A Botnet Detecting Infrastructure
Using a Beneficial Botnet
Takashi Yamanoue, Fukuyama University
ACM SIGUCCS 2018, @Orland Fl. USA, Oct. 9, 2018.
#siguccs18
#siguccs18
• An Eye for an Eye
• A Tooth for a Tooth
• A Bot for a Bot
• A Botnet for a Botnet
#siguccs18
Vs.
Vs.
Vs.
Vs.
Contents
• 1. Introduction
• 2. Beneficial Botnet
• 3. Experiment
• 4. Related Work
• 5. Conclusion
#siguccs18
1. Introduction (1/6)
• Network Managers and
People in charge of Network Security
…Troubled by Malicious Botnet
#siguccs18
• A Malicious Botnet does
– Spread SPAM mails
– DDoS Attack
– Click Fraud
– Steal bank account information of users
of zombie computers.
• A Malicious Botnet is
– Persistent
• Continues malicious things even if some
of bots of it were removed.
– Evolving
1. Introduction (2/6)
#siguccs18
• Technology of a Malicious Botnet
– It was rely on a single centralized command and control
(C2) server
• The botnet can be disrupted by removing the C2 server
–Peer To Peer (P2P)
• Middle years of 2000 …ex. Agobot/Phatbot
–Domain Generation Algorithm, (DGA)
• Late years of 2000 … ex. Conficker
• Our approach to cope with them
–Beneficial Botnet
1. Introduction (3/6)
#siguccs18
• Our Beneficial botnet is A group of beneficial bots
–Agent bots and an analyzing bot.
• An agent bot
– Located between a LAN and its NAT/ router.
– Collects and controls communication of hosts in the
LAN
• The analyzing bot
– Collects communication data from agent bots
– Analyzes the data.
– Can execute R programs for the analysis
1. Introduction (4/6)
#siguccs18
Agent bot
NAT/Router
Sub-LAN
• The P2P communication of malicious
botnet
– hard to detect by a single IDS at the entrance of
organizational network
• Because
– No single C2 Server
– Small amount of traffic of P2P communication
between P2P nodes at the inside of the
organizational network and the outside.
• Our beneficial botnet
– has the ability to detect P2P communication
using collaboration of our beneficial bots.
1. Introduction (5/6)
#siguccs18
• We have Made a Pseudo Botnet, Pseudo
Gameover ZeuS, for evaluation.
– Performs P2P communication between some nodes
of it without doing malicious things.
• Our beneficial botnet
– could detect P2P communication of the pseudo
Gameover ZeuS
– may also detect communication using DGA.
• Furthermore, our beneficial botnet
– has ability to cope with new technology of new
botnets, because our beneficial botnet has the
ability to evolving,
as same as malicious botnets.
1. Introduction (6/6)
#siguccs18
2. Beneficial Botnet(1/13)
• It is common
– to use an IDS or intrusion prevention system
(IPS)
– at the entrance of an organizational network
now.
• An IDS or IPS at the entrance of the
organizational network
– is effective for a malicious botnet with
centralized C2 server because every
communication between the C2 server and
all bots can be detected by the IDS or the IPS.
#siguccs18
• An Example of Botnets
– Gameover ZeuS
• Disrupted in 2014 by the international collaborative
investigating activity.
• The losses attributable to Gameover ZeuS were
estimated over one hundred million dollars according
to the FBI announcement
• “The second centralized version of Zeus mutated into
a peer-to-peer (P2P) variant, known as P2P Zeus or
Gameover. Since P2P Zeus does not rely on
centralized command and control server (C2), it is
immune to traditional countermeasures against Zeus”,
according to Andriesse and Bos
2. Beneficial Botnet (2/13)
#siguccs18
• Botnets with the P2P feature are
– difficult to detect, difficult to disrupt.
– No Centralized command and control server
(C2 server)
• A reason of great losses of the Gameover
Zeus can be considered that the Gameover
Zeus acquired the P2P feature.
– Gameover ZeuS has been disrupted.
However, there are chances that similar P2P
botnets intrude campuses and conduct their
malicious operation.
2. Beneficial Botnet(3/13)
#siguccs18
• By the single IDS at the network
entrance of the campus
–Hard to detect P2P
communication of bots in the
campus.
–Hard to locate such Bots in the
campus.
2. Beneficial Botnet(4/13)
#siguccs18
• We have designed our beneficial botnet to cope
with malicious botnet such as the Gameover
ZeuS.
2. Beneficial Botnet(5/13)
#siguccs18
• We want cope with technologies of Malicious
Botnets
– Developing The Beneficial Botnet using our
Beneficial Bots
• “Monitoring Servers With a Little From My Bots”
• “Capturing Malicious Bots using a Beneficial Bot”
– A Beneficial Botnet is a group of Beneficial Bots
• Agent Bots behind the NAT of a LAN
• Analyzing Bots which analyzes data which collected by
Agent Bots
– The Beneficial Botnet has ability to detect P2P
communication of malicious Bots, using
collaboration of Beneficial Bots.
2. Beneficial Botnet(6/13)
#siguccs18
Agent bot
NAT/Router
2. Beneficial Botnet(7/13)
#siguccs18
2. Beneficial Botnet(8/13)
2.1 A Bot of a Beneficial Botnet(1/2)
#siguccs18
It is a script interpreter
#siguccs18
Sensors, Actuators,
Traffic Controller
Ex. Of Other command
- set pageName <page-name>
- include <url>
An example of a wiki page
2. Beneficial Botnet(9/13)
2.1 A Bot of a Beneficial Botnet(2/2)
2. Beneficial Botnet(10/13)
2.1 Agent Bot (1/3)
#siguccs18
Agent bot
NAT/Router
Sub-LAN
• Buffers
– Packet History
• Sub buffers for every (Source IP, Destination
IP)
• Packet Information + Date/Time, Sha1 hash of
the Packet Payload
– MAC-list …
• which hosts are connected to this sub-LAN
and its IP addresses in this LAN.
2. Beneficial Botnet(11/13)
2.1 Agent Bot (2/3)
#siguccs18
– Domain-list … DNS queries
• which host in this sub-LAN communicated with
which host outside of this LAN.
• Can be used to detect the usage of Domain
Generation Algorithm (DGA).
– Dhcp-list … DHCP server queries
• Can be used to
–detect the DHCP spoofing
–detect un-authorized DHCP server.
– Arp-list
• Can be used to
–detect the Arp spoofing.
#siguccs18
2. Beneficial Botnet(12/13)
2.1 Agent Bot (3/3)
• The analyzing bot gathers information of each
agent bot and analyzes them.
• The language processor of Beneficial Bot
– CSV parser
– Spread Sheet Manipulation/Spread sheet functions.
• Analyzing Bot,
– The language processor of Beneficial Bot +
– R language processor
2. Beneficial Botnet(13/13)
2.2 Analyzing Bot (1/1)
#siguccs18
3. Experiment(1/18)
• Gameover
ZeuS
#siguccs18
We can download
the source code
However,
It is dangerous.
How shall we evaluate the beneficial botnet?
3. Experiment(2/18)
• Pseudo
Gameover
ZeuS
• Does Not
Do Bad
Things
#siguccs18
#siguccs18
3. Experiment(3/18)
#siguccs18
3. Experiment(4/18)
3.1 Script and results of Agent Bots(1/5)
3. Experiment(5/18)
3.1 Script and results of Agent Bots(2/5)
• Class
page
#siguccs18
UDP,
Not NTP,
Not DNS
Host-list
Domain-list
#siguccs18
3. Experiment(6/18)
3.1 Script and results of Agent Bots(3/5)
• Object page
#siguccs18
3. Experiment(7/18)
3.1 Script and results of Agent Bots(4/5)
Results
• A Part of the results
– cmd=get repeating, date="2018/04/14 17:03:19 +0900", no=3299,
if=1, smac="bc:5c:4c:5d:1c:cd", dmac="b8:27:eb:cb:d6:38",
prtcl=udp, sip="192.168.13.160", dip="192.168.2.100", sp=34724,
dp=33331,
sha1payload="9dac7a7beb944a7193847a3d0fbcc370d13a5838",
payloadLength=46, payload=broadcast id=3394824 ttl=3
cmd="message test".....
3. Experiment(8/18)
3.1 Script and results of Agent Bots(5/5)
#siguccs18
#siguccs18
3. Experiment(9/18)
3.2 Script and results of Analyzing Bots(1/10)
#siguccs18
3. Experiment(10/18)
3.2 Script and results of Analyzing Bots(2/10)
Find Out pairs of Packets
Satisfying
- different LAN
- same SHA1
- Near Time
Among
- All pair of Packets
(from Agent Bots)
3. Experiment(11/18)
3.2 Script and results of Analyzing Bots(3/10)
Define Arrays,
Prepare URLs of Object Pages of Agent Bots
…
3. Experiment(12/18)
3.2 Script and results of Analyzing Bots(4/10)
Choose Packet Info,
Read CSV into the Table,
Get Vectors of
date, sip, dip,
smac, dmac,
payload, sha1,
LAN-ID
For R processor
For each LAN,
3. Experiment(13/18)
3.2 Script and results of Analyzing Bots(5/10)
Prepare the Data Frame
of the Packet Information
In this LAN
For R
and combine the Data Frame
Of All LAN
3. Experiment(14/18)
3.2 Script and results of Analyzing Bots(6/10)
Find Out pairs of Packets
Satisfying
- different LAN
- same SHA1
- Near Time
Among
- All pair of Packets
(from Agent Bots)
Write the Results to the Wiki.
Results
• A part of the Object page of the Analyzing Bot
– Possible P2P communication Between LAN-1 and LAN-2
• lan1= 0 ,date= 2018/04/14 17:06:50 +0900 ,
smac= b8:27:eb:cb:d6:38 ,dmac= bc:5c:4c:5d:1c:cd ,
sip= 192.168.2.100 ,dip= 192.168.13.160 ,
lan2= 1 ,date= 2018/04/14 17:06:51 +0900 ,
smac= bc:5c:4c:5d:1a:c9 ,dmac= b8:27:eb:2f:33:cd ,
sip= 192.168.13.210 ,dip= 192.168.2.102 ,
sha1payload=
f14db4dae7a139cde5185267b8d353498850f22b ,
payload= broadcast id
3. Experiment(15/18)
3.2 Script and results of Analyzing Bots(7/10)
3. Experiment(16/18)
3.2 Script and results of Analyzing Bots(8/10)
#siguccs18
• A part of the Object page of the Analyzing Bot
– Possible P2P communication Between LAN-2 and LAN-5
• lan1= 1 ,date= 2018/04/14 17:03:18 +0900 ,
smac= b8:27:eb:2f:33:cd ,dmac= bc:5c:4c:5d:1a:c9 ,
sip= 192.168.2.102 ,dip= 192.168.13.150 ,
lan2= 4 ,date= 2018/04/14 17:03:17 +0900 ,
smac= bc:5c:4c:5d:1a:bf ,dmac= b8:27:eb:3a:6b:fa ,
sip= 192.168.13.160 ,dip= 192.168.2.102 ,
sha1payload=
9dac7a7beb944a7193847a3d0fbcc370d13a5838 ,
payload= broadcast id
3. Experiment(17/18)
3.2 Script and results of Analyzing Bots(9/10)
3. Experiment(18/18)
3.2 Script and results of Analyzing Bots(10/10)
#siguccs18
• A group of Agents of Distributed IDS.
• agents + transceivers + monitors
• An agent of the AAFID is similar to our agent bot
– both of them are controlled by commands and
collect traffic data.
• A monitor and a transceiver of the AAFID is similar
to our analyzing bot
– both of them are collecting data from agents,
transceivers, other monitors or agent bots, and
analyzing the data.
4. Related Work(1/5)
4.1 Autonomous Agents for Intrusion Detection (AAFID)
#siguccs18
(Purdue University)
• An agent of the AAFID is installed in a client host
– while our agent bot is placed between LAN and
its router or NAT router.
• The manager of our beneficial botnet
– does not need to install our agent bot to each client
host.
4. Related Work(2/5)
4.1 Autonomous Agents for Intrusion Detection (AAFID)
#siguccs18
(Purdue University)
• A monitor or a transceiver of the AAFID
– is not controlled by the script in a wiki page
– while our agent bots And the analyzing bot is
controlled by the script in a wiki page.
• Communication mechanism is not specified in
the AAFID architecture
– while our beneficial botnet uses wiki API.
#siguccs18
4. Related Work(3/5)
4.1 Autonomous Agents for Intrusion Detection (AAFID)
(Purdue University)
• The action of an agent bot of our
beneficial bot
– can be seen as the man-in-the-middle
attack.
• Many communications in the sub-
LAN
– can be controlled by the agent bot.
• We have to be careful
– so that the agent bot does not to go
to the dark side.
4. Related Work(4/5)
4.2 Man in the Middle Attack
#siguccs18
• Consist of
– agent programs at the devices, such as
PCs(KASEYA) or wi-fi access points(UNIFAS),
– a web site to manage them,
– as in our beneficial botnet.
• Their devices
– can also communicate with the web site over a
NAT as in our agent bots.
• However, they
– use a specialized web server,
– whereas our beneficial botnet uses a web site
with common wiki software.
4. Related Work(5/5)
4.3 KASEYA and UNIFAS
#siguccs18
• Get Flow Information of Network,
Find out similar communication
between hosts, discriminate
communication which seems
between malicious bots.
• Similar to our beneficial botnet
– Finding out similar communication
between hosts.
• Different to our beneficial botnet
– It uses flow information of network.
4. Related Work(6/6)
BotMiner
#siguccs18
• Beneficial Botnet
– P2P communication by malware can be detected
– The use of DGA algorithm can be also detected.
– Ability to cope with new technology of new botnets by
rewriting the scripts.
– Currently, very slow.
• have to improve the speed for real use of our beneficial
botnet.
– We also have to improve the security of our beneficial
botnet.
5. Conclusion
#siguccs18
Agent bot
NAT/Router
Sub-LAN
Acknowledgements
• A part of this research was supported by JSPS KAKENHI
Grant Number JP16K00197.
• Pcap4J, community of R.
• Students who helped us to conduct the experiment in this
paper.
#siguccs18
• An Eye for an Eye
• A Tooth for a Tooth
• A Bot for a Bot
• A Botnet for a Botnet
#siguccs18
Vs.
Vs.
Vs.
Vs.
Beneficial Botnet
#siguccs18

A Botnet Detecting Infrastructure Using a Beneficial Botnet

  • 1.
    A Botnet DetectingInfrastructure Using a Beneficial Botnet Takashi Yamanoue, Fukuyama University ACM SIGUCCS 2018, @Orland Fl. USA, Oct. 9, 2018.
  • 2.
  • 3.
    • An Eyefor an Eye • A Tooth for a Tooth • A Bot for a Bot • A Botnet for a Botnet #siguccs18 Vs. Vs. Vs. Vs.
  • 4.
    Contents • 1. Introduction •2. Beneficial Botnet • 3. Experiment • 4. Related Work • 5. Conclusion #siguccs18
  • 5.
    1. Introduction (1/6) •Network Managers and People in charge of Network Security …Troubled by Malicious Botnet #siguccs18
  • 6.
    • A MaliciousBotnet does – Spread SPAM mails – DDoS Attack – Click Fraud – Steal bank account information of users of zombie computers. • A Malicious Botnet is – Persistent • Continues malicious things even if some of bots of it were removed. – Evolving 1. Introduction (2/6) #siguccs18
  • 7.
    • Technology ofa Malicious Botnet – It was rely on a single centralized command and control (C2) server • The botnet can be disrupted by removing the C2 server –Peer To Peer (P2P) • Middle years of 2000 …ex. Agobot/Phatbot –Domain Generation Algorithm, (DGA) • Late years of 2000 … ex. Conficker • Our approach to cope with them –Beneficial Botnet 1. Introduction (3/6) #siguccs18
  • 8.
    • Our Beneficialbotnet is A group of beneficial bots –Agent bots and an analyzing bot. • An agent bot – Located between a LAN and its NAT/ router. – Collects and controls communication of hosts in the LAN • The analyzing bot – Collects communication data from agent bots – Analyzes the data. – Can execute R programs for the analysis 1. Introduction (4/6) #siguccs18 Agent bot NAT/Router Sub-LAN
  • 9.
    • The P2Pcommunication of malicious botnet – hard to detect by a single IDS at the entrance of organizational network • Because – No single C2 Server – Small amount of traffic of P2P communication between P2P nodes at the inside of the organizational network and the outside. • Our beneficial botnet – has the ability to detect P2P communication using collaboration of our beneficial bots. 1. Introduction (5/6) #siguccs18
  • 10.
    • We haveMade a Pseudo Botnet, Pseudo Gameover ZeuS, for evaluation. – Performs P2P communication between some nodes of it without doing malicious things. • Our beneficial botnet – could detect P2P communication of the pseudo Gameover ZeuS – may also detect communication using DGA. • Furthermore, our beneficial botnet – has ability to cope with new technology of new botnets, because our beneficial botnet has the ability to evolving, as same as malicious botnets. 1. Introduction (6/6) #siguccs18
  • 11.
    2. Beneficial Botnet(1/13) •It is common – to use an IDS or intrusion prevention system (IPS) – at the entrance of an organizational network now. • An IDS or IPS at the entrance of the organizational network – is effective for a malicious botnet with centralized C2 server because every communication between the C2 server and all bots can be detected by the IDS or the IPS. #siguccs18
  • 12.
    • An Exampleof Botnets – Gameover ZeuS • Disrupted in 2014 by the international collaborative investigating activity. • The losses attributable to Gameover ZeuS were estimated over one hundred million dollars according to the FBI announcement • “The second centralized version of Zeus mutated into a peer-to-peer (P2P) variant, known as P2P Zeus or Gameover. Since P2P Zeus does not rely on centralized command and control server (C2), it is immune to traditional countermeasures against Zeus”, according to Andriesse and Bos 2. Beneficial Botnet (2/13) #siguccs18
  • 13.
    • Botnets withthe P2P feature are – difficult to detect, difficult to disrupt. – No Centralized command and control server (C2 server) • A reason of great losses of the Gameover Zeus can be considered that the Gameover Zeus acquired the P2P feature. – Gameover ZeuS has been disrupted. However, there are chances that similar P2P botnets intrude campuses and conduct their malicious operation. 2. Beneficial Botnet(3/13) #siguccs18
  • 14.
    • By thesingle IDS at the network entrance of the campus –Hard to detect P2P communication of bots in the campus. –Hard to locate such Bots in the campus. 2. Beneficial Botnet(4/13) #siguccs18
  • 15.
    • We havedesigned our beneficial botnet to cope with malicious botnet such as the Gameover ZeuS. 2. Beneficial Botnet(5/13) #siguccs18
  • 16.
    • We wantcope with technologies of Malicious Botnets – Developing The Beneficial Botnet using our Beneficial Bots • “Monitoring Servers With a Little From My Bots” • “Capturing Malicious Bots using a Beneficial Bot” – A Beneficial Botnet is a group of Beneficial Bots • Agent Bots behind the NAT of a LAN • Analyzing Bots which analyzes data which collected by Agent Bots – The Beneficial Botnet has ability to detect P2P communication of malicious Bots, using collaboration of Beneficial Bots. 2. Beneficial Botnet(6/13) #siguccs18 Agent bot NAT/Router
  • 17.
  • 18.
    2. Beneficial Botnet(8/13) 2.1A Bot of a Beneficial Botnet(1/2) #siguccs18 It is a script interpreter #siguccs18 Sensors, Actuators, Traffic Controller
  • 19.
    Ex. Of Othercommand - set pageName <page-name> - include <url> An example of a wiki page 2. Beneficial Botnet(9/13) 2.1 A Bot of a Beneficial Botnet(2/2)
  • 20.
    2. Beneficial Botnet(10/13) 2.1Agent Bot (1/3) #siguccs18 Agent bot NAT/Router Sub-LAN
  • 21.
    • Buffers – PacketHistory • Sub buffers for every (Source IP, Destination IP) • Packet Information + Date/Time, Sha1 hash of the Packet Payload – MAC-list … • which hosts are connected to this sub-LAN and its IP addresses in this LAN. 2. Beneficial Botnet(11/13) 2.1 Agent Bot (2/3) #siguccs18
  • 22.
    – Domain-list …DNS queries • which host in this sub-LAN communicated with which host outside of this LAN. • Can be used to detect the usage of Domain Generation Algorithm (DGA). – Dhcp-list … DHCP server queries • Can be used to –detect the DHCP spoofing –detect un-authorized DHCP server. – Arp-list • Can be used to –detect the Arp spoofing. #siguccs18 2. Beneficial Botnet(12/13) 2.1 Agent Bot (3/3)
  • 23.
    • The analyzingbot gathers information of each agent bot and analyzes them. • The language processor of Beneficial Bot – CSV parser – Spread Sheet Manipulation/Spread sheet functions. • Analyzing Bot, – The language processor of Beneficial Bot + – R language processor 2. Beneficial Botnet(13/13) 2.2 Analyzing Bot (1/1) #siguccs18
  • 24.
    3. Experiment(1/18) • Gameover ZeuS #siguccs18 Wecan download the source code However, It is dangerous. How shall we evaluate the beneficial botnet?
  • 25.
    3. Experiment(2/18) • Pseudo Gameover ZeuS •Does Not Do Bad Things #siguccs18
  • 26.
  • 27.
    #siguccs18 3. Experiment(4/18) 3.1 Scriptand results of Agent Bots(1/5)
  • 28.
    3. Experiment(5/18) 3.1 Scriptand results of Agent Bots(2/5) • Class page #siguccs18 UDP, Not NTP, Not DNS Host-list Domain-list
  • 29.
    #siguccs18 3. Experiment(6/18) 3.1 Scriptand results of Agent Bots(3/5)
  • 30.
    • Object page #siguccs18 3.Experiment(7/18) 3.1 Script and results of Agent Bots(4/5) Results
  • 31.
    • A Partof the results – cmd=get repeating, date="2018/04/14 17:03:19 +0900", no=3299, if=1, smac="bc:5c:4c:5d:1c:cd", dmac="b8:27:eb:cb:d6:38", prtcl=udp, sip="192.168.13.160", dip="192.168.2.100", sp=34724, dp=33331, sha1payload="9dac7a7beb944a7193847a3d0fbcc370d13a5838", payloadLength=46, payload=broadcast id=3394824 ttl=3 cmd="message test"..... 3. Experiment(8/18) 3.1 Script and results of Agent Bots(5/5) #siguccs18
  • 32.
    #siguccs18 3. Experiment(9/18) 3.2 Scriptand results of Analyzing Bots(1/10)
  • 33.
    #siguccs18 3. Experiment(10/18) 3.2 Scriptand results of Analyzing Bots(2/10) Find Out pairs of Packets Satisfying - different LAN - same SHA1 - Near Time Among - All pair of Packets (from Agent Bots)
  • 34.
    3. Experiment(11/18) 3.2 Scriptand results of Analyzing Bots(3/10) Define Arrays, Prepare URLs of Object Pages of Agent Bots …
  • 35.
    3. Experiment(12/18) 3.2 Scriptand results of Analyzing Bots(4/10) Choose Packet Info, Read CSV into the Table, Get Vectors of date, sip, dip, smac, dmac, payload, sha1, LAN-ID For R processor For each LAN,
  • 36.
    3. Experiment(13/18) 3.2 Scriptand results of Analyzing Bots(5/10) Prepare the Data Frame of the Packet Information In this LAN For R and combine the Data Frame Of All LAN
  • 37.
    3. Experiment(14/18) 3.2 Scriptand results of Analyzing Bots(6/10) Find Out pairs of Packets Satisfying - different LAN - same SHA1 - Near Time Among - All pair of Packets (from Agent Bots) Write the Results to the Wiki. Results
  • 38.
    • A partof the Object page of the Analyzing Bot – Possible P2P communication Between LAN-1 and LAN-2 • lan1= 0 ,date= 2018/04/14 17:06:50 +0900 , smac= b8:27:eb:cb:d6:38 ,dmac= bc:5c:4c:5d:1c:cd , sip= 192.168.2.100 ,dip= 192.168.13.160 , lan2= 1 ,date= 2018/04/14 17:06:51 +0900 , smac= bc:5c:4c:5d:1a:c9 ,dmac= b8:27:eb:2f:33:cd , sip= 192.168.13.210 ,dip= 192.168.2.102 , sha1payload= f14db4dae7a139cde5185267b8d353498850f22b , payload= broadcast id 3. Experiment(15/18) 3.2 Script and results of Analyzing Bots(7/10)
  • 39.
    3. Experiment(16/18) 3.2 Scriptand results of Analyzing Bots(8/10) #siguccs18
  • 40.
    • A partof the Object page of the Analyzing Bot – Possible P2P communication Between LAN-2 and LAN-5 • lan1= 1 ,date= 2018/04/14 17:03:18 +0900 , smac= b8:27:eb:2f:33:cd ,dmac= bc:5c:4c:5d:1a:c9 , sip= 192.168.2.102 ,dip= 192.168.13.150 , lan2= 4 ,date= 2018/04/14 17:03:17 +0900 , smac= bc:5c:4c:5d:1a:bf ,dmac= b8:27:eb:3a:6b:fa , sip= 192.168.13.160 ,dip= 192.168.2.102 , sha1payload= 9dac7a7beb944a7193847a3d0fbcc370d13a5838 , payload= broadcast id 3. Experiment(17/18) 3.2 Script and results of Analyzing Bots(9/10)
  • 41.
    3. Experiment(18/18) 3.2 Scriptand results of Analyzing Bots(10/10) #siguccs18
  • 42.
    • A groupof Agents of Distributed IDS. • agents + transceivers + monitors • An agent of the AAFID is similar to our agent bot – both of them are controlled by commands and collect traffic data. • A monitor and a transceiver of the AAFID is similar to our analyzing bot – both of them are collecting data from agents, transceivers, other monitors or agent bots, and analyzing the data. 4. Related Work(1/5) 4.1 Autonomous Agents for Intrusion Detection (AAFID) #siguccs18 (Purdue University)
  • 43.
    • An agentof the AAFID is installed in a client host – while our agent bot is placed between LAN and its router or NAT router. • The manager of our beneficial botnet – does not need to install our agent bot to each client host. 4. Related Work(2/5) 4.1 Autonomous Agents for Intrusion Detection (AAFID) #siguccs18 (Purdue University)
  • 44.
    • A monitoror a transceiver of the AAFID – is not controlled by the script in a wiki page – while our agent bots And the analyzing bot is controlled by the script in a wiki page. • Communication mechanism is not specified in the AAFID architecture – while our beneficial botnet uses wiki API. #siguccs18 4. Related Work(3/5) 4.1 Autonomous Agents for Intrusion Detection (AAFID) (Purdue University)
  • 45.
    • The actionof an agent bot of our beneficial bot – can be seen as the man-in-the-middle attack. • Many communications in the sub- LAN – can be controlled by the agent bot. • We have to be careful – so that the agent bot does not to go to the dark side. 4. Related Work(4/5) 4.2 Man in the Middle Attack #siguccs18
  • 46.
    • Consist of –agent programs at the devices, such as PCs(KASEYA) or wi-fi access points(UNIFAS), – a web site to manage them, – as in our beneficial botnet. • Their devices – can also communicate with the web site over a NAT as in our agent bots. • However, they – use a specialized web server, – whereas our beneficial botnet uses a web site with common wiki software. 4. Related Work(5/5) 4.3 KASEYA and UNIFAS #siguccs18
  • 47.
    • Get FlowInformation of Network, Find out similar communication between hosts, discriminate communication which seems between malicious bots. • Similar to our beneficial botnet – Finding out similar communication between hosts. • Different to our beneficial botnet – It uses flow information of network. 4. Related Work(6/6) BotMiner #siguccs18
  • 48.
    • Beneficial Botnet –P2P communication by malware can be detected – The use of DGA algorithm can be also detected. – Ability to cope with new technology of new botnets by rewriting the scripts. – Currently, very slow. • have to improve the speed for real use of our beneficial botnet. – We also have to improve the security of our beneficial botnet. 5. Conclusion #siguccs18 Agent bot NAT/Router Sub-LAN
  • 49.
    Acknowledgements • A partof this research was supported by JSPS KAKENHI Grant Number JP16K00197. • Pcap4J, community of R. • Students who helped us to conduct the experiment in this paper. #siguccs18
  • 50.
    • An Eyefor an Eye • A Tooth for a Tooth • A Bot for a Bot • A Botnet for a Botnet #siguccs18 Vs. Vs. Vs. Vs.
  • 51.