
|
2017-10-29 11:52:59 +01:00
|
|
|
|
|
|
|
@property
|
|
|
|
def service(self):
|
|
|
|
# Resolve ports to their services, if known
|
2019-11-03 15:58:40 +01:00
|
|
|
default = "({} {})".format(self.src_port, self.dest_port)
|
|
|
|
if self.src_port > 10000:
|
|
|
|
return default
|
2019-10-05 09:06:06 +02:00
|
|
|
with contextlib.suppress(OSError):
|
|
|
|
return socket.getservbyport(self.src_port)
|
|
|
|
with contextlib.suppress(OSError):
|
|
|
|
return socket.getservbyport(self.dest_port)
|
2019-11-03 15:58:40 +01:00
|
|
|
return default
|
|
|
|
|
|
|
|
@property
|
|
|
|
def total_packets(self):
|
|
|
|
return self.src_flow["IN_PKTS"] + self.dest_flow["IN_PKTS"]
|
2019-10-05 09:06:06 +02:00
|
|
|
|
|
|
|
|
|
|
|
if __name__ == "__main__":
|
|
|
|
parser = argparse.ArgumentParser(description="Output a basic analysis of NetFlow data")
|
2019-11-03 11:59:28 +01:00
|
|
|
parser.add_argument('-f', '--file', dest='file', type=str, default=sys.stdin,
|
2019-10-05 09:06:06 +02:00
|
|
|
help="The file to analyze (defaults to stdin if not provided)")
|
2019-11-03 15:58:40 +01:00
|
|
|
parser.add_argument('-p', '--packets', dest='packets_threshold', type=int, default=10,
|
|
|
|
help="Number of packets representing the lower bound in connections to be processed")
|
2019-10-05 09:06:06 +02:00
|
|
|
args = parser.parse_args()
|
|
|
|
|
2019-11-03 11:59:28 +01:00
|
|
|
# Using a file and using stdin differ in their further usage for gzip.open
|
|
|
|
file = args.file
|
|
|
|
mode = "rb" # reading files
|
|
|
|
if file != sys.stdin and not os.path.exists(file):
|
|
|
|
exit("File {} does not exist!".format(file))
|
|
|
|
|
|
|
|
if file == sys.stdin:
|
|
|
|
file = sys.stdin.buffer
|
|
|
|
mode = "rt" # reading from stdin
|
|
|
|
|
|
|
|
data = {}
|
|
|
|
|
|
|
|
with gzip.open(file, mode) as gzipped:
|
|
|
|
# "for line in" lazy-loads all lines in the file
|
|
|
|
for line in gzipped:
|
|
|
|
entry = json.loads(line)
|
|
|
|
if len(entry.keys()) != 1:
|
|
|
|
logger.warning("Line \"{}\" does not have exactly one timestamp key.")
|
|
|
|
|
|
|
|
try:
|
|
|
|
ts = list(entry)[0] # timestamp from key
|
|
|
|
except KeyError:
|
|
|
|
logger.error("Saved line \"{}\" has no timestamp key!".format(line))
|
|
|
|
continue
|
|
|
|
|
|
|
|
data[ts] = entry[ts]
|
2019-10-05 09:06:06 +02:00
|
|
|
|
2019-11-03 11:59:28 +01:00
|
|
|
# Go through data and dissect every flow saved inside the dump
|
2019-11-03 15:58:40 +01:00
|
|
|
|
|
|
|
# The following dict holds flows which are looking for a peer, to analyze a duplex 'Connection'.
|
|
|
|
# For each flow, the destination address is looked up. If the peer is not in the list of pending peers,
|
|
|
|
# insert this flow, waiting for its peer. If found, take the waiting peer and create a Connection object.
|
|
|
|
pending = {}
|
|
|
|
skipped = 0
|
|
|
|
skipped_threshold = args.packets_threshold
|
|
|
|
|
2019-10-05 09:06:06 +02:00
|
|
|
for key in sorted(data):
|
|
|
|
timestamp = datetime.fromtimestamp(float(key)).strftime("%Y-%m-%d %H:%M.%S")
|
2019-11-03 13:57:06 +01:00
|
|
|
client = data[key]["client"]
|
|
|
|
flows = data[key]["flows"]
|
2019-11-03 15:58:40 +01:00
|
|
|
|
|
|
|
for flow in sorted(flows, key=lambda x: x["FIRST_SWITCHED"]):
|
|
|
|
first_switched = flow["FIRST_SWITCHED"]
|
|
|
|
|
|
|
|
if first_switched - 1 in pending:
|
|
|
|
# TODO: handle fitting, yet mismatching (here: 1 second) pairs
|
|
|
|
pass
|
|
|
|
|
|
|
|
if first_switched not in pending:
|
|
|
|
pending[first_switched] = {}
|
|
|
|
|
|
|
|
# Find the peer for this connection
|
|
|
|
if flow["IP_PROTOCOL_VERSION"] == 4:
|
|
|
|
local_peer = flow["IPV4_SRC_ADDR"]
|
|
|
|
remote_peer = flow["IPV4_DST_ADDR"]
|
|
|
|
else:
|
|
|
|
local_peer = flow["IPV6_SRC_ADDR"]
|
|
|
|
remote_peer = flow["IPV6_DST_ADDR"]
|
|
|
|
|
|
|
|
if remote_peer in pending[first_switched]:
|
|
|
|
# The destination peer put itself into the pending dict, getting and removing entry
|
|
|
|
peer_flow = pending[first_switched].pop(remote_peer)
|
|
|
|
if len(pending[first_switched]) == 0:
|
|
|
|
del pending[first_switched]
|
|
|
|
else:
|
|
|
|
# Flow did not find a matching, pending peer - inserting itself
|
|
|
|
pending[first_switched][local_peer] = flow
|
|
|
|
continue
|
|
|
|
|
|
|
|
con = Connection(flow, peer_flow)
|
|
|
|
if con.total_packets < skipped_threshold:
|
|
|
|
skipped += 1
|
2019-10-05 09:06:06 +02:00
|
|
|
continue
|
2019-11-03 15:58:40 +01:00
|
|
|
|
|
|
|
print("{timestamp}: {service:<14} | {size:8} | {duration:9} | {packets:5} | Between {src_host} ({src}) and {dest_host} ({dest})" \
|
|
|
|
.format(timestamp=timestamp, service=con.service.upper(), src_host=con.hostnames.src, src=con.src,
|
|
|
|
dest_host=con.hostnames.dest, dest=con.dest, size=con.human_size, duration=con.human_duration,
|
|
|
|
packets=con.total_packets))
|
|
|
|
|
|
|
|
if skipped > 0:
|
|
|
|
print(f"{skipped} connections skipped, because they had less than {skipped_threshold} packets.")
|
|
|
|
|
|
|
|
if len(pending) > 0:
|
|
|
|
print(f"There are {len(pending)} first_switched entries left in the pending dict!")
|
|
|
|
all_noise = True
|
|
|
|
for first_switched, flows in sorted(pending.items(), key=lambda x: x[0]):
|
|
|
|
for peer, flow in flows.items():
|
|
|
|
# Ignore all pings, SYN scans and other noise to find only those peers left over which need a fix
|
|
|
|
if flow["IN_PKTS"] < skipped_threshold:
|
|
|
|
continue
|
|
|
|
all_noise = False
|
|
|
|
|
|
|
|
if flow["IP_PROTOCOL_VERSION"] == 4:
|
|
|
|
print(first_switched, peer, flow["IPV4_DST_ADDR"], flow["IN_PKTS"])
|
|
|
|
else:
|
|
|
|
print(first_switched, peer, flow["IPV6_DST_ADDR"], flow["IN_PKTS"])
|
|
|
|
|
|
|
|
if all_noise:
|
|
|
|
print("They were all noise!")
|