NAVAL POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA
THESIS
ENERGY-GRID THREAT ANALYSIS USING HONEYPOTS
by
Marian M. Kendrick and
Zaki A. Rucker
June 2019
Thesis Advisor: Neil C. Rowe
Co-Advisor: Thuy D. Nguyen
Approved for public release. Distribution is
unlimited.
ENERGY-GRID THREAT ANALYSIS USING HONEYPOTS
Marian M. Kendrick
Lieutenant, United States Navy
BS, Ohio University, 2007
Zaki A. Rucker
Lieutenant, United States Navy
BA, The Citadel, 2012
Submitted in partial fulfillment of the
requirements for the degree of
MASTER OF SCIENCE IN COMPUTER SCIENCE
rom the
NAVAL POSTGRADUATE SCHOOL
June 2019
Approved by:���� Neil C. Rowe
�������������������������� Advisor
�������������������������� Thuy D. Nguyen
�������������������������� Co-Advisor
�������������������������� Peter J. Denning
�������������������������� Chair, Department of Computer Science
ABSTRACT
����������� This research addresses cyber threats to smart energy grids.
This research identified and characterized threats to electric-power grids by
analysis of traffic in a simulated grid. We deployed a high-interaction
honeypot that simulates a grid named GridPot. Network-traffic captures and
honeypot-activity logs were analyzed to demonstrate the effectiveness of our
honeypot at collecting intelligence for threat analysis. This study will
contribute to the efforts of the Department of Defense and Department of
Homeland Security to protect U.S. critical infrastructure from cyber threats.
TABLE OF CONTENTS
I.������������ Introduction................................................................................................. 1
A.���������� motivation.......................................................................................... 1
B.���������� Research Plan................................................................................... 3
C.���������� thesis OUtline................................................................................... 3
II.���������� BACKGROUND and
related work........................................................ 5
A.���������� honeypots............................................................................................ 5
B.���������� Previous work.................................................................................. 5
1.����������� The Honeynet Project................................................................... 5
2.����������� Honeypot-Aware Botnets.............................................................. 7
3.����������� Privacy Concerns........................................................................... 8
4.����������� Cloud-Based Industrial Honeypot............................................... 8
5.����������� HoneyPhy....................................................................................... 9
III.�������� GRIDPOT............................................................................................................ 11
A.���������� bulk power system...................................................................... 11
B.���������� gridpot architecture............................................................... 13
1.����������� Conpot.......................................................................................... 13
2.����������� GridLAB-D.................................................................................. 14
3.����������� GridPot......................................................................................... 15
C.���������� Network protocols.................................................................... 16
1.����������� HTTP............................................................................................ 16
2.����������� MODBUS...................................................................................... 17
3.����������� S7COMM..................................................................................... 18
4.����������� SNMP............................................................................................ 18
5.����������� IEC 61850..................................................................................... 18
D.���������� PREVIOUS GRIDPOT AND
CONPOT WORK................................. 19
IV.�������� Methodology.............................................................................................. 21
A.���������� host environment........................................................................ 21
B.���������� Gridpot................................................................................................ 22
1.����������� Honeypot Layer........................................................................... 22
2.����������� Modeling Layer............................................................................ 23
C.���������� DATA COLLECTION............................................................................ 24
1.����������� Internal Testing........................................................................... 25
2.����������� Host-To-Virtual-Machine Baseline............................................ 25
3.����������� Live Data Collection.................................................................... 26
D.���������� PARSER DESIGN................................................................................... 27
E.���������� UNEXPECTED
COMPLICATIONS.................................................... 27
V.���������� Data and results....................................................................................... 29
A.���������� PROTOCOL DATA AND
RESULTS................................................... 29
1.����������� Overall Statistics.......................................................................... 29
2.����������� HTTP............................................................................................ 32
3.����������� MODBUS...................................................................................... 36
4.����������� S7Comm....................................................................................... 39
B.���������� DECEPTION RESULTS........................................................................ 40
C.���������� BENEFITS OF MULTI-TOOL
LOGGING......................................... 40
VI.�������� Conclusion and future
work........................................................... 43
A.���������� conclusions...................................................................................... 43
B.���������� future work..................................................................................... 43
appendix A. GRIDPOT SETUP................................................................................ 45
Appendix B. Running gridpot.......................................................................... 47
APPENDIX C. MODIFIED GPM FILE....................................................................... 49
APPENDIX D. INTERNAL TESTING......................................................................... 51
APPENDIX E. HOST-TO-VIRTUAL MACHINE TESTING................................... 55
List of References................................................................................................ 57
initial distribution list................................................................................... 63
LIST OF FIGURES
Figure 1.������������ A
Honeynet. Source: [16]............................................................................ 6
Figure 2.������������ A
Botnet Layout. Source: [17].................................................................... 7
Figure 3.������������ Botnet
Detection of a Honeypot. Source: [17]............................................ 8
Figure 4.������������ Results
from Cloud-Based Industrial Honeypot Deployment. Source: [19]. 9
Figure 5.������������ North
American Bulk Power System Interconnections. Source: [22]....... 11
Figure 6.������������ Basic
Structure of the Electric System. Source: [9].................................. 12
Figure 7.������������ GridPot
Architecture................................................................................. 16
Figure 8.������������ MODBUS
Communication of Serial Line (left) and Ethernet Using TCP/IP (right). Source:
[38]............................................................................................... 17
Figure 9.������������ NPS
Conpot Data. Source: [48]................................................................. 20
Figure 10.��������� Live
Environment Setup. Adapted from [48]............................................ 22
Figure 11.��������� Launching
Conpot Using the GridPot Template....................................... 23
Figure 12.��������� Partial
Layout of our IEEE 13 Node Grid Model. Source: [46]................ 24
Figure 13.��������� Protocol
Count by Source IP Address from April 11-30, 2019................ 30
Figure 14.��������� Source
IP Address Endpoints.................................................................... 31
Figure 15.��������� Countries
Using More Than 100 Unique IP Source Addresses................ 31
Figure 16.��������� HTTP
Request Method Distribution......................................................... 32
Figure 17.��������� HTTP
Country Count from April 11-30, 2019.......................................... 33
Figure 18.��������� HTTP
Requests from April 11-30, 2019................................................... 34
Figure 19.��������� CDF
of Attacks Using HTTP GET........................................................... 34
Figure 20.��������� CDF
of Attacks Using HTTP POST......................................................... 35
Figure 21.��������� CDF
of Attacks Using HTTP.................................................................... 35
Figure 22.��������� MODBUS
Connections and MODBUS Traffic from April 11-30, 2019. 36
Figure 23.��������� Conpot
Log from Host-to-Virtual-Machine Nmap Scan Using MODBUS Script��� 37
Figure 24.��������� First
Conpot Log of Likely Nmap Scan Using MODBUS Script............. 37
Figure 25.��������� Second
Conpot Log of Likely Nmap Scan Using MODBUS Script........ 38
Figure 26.��������� MODBUS
Traffic Distribution by Country.............................................. 38
Figure 27.��������� S7Comm
Connections from April 11-30, 2019........................................ 39
Figure 28.��������� S7Comm
Connections by Country............................................................ 40
LIST OF TABLES
Table 1.������������� S7comm
OSI Layer Model. Adapted from [41]........................................ 18
Table 2.������������� Protocol
Servers Used by Conpot�s GridPot Template............................. 23
THIS PAGE INTENTIONALLY LEFT
BLANK
LIST OF ACRONYMS AND ABBREVIATIONS
ADU�������������������������� application
data unit
CDF��������������������������� cumulative
distribution function
COTP ������������������������ connection-oriented
transport protocol
CPS���������������������������� cyber-physical
system
DOS��������������������������� denial-of-service
FERC������������������������� Federal
Energy Regulatory Commission
GOOSE��������������������� Generic
Object Oriented Substation Event
GLM�������������������������� GridLAB-D
model
GPM ������������������������� GridPot
model
HTTP������������������������� Hypertext
Transfer Protocol
ICS����������������������������� industrial-control
system
IEC���������������������������� International
Electrotechnical Commission
IED���������������������������� intelligent
electronic device
I/O ����������������������������� input/output
MBAP������������������������ Modbus
Application Protocol
MIB��������������������������� management
information base
MMS�������������������������� Manufacturing
Message Specification
NAT��������������������������� Network
Address Translation
NCCIC���������������������� National
Cybersecurity and Communications Integration Center
NERC������������������������ North
American Electric Reliability Corporation
NMAP����������������������� Network
Mapper
OS ����������������������������� operating
system
PCAP������������������������� packet
capture
PDU��������������������������� protocol
data unit
PLC���������������������������� programmable
logic controller
SCADA��������������������� supervisory
control and data acquisition
SNMP������������������������ Simple
Network Management Protocol
S7comm��������������������� S7
communication
TCP���������������������������� Transmission
Control Protocol
UDP �������������������������� User
Datagram Protocol
VM����������������������������� virtual
machine
XML�������������������������� Extensible
Markup Language
ACKNOWLEDGMENTS
We would like to thank you, the reader.
THIS PAGE INTENTIONALLY LEFT
BLANK
In support of the National Security Strategy, this work
addresses the defenses of cyber-physical systems and critical infrastructure
pertaining to bulk electric systems. It examines cyber threats to smart energy
grids and aims to identify and characterize threats as learned through data
analysis of traffic within a simulated grid. This threat survey aims to build
awareness of potential vulnerabilities, thereby reducing the risk to actual
critical systems. Identification and characterization of threats applicable to
energy resilience contribute to the risk management of critical infrastructure
as directed in SECNAVINST 3501.1D [1].
For most of the twentieth century, security for a
critical-infrastructure operation such as the bulk power system meant only
physical security. Securing of that infrastructure was achieved with a fence
around dangerous or critical equipment such as regulators and transformers.
Later in the twentieth century, locks and barbed wire were added to increase
security in light of more direct threats of domestic terrorism. Current
security postures have expanded to include a cybersecurity architecture for
risk management and assessment [2].
In the early days of cyber-physical system development,
insufficient resources were spent mitigating risks to industrial systems [3].
Added risk can be seen in the loss of expertise in the actual operation of the
power equipment. A switch or transformer that was once operated manually by an
experienced operator may now be operated by an individual who has not
physically visited the site or has a limited understanding of the safe
operation of the power-system equipment. Alternatively, an experienced operator
is now interacting with a computer system instead of the physical equipment and
must learn how to correctly operate the computer equipment along with the power
system. An increasing risk is that the cyber-physical system could be
controlled by an unauthorized operator. Before the advent of cyber-physical
systems, to control an industrial system an unauthorized operator had to gain
physical access to the system. Now any device connected to the Internet risks
unauthorized access.
There is an elevated risk to network-connected critical
infrastructure, over other information technology systems, due to the potential
catastrophic impact if it were to fail. Cyber security of critical
infrastructure systems cannot be limited to traditional information-technology
defense measures, and defense in depth must be applied. An industrial-control
systems honeypot can be deployed as an extra layer for defense in depth, for
either all the system or a single high-value component. Hackers could be fooled
into thinking a honeypot is the real system. Adversarial targets, objectives,
or techniques could be learned by capturing the traffic patterns or noting
system changes within the honeypot. Not only would this be valuable
intelligence to ensure our limited time and resources are being applied against
the most targeted systems, but honeypots waste attacker time.
The National Cybersecurity and Communications Integration
Center (NCCIC) reports approximately 1.3 million reports of indications of
compromise (IOCs) related to cyber and communications since March 2016 [4].
Industrial Control Systems (ICS) have been targeted by four families of
malware; the most recent of which is CRASHOVERRIDE, aka Industroyer [5].
Stuxnet, BlackEnergy 2, and Havex were earlier tailored malware [6].
CRASHOVERRIDE used methods from these three previous malwares, creating a new
framework designed specifically to attack electric grids [6]. In December 2016,
a transmission-level substation was attacked in the Ukraine using CRASHOVERRIDE
[6]. CRASHOVERRIDE exploits network-communications and grid-operations
knowledge not specific to any configuration or vendor, and with minimal
tailoring can be repurposed to affect grid operations in North America, Europe,
and portions of Asia, and the Middle East [6]. Further proof that grid
operations can be affected by a cyberattack was demonstrated in a U.S.
Department of Energy test at Idaho Labs in 2007 known as the Aurora Experiment,
which caused the self-destruction of a replica power plant generator by means
of a cyberattack [7].
Sridhar et al. [8] identified cyber threats that target
the distribution portion of the bulk power system as load shedding, advanced
metering infrastructures, and demand-side management. The United States has
seen load-shedding incidents in recent years that have caused cascading power
outages affecting tens of thousands if not millions of customers. In 2007,
Tempe, Arizona experienced large-scale load shedding which affected 98,700
customers for almost an hour [8]. In 2003, the North America experienced its
most severe blackout on record [9].
Our research constructed a honeypot, a device with only
the purpose of collecting information about those that interact with the
honeypot [10]. This research implemented a high-interaction honeypot simulating
a smart energy grid. We used a high interaction honeypot to draw in more
advanced attackers. We used GridPot, an open-source symbolic cyber-physical
honeynet framework, to emulate realistic protocols common to industrial-control
systems [11]. The primary module we studied was Powerflow created by GridLAB-D,
which models an electrical distribution system [12].
Chapter II provides background information about
honeypots and previous work involving honeypots. Chapter III addresses the
North American Bulk Power System, industrial-control systems (ICS), GridPot�s
architecture, five common ICS network protocols, and similar work. Chapter IV
outlines our research methodology, with host-environment details, GridPot
modifications, data collection phases, and design of parsers for data analysis.
Chapter V discusses the results of our collection and data analysis. Chapter VI
contains our conclusions and recommendations for future work.
this page intentionally left
blank
A honeypot is a network node with no purpose other than
to detect and analyze unauthorized activity directed at computers and digital
devices [13]. Honeypots can provide data and information about an intrusion
afterwards to simplify its future
detection or provide information about an attacker [10]. Honeypots can be
classified based on their reactivity of low, medium, and high. Honeypots are a
safe way, with limited risk to a systems infrastructure, to capture, analyze,
and characterize threats against an information system.
Honeypots have two general uses as either a tool for
research or production [14]. Research honeypots collect data to be analyzed by
a researcher instead of notifying an administrator. Production honeypots work
with other security measures, like an intrusion detection system (IDS), to
provide defense-in-depth of networks. An intrusion-detection system uses known
signatures and threat characteristics to alert administrators of unauthorized
users or tools on a network. A honeypot could be employed in conjunction with
such a system to detect activity for which the system has no signature.
The usefulness of honeypots is related to how much data
they collect. Honeypots can use deception to entice attackers into revealing a
richer set of information about their attacks [15]. Honeypots that conceal
their purpose through deception are more productive because attackers do not
want to interact with honeypots [15]. High-interaction honeypots can confuse
attackers through program-based or scripted interaction designed to encourage
further exploration. The longer an attacker interacts with a honeypot, the more
complete the data set will be.
A honeynet is a network of
honeypots to simulate production networks without actually producing anything.
The Honeynet Project worked to discover using honeynets the tools and tactics
of potential attackers, to improve cyber-security methods [16]. How an
organization might integrate a honeynet into an existing production network is
depicted in Figure 1. A sensor captures key information from attempted
accessing of the honeynet from the Internet. A honeynet deployed in this manner
could collect information from attackers outside of the organization as well as
insider threats.
Figure 1.
A Honeynet. Source: [16].
To collect data from attackers, one could focus on
logging key strokes of attackers [16]. However, keystrokes do not capture the
activity of a script that the attacker might deploy. Focusing on any one layer
of the TCP/IP stack with focused tools has weaknesses since no single layer
tells a complete picture. Instead, several collection points are necessary to
understand motives and effects of attackers as with packets, log flows, and log
keystrokes.
A botnet is a collection of compromised computers
controlled by a single site (Figure 2). A botnet might be used to send spam,
mine cryptocurrencies, or promote accounts on social media.
Figure 2.
A Botnet Layout. Source: [17].
Botnets have evolved to become aware of honeypots by
detecting firewalls and filters on outbound traffic [17]. Honeypots desire to
limit their liability in case they are used to launch an attack on a third
party. This could be done with an intrusion-detection system that filters for
outbound activity. However, a bot controller can detect if a bot is prevented
from sending malicious data (Figure 3).
Figure 3.
Botnet Detection of a Honeypot. Source: [17].
Despite the benefits in deploying a honeypot to gather
information about a potential vulnerability or threat to a network, there are
some risks as well. Some researchers would argue a honeypot that collects IP
addresses, timestamps, and protocol data could be considered a personal-data
collection system [18]. This article states there is a privacy concern when
collecting network data, since IP addresses have been used to link criminal
behavior to individuals. The article mentions the existence of potential
privacy concerns in the deployment of honeypots and honeynets and in their
ability to share threat data with peers and partners. Finally, this article
mentions publishing information collected from a honeypot could provide
attackers with information to launch an improved attack.
A cloud-based industrial honeypot project deployed a
large-scale low-interaction honeypot system for 28 days using Amazon�s EC2
cloud environment [19]. This experiment monitored the protocols DNP3, ICCP,
IEC-104, MODBUS, SNMP, TFTP, and XMPP. This experiment also categorized
interaction as either related to Shodan (or Shodan�s subdomain) or not. The
researchers concluded that reconnaissance activities occurred more often than
actual attacks and this reconnaissance targeted individual industrial protocols
rather than combinations of different protocols. Counts by protocol are shown
in Figure 4. Additionally, the researchers identified a positive correlation
between MODBUS reconnaissance activity by non-Shodan sources following Shodan
discovery of MODBUS-enabled devices.
Figure 4.
Results from Cloud-Based Industrial Honeypot Deployment. Source: [19].
HoneyPhy,
a cyber-physical system honeypot framework, was developed to better understand
attacks against cyber-physical systems [20]. It addressed the problem that
existing frameworks could be unrealistic in modeling device physics and
device-actuation times and therefore could be easily identifiable as a
honeypot. The authors designed two proof-of-concept systems aimed at presenting
convincing honeypots. The initial version provided general structure-modeling
processes and devices implementing a simple heating-ventilation system. The
extended version provided deployment and log-collection specifications for a
simple water-treatment system. The authors claimed both versions appeared to
work well [20].
THIS PAGE INTENTIONALLY LEFT
BLANK.
We used GridPot as the base for our experiments. It is
the only open-source high-interaction industrial-control system honeypot available
at the time of this writing.
A.
bulk power system
Within the utilities industry, the term �electric grid�
refers to the high-voltage transmission lines which crisscross the area of
transmission of a bulk power system, connecting generation sources to local
distribution sub-stations [21]. The bulk power system of the United States is
owned and operated by dozens of individual corporations, municipalities, and
cooperatives. The U.S. Department of Energy is responsible for policies
concerning power generation and transmission in the United States, among other
things. The Federal Energy Regulatory Commission (FERC), an independent
regulating agency, is responsible for the safety, security, and availability of
the whole of the U.S. bulk power system. FERC has appointed the North American
Electric Reliability Corporation (NERC) as its agent to develop and monitor
standards for reliability and safety. NERC provides guidance to the regional
authorities responsible for geographic zones depicted in Figure 5.
Figure 5.
North American Bulk Power System Interconnections.
Source: [22].
A bulk power system is divided into four sectors:
generation, transmission, distribution, and end-use. The generation of
electricity occurs in coal-fired plants, natural-gas plants, solar farms, wind
turbines, and hydroelectric plants. Generated electricity is passed through
transformers to step up voltage to a very high level where it is then
transferred to transmission lines [21]. Transmission lines deliver electricity
to substations. Substations use a transformer to step down the voltage from
high to low voltage before it is distributed to end users. The process of power
flow from generation to customer is depicted in Figure 6.
Key components of substations include transformers and
switches [23]. Transformers change voltage when passing current from one
circuit to the next. A switch is a device used to
direct the flow of current by opening and closing a circuit. A
substation employs each of these devices to safely control the transfer of current
from transmission to distribution. We also used regulators in our model.
Regulators ensure a constant and safe voltage level is maintained throughout
the bulk power system.
Figure 6.
Basic Structure of the Electric System. Source: [9].
In the second half of the twentieth century,
technological advances were made that allowed grid operators to monitor and
control portions of the electric grid without being at each physical location.
Industrial control systems (ICS), Supervisory Control and Data Acquisition (SCADA)
systems, cyber-physical systems (CPS), and Intelligent Electronic Devices
(IEDs) were part of a larger movement of industrial automation.
Industrial-control systems provide the ability to control multiple physical
devices through connected communications. SCADA devices, a subset of
industrial-control systems, allow operators to monitor many devices over a wide
area. Cyber-physical systems integrate physics and logic to allow interaction
between digital, analog, physical, and human components [24]. An IED, like a
controller or a digital relay, is a device capable of sending or receiving data
or control to or from an external source or a combination thereof [3].
U.S. Navy ships and shore facilities depend on critical
infrastructure that includes industrial-control systems. Shipboard
cyber-physical systems include mechanical and electrical-control systems used
to generate power for driving the ship. Critical power-supply systems and
components necessary for operations are found on many military bases, including
naval facilities. All are vulnerable to the same threats as a shore-based
industrial-control system.
Internet-connected devices including industrial-control
systems, can be found using Shodan, an Internet-connected device-search engine.
Released in 2009, Shodan provides a suite of services used by researchers and
technical professionals [54]. Tools offered on Shodan�s subscription site
include location mapping of connected devices and searching using protocol
headers to fingerprint devices. Shodan's tool �Honeypot Or Not� returns a
report indicating if a device at a particular IP address is a honeypot or
another type of connected device [26].
Conpot is a low-interaction industrial-control-system
honeypot associated with the Honeynet Project, an international organization
dedicated to security research [27], [28]. Conpot was developed in 2014 to
provide researchers with attack data on supervisory control and data
acquisition by simulating a Siemens SIMATIC S7-200 programmable logic
controller (PLC) with common industrial-control-system protocols [29]. Conpot was
written in Python 2.7 and later upgraded to Python 3. Conpot is connected to
the Internet and listens on common ports used by industrial-control system
devices; it records data from other computers and devices that attempt to
connect to those ports over the Internet. The data is recoded in a log for a
researcher to analyze. The latest version of Conpot, version 0.6.0, runs nine
protocol-related servers with the default template [30]. These protocol-related
servers are MODBUS, S7Comm, HTTP, SNMP, BACNET, IPMI, ENIP, FTP, and TFTP. Our
honeypot uses Conpot version 0.4.0, enabling MODBUS, S7Comm, HTTP, and SNMP
servers on startup.
GridLAB-D is a simulation and analysis tool for
power-distribution systems [31]. It was developed by the Department of Energy
and Pacific Northwest National Laboratory to offer users algorithms to model
and test distribution systems at a low cost. We used the Powerflow module to
simulate voltage and current values across an IEEE 13 node grid model with 15
houses [32]. The Powerflow module is written using GridLAB-D model and
Extensible Markup Language (XML) syntax and contains Powerflow objects and
schedules [33]. GridLAB-D model syntax is similar to C++ though it is not a
procedural language [34]; objects are described in terms of properties and
parameters. Schedules use local time to change values in a predefined manner [35].
GridLAB-D model objects that were important to our
honeypot were node, link, switch, transformer, and regulator. Object node
properties include �phases�, which is used to represent a three-phase
connection in terms of �A�, �B�, and �C�, as well as, �N� for neutral phase.
Object link properties include the following: �from� and �to� for referencing
node object connections; �status� used in terms of open or closed; and
�power-in� and �power_out� to express power flow in volt-amperes. The switch,
transformer, and regulator objects all include properties inherited from the
node and link objects. Inherited object switch properties include �status�,
�phases�, �from�, �to�, �power_in�, and �power_out�. Inherited object
transformer properties include �status�, �phases�, �from�, and �to�, and
contain three additional properties: �ambient_temperature� to express the
temperature around the transformer, �winding_hot_spot_temparture� to express
the temperature of the transformer windings, and �configuration� to describe
the specific transformer implementation. Inherited object regulator properties
include �status�, �phases�, �from�, and �to�, with four additional properties:
�tap_A�, �tap_B�, and �tap_C� to express the position of the tap, and
�configuration� to describe the specific regulator implementation. GridLAB-D
model objects that load or transform values can include schedules as parameters
to change values over time. GridLAB-D model schedules are defined in
<minutes hours days months weekdays value> form.
GridPot is an open-source honeypot framework for modeling
electric grids [36]. GridPot uses a honeypot layer and a modeling layer to add
electrical components and integration between GridLAB-D and Conpot, including
IEC 61850 communication [36]. GridPot�s honeypot layer is derived from Conpot,
adding an XML-formatted GridPot template that specifies to which GridLAB-D
model (GLM) to link. Additional Python-coded GridPot files are included in the
honeypot layer to retrieve parameter values from the running model in real-time
using TCP port 6267. GridPot�s primary modeling layer uses GridLAB-D�s
Powerflow module. GridPot model (GPM) configuration files were added to the
modeling layer. A GPM configuration file names objects specific to a GLM and is
referenced by the GridPot template, thus linking the two layers together. A
visual of the linkages between GridPot�s honeypot and modeling layers is shown
in Figure 7. We obtained the GridPot source code at
https://github.com/sk4ld/gridpot [36].
Our GridPot configuration will be discussed more in Chapter IV.
GridPot source code includes additional modeling features
for intelligent electronic devices (IEDs) under an electric-components
subdirectory [36]. This subdirectory contains code written in C and XML syntax
to simulate a GE Brick Merging Unit and a generic input/output (I/O) switch
control device. GridPot demo commands state to �start configured IEDs,� which
implies to start the GE Brick Merging Unit and the switch control device [36].
Our honeypot does not utilize the GE Brick device and we left this additional
modeling piece for future work.
Figure 7.
GridPot Architecture
The Hypertext Transfer Protocol (HTTP) is an
application-layer protocol used to exchange messages between two programs, a
client and a server. The World-Wide Web uses HTTP to send request and response
messages across the Internet, most often using transmission control protocol (TCP)
port 80 [37]. Clients can make requests of the server via pre-defined methods,
such as GET, POST, OPTIONS, HEAD, and TRACE. The GET method retrieves the
requested specified information. The POST method requests the identified
resource accepts the enclosed entity and is designed to add, append, or make
comments to resources. The OPTIONS method requests information (i.e., server
capabilities) about a specific resource. The HEAD method requests a return of
meta-information (header fields) of the requested entity and is identical to
GET except without the message-body. The TRACE method is used by a client to
gain diagnostic information by requesting a reflection of what the server
receives in the response body of the message. Status codes are sent by the server
in response to client method requests.
MODBUS is an industrial-control-system application-layer
protocol for client/server communication shown in Figure 8.� MODBUS was
introduced in 1979 and is the most common industrial-control-system protocol
currently [38]. It is used to remotely start operations by devices including
programmable logic controllers (PLC), control panels, and input/output (I/O)
devices. [39].
Figure 8.
MODBUS Communication of Serial Line (left) and Ethernet Using TCP/IP
(right). Source: [38].
Four MODBUS message types are sent between client and
servers on an Ethernet TCP/IP network: MODBUS Request, MODBUS Confirmation,
MODBUS Indication, and MODBUS Response [40]. Information exchange is used to
read and write data access and for diagnostics purposes. MODBUS messaging
consists of a MODBUS Application Protocol header (MBAP header) and a protocol
data unit (PDU) consisting of a function code and data. All of these are
encapsulated by the MODBUS TCP/IP application data unit (ADU).
S7comm is a Siemens proprietary protocol used for data
exchange between PLCs and PLC programming. It is used to enable supervisory
control and data acquisition (SCADA) systems to access PLC data [41]. It uses
the Connection-Oriented Transport Protocol (COTP), the International Standards
Organization ISO-on-TCP transport services, and TCP port 102. Message types
include 0x01-Job Request, 0x02-Ack, 0x03-Ack-Data, and 0x07-Userdata [42]. A
basic model for S7comm is shown in Table 1.
Table 1.�
S7comm OSI Layer Model. Adapted from [41].
The Simple Network Management Protocol (SNMP) uses the
User Datagram Protocol (UDP) on port 161 and allows network administrators to
remotely manage, monitor, and configure network devices [43]. SNMP uses an SNMP
Manager on the controller, an SNMP Agent on the device, and a �management
information base� [44]. Central management consoles for industrial control
systems use SNMP to manage and maintain a network of PLCs [3].
Manufacturing Message Specification (MMS) and Generic Object
Oriented Substation Event (GOOSE) protocols are nested under IEC 61850. MMS is
similar to the proprietary S7Comm protocol in that it uses TCP port 102 for
SCADA monitoring and supports client/server communications [45]. GOOSE is used
for sending command requests and status updates between IEDs and controllers
using Ethernet-based multicast communications [45].
D.
PREVIOUS GRIDPOT AND CONPOT WORK
A 2015 Florida State University study about
vulnerabilities of cyber-physical systems created GridPot as a proof of concept
for physics-based intrusion detection, threat intelligence collection, and as
an additional capability for CPS real-time situational awareness [46]. This
experiment integrated a modified GridLAB-D with simulated substations, a SCADA
operator interface, and IEDs using IEC 61850 protocols. A switching attack
against IEC 61850 protocols was replicated during this experiment to illustrate
the real-time physics analysis performed by GridPot. Our research differs in
that it is focused more on network-based threat characterization rather than
physics-based impact provided by a modified GridPot.
The effectiveness of Conpot was analyzed in another
project [47]. This experiment ran multiple virtualized Conpot instances using
both the default template and a template simulating a gas-tank level, which
were deployed globally using Amazon Web Services for an 18-day period. Conpot
instances that were dependent on obsolete repositories had to be manually
added. Analysis was conducted on scans using multiple Nmap flag settings in
addition to results of Shodan scans. They concluded that Conpot accurately
simulated SCADA ports but could be identified as a honeypot based on other open
ports that appear when the system is scanned.
Work at NPS conducted a study to generate cyberattack
data specific to industrial control systems using Conpot version 0.5.1 [48].
The objective was to better understand indications of compromise on industrial
control systems. This honeypot was deployed outside of the Naval Postgraduate
School�s firewall from October 2017 to February 2018. It used Conpot�s default
template to monitor the protocols HTTP, EtherNet/IP, MODBUS, S7Comm, SNMP,
BACnet, and IPMI. This honeypot averaged 10.51 attacks per day from 54
countries with a decrease in traffic over time, as shown in Figure 9. This
research showed that Conpot is a viable platform for honeypot research for
industrial control systems, and suggested that the logs provided by Conpot need
to be supplemented with a packet capture tool like Wireshark to extract more
information about attacks.
Figure 9.
NPS Conpot Data. Source: [48].
A.
host environment
Our experiment used two environments, a test environment
and a live environment. The test environment was used for the initial code
review and baseline data collection. It also enabled internal testing and
host-to-virtual-machine testing to establish baseline data for analysis and
comparison. The live environment enabled external user access and threat data
collection from outside of the campus firewall.
Our live environment was a Dell XPS 8910 desktop computer
running a Windows 10 Home OS with 16 GM RAM and a 925 GM hard disk. We used
Oracle VM Virtualbox 5.2.22 to import and open our GridPot VM from the test
environment. Our test environment was a Dell Precision M6800 laptop computer
running a Linux x86-64 Ubuntu 18.04.01 LTS operating system with 16 GB RAM and
a 750 GB hard disk. We used Oracle VM Virtualbox 5.2.22 to install a virtual
machine in which GridPot was installed. It ran the same operating system as the
host and was configured with 10.7 GB of RAM and a 300 GB virtualized hard disk.
We used Network Address Translation (NAT), Host-Only-Adapter, and
Bridged-Adapter network settings in our live and test environments. The design
and layout of our live setup is depicted in Figure 10.
Figure 10.
Live Environment Setup. Adapted from [48].
Our GridPot had a honeypot layer and a modeling layer as
discussed in Chapter III. The honeypot layer initialized Conpot using the
GridPot template and the modeling layer initialized a GridLAB-D model named
IEEE_13_Node_With_Houses. Our GridPot installation steps are provided in
Appendix A.
Our Conpot was launched using a modified GridPot
template. An example output of a successful launch is shown in Figure 11. We
updated the �gridpotmodel_file� field value to link with our modified GPM file
named IEEE_13_Node_With_Houses shown in Appendix C. Four protocol servers were
started upon launch as written in the original source code and are depicted in
Table 2.�� MODBUS is used on TCP/IP port 502 connecting to one MODBUS client
and two MODBUS servers. Details of how we started Conpot using the GridPot
template can be found in Appendix B.
Figure 11.
Launching Conpot Using the GridPot Template
Table 2.�
Protocol Servers Used by Conpot�s GridPot Template
2.
Modeling Layer
We used the IEEE_13_Node_With_Houses model, which
simulates an IEEE 8500 distribution model with 13 nodes and 15 houses. A
partial diagram of this distribution model is shown in Figure 12. We chose this
model since it contained switch, transformer, and regulator objects used for
the Conpot integration and required minimal code modifications. A schedule was
used to alter power flow readings across the switch. Real-time power-in and
power-out simulated switch parameter values, which were displayed on our
web-based interface. We created a GPM file to link with the switch,
transformer, and regulator objects specified in the IEEE_13_Node_With_Houses
GLM file by modifying an existing GPM file to ensure proper formatting. These
modifications are shown in Appendix C. Chapter III explained how the honeypot
layer and modeling layer were linked using the GPM and GLM files.
The switch is located between nodes 671 and
692. The transformer is located between nodes 633 and 634. The regulator is
located between nodes 630 and 650.
Figure 12.
Partial Layout of our IEEE 13 Node Grid Model.
Source: [46].
We collected data in three phases; an internal testing
phase from within the virtual machine, a host-to-virtual-machine phase, and a
live data collection phase. The internal phase tested GridPot�s source code and
how Conpot and GridLAB-D interacted. The host-to-virtual-machine phase
established a baseline data repository to compare with live data results. The
live data collection phase was the period in which the honeypot was connected
to the Internet.
We used the tools Wireshark, Nmap, Nessus, Metasploit,
and Netstat. Wireshark, a network-protocol analyzer, captured packets on the
network [49]. Nmap probed a target system to determine port status, active
services, operating system, and MODBUS identification and device information [50].
Nessus, a vulnerability-assessment tool, scanned for operational-technology
devices and SCADA protocols active on the network [51]. Metasploit, a penetration-testing
framework, detected MODBUS, found MODBUS units, and read MODBUS client data [52].
Netstat, a command line tool, printed networking subsystem information and
showed listening and non-listening network sockets [53].
We conducted internal tests of our GridPot installation
using a variety of tools. We altered the Conpot code to use its localhost IP
address instead of retrieving the host environment�s external IP address upon
start. This enabled our honeypot to keep all network traffic internal to our
machine. First, we tested if our web-based interface display was accurate to
the running model by pointing a web-browser to GridPot�s HTTP server using
localhost IP address and TCP port 80 and comparing the results to the GridLAB-D
model instance that listened on port 6267. We then used the Netstat tool to
determine which ports were opened by GridPot. We tested this by executing
�netstat �ano� from the command line before and after starting GridPot. We then
ran scans using Nmap, Nessus, and Metasploit against our honeypot. We focused
these scans on open ports and probed for operational-technology devices using
the MODBUS protocol by running MODBUS detection, MODBUS discovery, and MODBUS
interaction scans. We focused on MODBUS since it is the most commonly known
industrial-control-system protocol. Details of our internal testing are in
Appendix D.
Host-to-virtual-machine baseline testing also used Nmap,
Nessus, and Metasploit. Before starting this phase, we altered our network
connectivity from disabled-network status to host-only status. This allowed for
connectivity between our host machine and our GridPot virtual machine without
allowing the GridPot virtual machine to connect to the Internet. To prevent
GridPot from being viewable from outside our host machine, we left the previous
change to the Conpot startup using its localhost address.
We tested the host-only network status using ping
commands between our host and our GridPot virtual machine and confirmed receipt
of a �network is unreachable� error when
trying to ping an arbitrary IP address. The same scans using Nmap, Nessus, and
Metasploit were then performed with the target IP address set to the IP address
of our GridPot virtual machine instead of the localhost IP address. To increase
our log data for comparison against potential live denial-of-service (DOS)
attacks, we conducted an additional scan using an auxiliary DOS Metasploit
module during this baseline testing phase. Details are found in Appendix E.
We modified portions of the Conpot configuration that
are common across Conpot instances to attract more interaction from would-be
attackers. It was our desire that removing the simple indicators may prolong
the attacker's identification of our system as a honeypot. The longer we can
keep an attacker interacting with GridPot, the more data we have to determine
what their intentions may be. Live data collection altered the static default
configurations that could be used to fingerprint an instance of Conpot. These
changes were accomplished by modifying the GridPot template XML file found in
the Conpot subdirectory. We assumed by altering these default values we would
increase our honeypot�s deception to both attackers and web crawlers. We used
Shodan to determine the success of our deception efforts to suggest our
honeypot was an actual energy distribution system [54]. To do this, we surveyed
our IP address using Shodan�s Honeyscore website [26].
Our honeypot collected data over 19-days from April
11-30, 2019. GridPot ran continuously during this time period except when we
brought down the honeypot layer on April 12 due to a broken link between our
honeypot and our modeling layer. We fixed this issue and re-launched Conpot
using the GridPot template, using the same procedure shown in Appendix B. We
used Wireshark to complement the Conpot log to obtain more details about
external source interactions with our honeypot. Wireshark packet capture (Pcap)
files were saved every three hours. The Pcap files and honeypot logs were
backed up to the host machine and stored on a NPS web-based shared platform
twice a day on weekdays and once a day on weekends. GridPot was not affected by
this back-up procedure and Pcap files older than two days were deleted from the
live environment to free disk space.
Wireshark has many built-in capabilities to analyze and
report information from a packet capture (Pcap) file. However, these
capabilities are limited to single Pcap file. To analyze sets of files, we
wrote a Python program that uses the DPKT Python package [55] to examine the
packets to determine the source IP address of traffic, the unique ports with
which each address is attempting to connect, and the time period in which each
IP address is actively sending packets to the honeypot. This data indicates
actors that persisted and were searching for vulnerabilities, as well as actors
conducting port scans. Results of our analysis tool were compared to the Conpot
logs which provide source information as well as the payload of each packet.
However, these logs do not provide enough information without an additional
layer of analysis to monitor trends over time. We left this additional
capability as future work. The Conpot log is sufficient, however, to provide an
indication that an actor is active on a network without authorization.
We used Microsoft Excel to filter the Conpot log by
protocol. We then were able to count specific protocol instances to include the
number of HTTP, MODBUS, and S7Comm sessions, HTTP requests and responses, HTTP
versions, HTTP methods, MODBUS connections, MODBUS function codes, S7Comm
connections, COTP connection requests, and S7 packets and PDU Types. We also
used comparison functions to count repeat attacks, and graphing functions to
show HTTP requests by source IP address, MODBUS connections and traffic by
source IP address, S7Comm connections by source IP address, HTTP request method
distribution, country interaction counts using HTTP, MODBUS, and S7Comm, and
cumulative distribution functions (CDFs) of obvious scanning attacks using
HTTP.
The GridPot source code pulled from GitHub was last
updated in March 2015. There were many updates to Conpot and GridLAB-D since
then, including GRIDLAB-D upgrading to Python 3.6. This caused many broken
connections between the honeypot and the modeling layers which resulted in
additional dependencies needing installation before we could successfully run
our GridPot. We used the latest release of Ubuntu which probably contributed to
some of the additional older dependencies needing installation. A list of the
additional dependencies we installed are listed in Appendix A.
The GridPot source code also did not include all of the
code needed to run the IEDs found in the electric-components subdirectory or
reference what open-source tool was used to integrate a functioning SCADA
operator interface. In-depth code review leads us to believe IEC 61850
communications would occur by launching the IEDs. We assume there is additional
software needed to implement both the IEDs and SCADA operator interface and
suggest they be implemented in future work.
Results of our internal and host-to-virtual-machine
testing eventually proved successful. It confirmed our web-based interface
accurately displayed values of the running GridLAB-D model. Netstat results
confirmed that the four protocol ports shown in Table 2 were open. Results of
the Nmap, Nessus and Metasploit scans also saw these ports as open and saw that
MODBUS-enabled devices were running on our honeypot. We confirmed that results
received from internal testing were seen again during host-to-virtual-machine
testing.
The honeypot first went online on April 11, 2019. The
data collected by Wireshark as of April 29 totaled 9,240,989 packets. The bulk
of this data was the result of network broadcast messages, address resolution
protocol messages, and standard administrative traffic from the host machine,
such as software updates. Filtering this data for network traffic to and from
our honeypot reduced the data to 1,525,059 packets and 165 MBs. The traffic
flow averaged 1.1 packets per second, and the average packet size was 108 bytes
for 113 bytes per second. The greatest number of packets to the honeypot came from
an IP address to a California-based cloud-hosting corporation which accounted
for 1,013,726 packets and 63 MB of data. The second-highest source of packets
was an IP address registered to an LLC in St. Petersburg, Russia, which was
responsible for 56,280 packets and 3,221KB. Censys.io traced this address to a
Debian-based SSH server in Amsterdam. This address sent 45,657 packets to
GridPot over 12 days of which 38,754 were SYN packets sent to GridPot; also RST
packets were sent to ports 5160, 5164, 5110, 5093, 5094, 5112, 5134, and 5156.
Our Conpot log recorded 9,641 HTTP requests, 621 MODBUS
connections, 606 MODBUS traffic instances, and 102 S7Comm connections from
April 11-30, 2019. HTTP was the most commonly used protocol during this 19-day
period with heavy scanning almost every day. Heavy scanning using MODBUS was
seen twice. Thirty-nine unique source addresses sent packets to our honeypot
multiple times, with return visits either repeating the initial interaction or
showing more HTTP requests which could be attributed to further reconnaissance
efforts using learned information from the initial contact. Overall honeypot
activity by protocol from April 11-30, 2019, is shown in Figure 13.
Figure 13.
Protocol Count by Source IP Address from April 11-30, 2019
The source addresses observed are shown in Figure 14.
This was determined using the GeoLite2 plugin for Wireshark, developed by
MaxMind [56]. The countries with more than 100 unique source addresses that
probed our GridPot are depicted in Figure 15.
Figure 14.
Source IP Address Endpoints
Figure 15.
Countries Using More Than 100 Unique IP Source Addresses
The greatest number of packets came from an address
registered to Fastly, a content delivery network provider. GridPot exchanged
84,588 packets with just one Fastly address which MaxMind attributes to a
location in Seattle. Traffic from this address covered the entire duration of
our collection. The traffic can be distinguished by its volume and persistence
on our honeypot, but it contained over 26,000 retransmissions of nearly
identical ACK messages.
Our Conpot log recorded 453 new HTTP sessions, 9,641 HTTP
requests, and 9,591 HTTP responses in April 11-30, 2019. Of the 19,232 combined
HTTP requests and responses, 18,886 were HTTP v1.1, 160 were HTTP v0.9, 108
were HTTP v1.0, and 78 were either bad requests or had no HTTP version listed.
Seven different HTTP methods were seen in the 9,641 HTTP requests (Figure 16),
including 79 requests that did not use any methods and 78 invalid requests,
shown as �None� and �Bad�, respectively.
Figure 16.
HTTP Request Method Distribution
Sixty-seven different countries interacted with our
honeypot using HTTP during April 11-30, 2019. The top three were Brazil, the
United States, and China. Eleven countries interacted with our honeypot at
least 10 times, and 42 of the countries only interacted with our honeypot three
times or less (Figure 17).
Figure 17.
HTTP Country Count from April 11-30, 2019
Significant spikes in the number of HTTP requests were
seen almost daily as shown in Figure 18.� Each peak contained roughly the same
number of GET and POST request methods in the same order with varying speeds.
This led us to believe these attackers used the same HTTP scanning tool to
conduct the attacks. Fourteen attacks were using IP addresses associated with
China, two with Hong Kong, and one with Mexico.
We compared this apparent scanning to our
host-to-virtual-machine logs and determined that the real scans did not match
any of our test scans. We assess these scans are likely all using the same
scanning tool. The real scans presented a strong correlation in the quantity
and order of GET and POST request methods sent. This trend was not seen in any
of our test scan results. To show the real scan correlation, we plotted CDFs of
these attacks using distributions of HTTP GET shown in Figure 19, HTTP Post
shown in Figure 20, and the combination of HTTP GET and HTTP POST shown in
Figure 21. The x-axis for each of these figures is calculated using the start
and end time of the attacks; the y-axis shows the distributions of the protocol
method.
Figure 18.
HTTP Requests from April 11-30, 2019
Figure 19.
CDF of Attacks Using HTTP GET
Figure 20.
CDF of Attacks Using HTTP POST
Figure 21.
CDF of Attacks Using HTTP
Our Conpot log recorded 621 MODBUS connections and 606
MODBUS log entries during April 11-30, 2019. The majority of the MODBUS
traffic, 597 entries, did not use a function code and the remainder of the
traffic count was split between function codes 17 and 43, totaling 5 and 2,
respectively. MODBUS connections and traffic over time are shown in Figure 22,
highlighting outliers we investigated further.
Figure 22.
MODBUS Connections and MODBUS Traffic from April 11-30, 2019
Protocol scanning using MODBUS was apparent in a manual
review of the Conpot log, and was indicated by incrementing slave ID numbers
with each new request seen with both April 17 and April 25. We compared the
observed scanning to our host-to-virtual-machine logs and inferred that Nmap
and the �modbus-discover.nse� script were used in both cases. The similarity in
the sequencing of function code, slave ID, request values, and response values
can be seen in the log of our Nmap scan shown in Figure 23 compared to the logs
of attack scanning shown in Figures 24 and 25.
Figure 23.
Conpot Log from Host-to-Virtual-Machine Nmap Scan Using MODBUS Script
Figure 24.
First Conpot Log of Likely Nmap Scan Using MODBUS Script
Figure 25.
Second Conpot Log of Likely Nmap Scan Using MODBUS Script
MODBUS traffic originated from eight different countries,
with the United States and Romania being the top two. The MODBUS traffic
breakdown by country is shown in Figure 26.
Figure 26.
MODBUS Traffic Distribution by Country
During April 11-30, our Conpot log showed 20 new S7Comm
sessions, 102 S7Comm connections, 13 COTP connection requests, and 19 S7
packets. Messages of type-1 and type-7 were the only ones seen, of counts 6 and
13, respectively. S7Comm connections over time are shown in Figure 27.
Figure 27.
S7Comm Connections from April 11-30, 2019
Eleven different countries connected to our honeypot
using S7Comm; Japan connecting the most totaling 49 times, followed by the
United Kingdom and the United States with 28 and 10, respectively. S7Comm
connections by country is shown in Figure 28.
Figure 28.
S7Comm Connections by Country
We used Shodan's �Honeypot Or Not� tool to determine the
convincingness of our GridPot implementation. This tool calculates the
probability (ranging from 0.0 to 1.0) that an IP address is a honeypot [26].
Our IP address returned a score of 1.0, indicating with a strong probability
that it is a honeypot. Even though our GridPot was not deceptive enough to fool
Shodan, this did not prevent attackers from interacting with our honeypot.
Either would be attackers were not aware that Shodan had classified our IP
address as being attached to a honeypot or they chose to interact with it
anyway.
We used both Conpot log data and Wireshark Pcap files to
analyze the data captured by our GridPot. The Conpot log captured high-level
interaction data including timestamps, source IP addresses and ports, protocols
used, and basic protocol information, and it tagged each interaction with a
unique descriptor for source-flow analysis. Wireshark measured traffic patterns
and also provided more detailed information for anomalies highlighted with the
Conpot log. Wireshark tools provided several useful scripts for conducting
analysis. The Conpot log was specific only to our monitored protocols.
Additional knowledge of other attack vectors, like a high
number of packets sent to destination port 443, could prove useful for future
work. Abnormally large traffic periods were easily found by saving Wireshark
Pcaps every three hours. Pcap file sizes 10 times greater than the others stood
out for immediate analysis. Hard drive and virtual-memory size constraints were
not an issue during our research, but could become a problem the longer the
honeypot remained active due to multi-tool logging using Wireshark.
THIS PAGE INTENTIONALLY LEFT
BLANK
Much time was spent reconfiguring GridPot to ensure our
honeypot and modeling layers were working and integrating properly. This left
only a 19-day period for live data collection. Our GridPot implementation
proved successful at collecting intelligence for threat analysis during this
period. It is apparent that attackers will exploit any web-based
vulnerabilities in a SCADA external interface to gain access to the system, as
seen in the large amount of HTTP traffic captured by our honeypot. This
highlights the need to have SCADA administrators and operators trained on
information security to ensure unauthorized web-based access attempts are
stopped. Our GridPot also captured MODBUS scanning, indicating our simulated
grid was realistic enough to encourage specialized-protocol reconnaissance even
though it was fingerprinted as a honeypot by Shodan.
There are many different paths for follow-on work using
GridPot. As mentioned, GridPot runs on comparatively old versions of Conpot and
GridLAB-D. We recommend updating GridPot using the newest versions of Conpot
and GridLAB-D. Rebuilding efforts would need to include the GridPot layer links
shown in Figure 7 as well as the additional GridPot Python scripts in the
original source code under the Conpot protocol subdirectory.
Adding a SCADA operator interface is desirable since most
attacks were using HTTP. This would add an additional layer of deception, which
could increase the length of time attacks would interact with the honeypot.
Presenting a SCADA operator interface would also make a more attractive target
as it would be more believable than the read-only web page which GridPot
currently uses.
We recommend integrating IED device simulations to learn
more about attacks using IEC 61850 communications. This could be simulated
similar to the original GridPot source code under electric-components
subdirectory, or by connecting a real physical device which could provide real-time
visual feedback.
Future work could also increase deception within GridPot
or increase interaction through use of honeynets. Deception could be added in
the Internet registry or DNS information as a better way to hide the honeypot�s
location. Attackers can easily fingerprint our honeypot currently by searching
a WhoIs registry to determine ownership and note the tie to NPS. Using
honeynets, multiple connected honeypots, could increase the number of attacks
by presenting more interfaces with which to interact, although it would be
harder to set up and maintain a honeynet.
A final suggestion is to use a low-maintenance and easily
deployable honeypot like T-Pot, an open-source multi-honeypot platform that can
be configured to run dockerized versions of well-established honeypots
including Conpot [57].
appendix A. GRIDPOT SETUP
This appendix summarizes the steps required to install
GridPot. We changed our virtual-machine network setting to use NAT during this
setup.
1.�� Installing
dependencies (from the command line)
$ sudo apt-get update
$ sudo apt-get upgrade
$ sudo apt-get install autoconf (this includes
automake)
$ sudo apt-get install libtool subversion python-dev
mysql-server
$ sudo apt-get install libmysqlclient-dev
libmysqld-dev
$ sudo apt-get install libxerces-c-dev
python-mysqldb python-pip
$ sudo apt-get install libcurl3 libcurl4-openssl-dev
libcurl4-gnutls-dev
$ pip install �U sphinx
$ sudo apt-get install python3-sphinx libxml2-dev
libxslt1-dev
$ sudo pip install lxml gevent python-dateutil mixbox
$ sudo pip install pyasn1 pycryptodomex pysmi
$ sudo apt-get install doxygen
$ sudo apt-get install libcppunit-dev libcppunit-doc
$ sudo apt-get install libncurses5-dev
libncursesw5-dev
2.�� Pulling
GridPot from GitHub
$ sudo apt install git
$ git clone [36]
3.�� Setup
Conpot
$ cd gridpot/conpot/
$ conpot/ sudo make clean (if rebuilding)
$ conpot/ sudo python setup.py install
4.�� Setup
GridLAB-D
$ cd ../gridlabd/3.1
$ autoreconf -isf
$ make
$ sudo make install
5.�� Setup
libiec61850
$ cd ../libiec61850
$ make
$ sudo make INSTALL_PREFIX=/usr/local install
This appendix summarizes the steps required to clear and
enable logging from both the honeypot and modeling layers, enable packet
captures, and run GridPot.
0. Save
& clear Conpot log
$ cat conpot.log >> conpot_original.log
$ vim conpot.log
:0 (to go to beginning of file)
dG (to erase file)
shift+zz (to save & exit)
1. Start Wireshark
$ sudo wireshark
2. Start
GridLAB-D model with output to screen and file
$ gridlabd -D run_realtime=1 --server --debug
--verbose IEEE_13_Node_With_Houses.glm 2>&1 | tee HousesOutput.txt
3. Start
Conpot using GridPot template
$ sudo conpot -t gridpot �v
this page intentionally left
blank
This appendix summarizes the modifications made to the
GPM file. GPM format is a configuration file written using Python�s
ConfigParser module. This module contains sections beginning with [section]
headers, trailed by �name: value� entries, and ignores comment lines beginning
with a �#� symbol [58].
IEEE_13_Node_With_Houses.gpm
#GRIDPOT model
for conpot instances
[conpot1]
switch:
671-692
transformer:
633-634
regulator: 650630
this page intentionally left
blank
This appendix summarizes the steps we completed for
internal testing using our interface, the NETSTAT command, and running Nmap,
Nessus, and Metasploit scans of our network. We disabled our virtual-machine
network settings during this internal testing phase.
1.� Web-based
Interface
Run GridPot (see Appendix B.)
Open web-browser to localhost:80
Open web-browser to localhost:6267/650630/status
Open web-browser to
localhost:6267/tp_line_conductor/resistance
Stop GridPot (ctrl+c)
Pull Logs (Wireshark, Conpot, HousesOutput)
2.� Netstat
$ netstat -ano
>gridpot/conpot/Print/netstat_beforeStart.txt
Run GridPot (see Appendix B.)
$ netstat -ano
>gridpot/conpot/Print/netstat_afterStart.txt
Pull Logs (Wireshark, Conpot, HousesOutput, Netstat
output files)
3.� NMAP
Scans
Run GridPot (see Appendix B.)
$ sudo nmap -p- -oN nmap_port.txt localhost
$ sudo nmap -v -sV -O -oN nmap_v-sV-O.txt localhost
$ sudo nmap --script modbus-discover.nse
--script-args='modbus-discover.aggressive=true' -p 502 -oN nmap_modbus.txt
localhost
Stop GridPot (ctrl+c)
Pull Logs (Wireshark, Conpot, HousesOutput, nmap
output files)
4.� Nessus
Scans
$ /etc/init.d/nessusd start
Open web-browser to https://localhost:8834
Run GridPot (see Appendix B.)
Run Nessus Advanced Dynamic Scan: set 127.0.0.1 as
the target, select �scan operational technology devices� (found under Host
Discovery) and match plugin description contains SCADA
Run Nessus Advanced Scan: set 127.0.0.1 as the
target, select �scan operational technology devices� (found under Host
Discovery) and enable all plugins including SCADA family
$ /etc/init.d/nessusd stop
Stop GridPot (ctrl+c)
Pull Logs (Wireshark, Conpot, HousesOutput, Nessus
reports)
5.� Metasploit Scans
Run GridPot (see Appendix B.)
$ msfconsole
$ use auxiliary/scanner/scada/modbus_findunitid
$ show options
$ set RHOST 127.0.0.1
$ run
$ back
$ use auxiliary/scanner/scada/modbusdetect
$ show options
$ set RHOST 127.0.0.1
$ run
$ back
$ use auxiliary/scanner/scada/modbusclient
$ show actions
$ set ACTION READ_COILS (READ_REGISTERS)
$ show options
$ set RHOST 127.0.0.1, RPORT 502, UNIT_NUMBER 1,
NUMBER 1 (10, 100)
$ set DATA_ADDRESS 1 (10001, 30001, 40001)
$ run
$ back
$ exit
Stop GridPot (ctrl+c)
Pull Logs (Wireshark, Conpot, HousesOutput, msf
output files)
THIS PAGE INTENTIONALLY LEFT
BLANK
This appendix summarizes the steps we completed for
host-to-virtual-machine scanning using Nmap, Nessus, and Metasploit. We changed
our virtual-machine network setting to use Host-Only Adapter during this
testing phase. We testing our connectivity using the ping command to/from our
host (X.X.X.1) and client (X.X.X.101) and confirmed receipt of a �network is
unreachable� error when trying to ping an arbitrary IP address.
1.� NMAP scans
See Appendix D.3., change localhost to X.X.X.101
2.� Nessus Scans
See Appendix D.4., change target from 127.0.0.1 to
X.X.X.101
3.� Metasploit Scans
See Appendix D.5., change RHOST from 127.0.0.1 to
X.X.X.101 and run all scans, including additional scan and dos attack listed
below before exiting.
$ use auxiliary/scanner/scada/profinet_siemens
$ show actions
$ show options
$ set INTERFACE enp0s3
$ run
$ back
$ use auxiliary/dos/scada/siemens_siprotec4
$ show options
$ set RHOST X.X.X.101
$ run
THIS PAGE INTENTIONALLY LEFT
BLANK
[1] ����� Department of the Navy Critical Infrastructure
Protection Program, SECNAV Instruction 3501.1D, Secretary of the Navy,
Washington, DC, USA, 2018.
[2] ����� NIST, NIST roadmap for improving critical
infrastructure cybersecurity, 2014. [Online]. Available: https://www.nist.gov/sites/default/files/roadmap-021214.pdf
[3]������� K. Stouffer, V. Pillitteri, S. Lightman, M.
Abrams, and A. Hahn, �Guide to Industrial Control Systems (ICS) security,�
NIST, Gaithersburg, MD, USA, NIST SP 800-82r2, 2015.
[4]������� J. Manfra, �NCCIC year in review 2017: Operation
Cyber Guardian,� DHS, NCCIC, 2017.
[5]������� NCCIC, �CRASHOVERRIDE malware,� Washington, DC,
USA, Alert ICS-ALERT-17-206-01, 2017.
[6]������� R. Lee, �CRASHOVERRIDE: analysis of the threat to
electric grid operations,� DRAGOS Inc, Hanover, MD, USA, version 2.20170613,
2017.
[7]������� C. A. Theohary, �Cyber operations in DoD policy
and plans: issues for congress,� CRS Report No. R43848, 2015.
[8]������� S. Sridhar, A. Hahn, and M. Govindarasu,
�Cyber�physical system security for the electric power grid,� Proceedings of
the IEEE, vol. 100, no. 1, pp. 210�224, Jan. 2012.
[9]������� U.S.-Canada Power System Outage Task Force,
�Final report on the August 14, 2003 blackout in the United States and Canada
causes and recommendations,� U.S. Dept. of Energy, Washington, DC, USA,
2004.
[10]����� M. Nawrocki, M. W�hlisch, T. C. Schmidt, C. Keil,
and J. Sch�nfelder, �A survey on honeypot software and data analysis,� Cornell
University, vol. arXiv:1608.06249v1, Aug. 2016.
[11] GridPot, �GridPot: symbolic cyber-physical honeynet
framework,� Accessed November 3, 2018. [Online]. Available: gridpot.org
[12]����� Pacific Northwest National Laboratory, �GridLAB-D
main page,� Accessed February 22, 2019. [Online]. Available: http://gridlab-d.shoutwiki.com/wiki/Main_Page
[13]����� R. A. Grimes, Hacking the Hacker.
Indianapolis, IN, USA: John Wiley & Sons, Inc., 2017.
[14]����� F. Zhang, S. Zhou, Z. Qin, and J. Liu, �Honeypot:
a supplemented active defense system for network security,� in Proceedings
of the Fourth International Conference on Parallel and Distributed Computing,
Applications and Technologies, 2003, pp. 231�235.
[15]����� N. C. Rowe, Honeypot deception tactics.� Chapter 3
in E. Al-Shaer, J. Wei, K. Hamlen, and C. Wang (Eds.), Autonomous Cyber
Deception: Reasoning, Adaptive Planning, and Evaluation of HoneyThings,
Springer, Chaum, Switzerland, 2018, pp. 35-45.
[16]����� L. Spitzner, �The Honeynet Project: trapping the
hackers,� IEEE Security Privacy, vol. 1, no. 2, pp. 15�23, Mar. 2003.
[17]����� C. C. Zou and R. Cunningham, �Honeypot-aware
advanced botnet construction and maintenance,� in International Conference
on Dependable Systems and Networks (DSN�06), 2006, pp. 199�208.
[18]����� P. Sokol, J. M�ek, and M. Hus�k, �Honeypots and
honeynets: issues of privacy,� EURASIP J. on Info. Security, vol. 2017,
no. 1, pp. 1�9, 2017.
[19]����� A. V. Serbanescu, S. Obermeier, and Y. Der-Yeuan,
�Threat analysis using a large-scale honeynet,� in Proc. of the 3rd Int.
Sym. for ICS & SCADA Cyber Sec. Res., 2015. [Online]. doi: http://dx.doi.org/10.14236/ewic/ICS2015.3
[20]����� S. Litchfield, �Honeyphy: a physics-aware
CPS honeypot framework,� M.S. thesis, Dept. of Elec. and Comp. Eng., Georgia
Inst. of Tech., Atlanta, GA, USA, 2017.
[21]����� Department of Energy, �Infographic: understanding
the grid,� Energy.gov, November 17, 2014. [Online]. Available: https://www.energy.gov/articles/infographic-understanding-grid
[22]����� North American Electric Reliability Corporation,
�Key players,� Accessed February 28, 2019. [Online]. Available:
https://www.nerc.com/AboutNERC/keyplayers/Pages/default.aspx
[23]����� �Substations,� in Electric Power System Basics
for the Nonelectrical Professional, John Wiley & Sons, Ltd, 2016, pp.
53�89.
[24]����� K. D. Thompson, �Cyber-physical systems,� NIST,
June 20, 2014. [Online]. Available: https://www.nist.gov/el/cyber-physical-systems
[25]����� T. Kiravuo, S. Tiilikainen, M. S�rel�, and J.
Manner, �Peeking under the skirts of a nation: finding ICS vulnerabilities in
the critical digital infrastructure,� in European Conference on Cyber
Warfare and Security; Reading, Reading, United Kingdom, Reading, 2015, pp.
137�144.
[26]����� Jmath, �Honeypot or not?� Accessed March 26, 2019.
[Online]. Available: https://honeyscore.shodan.io/
[27]����� L. Rist, �About,� Conpot ICS/SCADA Honeypot.
Accessed February 23, 2019. [Online]. Available: http://conpot.org/
[28]����� F. Leder, �About the honeynet project,� The
Honeynet Project. Accessed March 19, 2019. [Online]. Available: https://www.honeynet.org/about
[29]����� C. Scott, �Designing and implementing a honeypot
for a SCADA network,� SANS Institute Reading Room, p. 39, Jun. 2014.
[30]����� L. Rist, �GitHub - mushorg/conpot: ICS/SCADA
honeypot,� February 04, 2019. [Online]. Available: https://github.com/mushorg/conpot
[31]����� Department of Energy, �GridLAB-D a unique tool to
design the smart grid,� August 2018. [Online]. Available:
https://www.gridlabd.org/
[32]����� Pacific Northwest National Laboratory, �Power flow
user guide,� GridLAB-D ShoutWiki, October 20, 2017. [Online]. Available:
http://gridlab-d.shoutwiki.com/wiki/Power_Flow_User_Guide
[33]����� Pacific Northwest National Laboratory, �GridLAB-D
basics,� October 13, 2012. [Online]. Available: http://gridlab-d.shoutwiki.com/wiki/GridLAB-D_basics
[34]����� Pacific Northwest National Laboratory, �GLM
syntax,� GridLAB-D Wiki, September 25, 2011. [Online]. Available:
http://gridlab-d.shoutwiki.com/wiki/GLM_syntax
[35]����� Pacific Northwest National Laboratory, �Schedule,�
October 05, 2014. [Online]. Available: http://gridlab-d.shoutwiki.com/wiki/Schedule
[36]����� Sk4ld, Gridpot. GitHub, 2015.
[37]����� R. Fielding et al., �Hypertext transfer protocol
-- HTTP/1.1.� The Internet Society, 1999.
[38]����� P. Ackerman, Industrial Cybersecurity.
Birmingham, AL, USA: Packt Publishing, 2017.
[39]����� �Modbus application protocol specification
v1.1b3.� Modbus, 2012.
[40]����� �Modbus messaging on TCP/IP implementation guide
v1.0b.� Modbus-IDA, 2006.
[41]����� T. Wiens, �S7 communication (s7comm),� Wireshark,
May 13, 2016. [Online]. Available: https://wiki.wireshark.org/S7comm
[42]����� G. Miru, �The Siemens S7 communication - part 1
general structure,� GyM�s Personal Blog, April 08, 2018. [Online].
Available: http://gmiru.com/article/s7comm/
[43]����� J. Case, M. Fedor, M. Schoffstall, and J. Davin,
�A simple network management protocol (SNMP).� Network Working Group, 1990.
[44]����� R. Latchmepersad and G. Singh, CompTIA Network+
Certification Guide. Birmingham, AL, USA: Packt Publishing, 2018.
[45]����� P. Matousek, �Description of IEC 61850
communication,� Brno University of Technology, Brno, Czech Republic, Technical
FIT-TR-2018-01, 2018.
[46]����� W. O. Redwood, �Cyber physical system
vulnerability research,� Ph.D. dissertation, Dept. of Comp. Sci., Florida State
Univ., Tallahassee, FL, USA, 2016.
[47]����� A. Jicha, M. Patton, and H. Chen, �SCADA
honeypots: An in-depth analysis of Conpot,� in Proc. of the 2016 IEEE Conf.
on Intel. and Sec. Inf. (ISI), Tucson, AZ, USA, 2016, pp. 196�198.
[48]����� D. Hyun, �Collecting cyberattack data for
industrial control systems using honeypots,� M.S. thesis, Dept. of Comp. Sci.,
NPS, Monterey, CA, USA, 2018.
[49]����� G. Combs, �About Wireshark.� Accessed March 26,
2019. [Online]. Available: https://www.wireshark.org/
[50]����� G. Lyon, �Nmap: the network mapper - free security
scanner.� Accessed March 26, 2019. [Online]. Available: https://nmap.org/
[51]����� �Nessus Professional,� Tenable�, February
04, 2015. [Online]. Available: https://www.tenable.com/products/nessus/nessus-professional
[52]����� �Metasploit | penetration testing software, pen
testing security,� Rapid7. Accessed March 26, 2019. [Online]. Available:
https://www.metasploit.com/
[53]����� F. Baumgarten, �Netstat(8) - Linux man page.� Accessed
March 26, 2019. [Online]. Available: https://linux.die.net/man/8/netstat
[54]����� �Shodan.� Accessed March 26, 2019. [Online]. Available:
https://www.shodan.io/
[55]����� �API Reference � dpkt 1.9.2 documentation.� Accessed
May 13, 2019. [Online]. Available: https://dpkt.readthedocs.io/en/latest/api/index.html
[56]����� �GeoLite2 Free Downloadable Databases
� MaxMind Developer Site.� Accessed May 14, 2019. [Online]. Available:
https://dev.maxmind.com/geoip/geoip2/geolite2/
[57]����� Deutsche Telekom AG Honeypot Project, �T-pot: a
multi-honeypot platform.� Accessed May 06, 2019. [Online]. Available:
https://dtag-dev-sec.github.io/mediator/feature/2015/03/17/concept.html
[58]����� The
Python Software Foundation, �13.2. ConfigParser � configuration file parser,�
March 18, 2019. [Online]. Available:
https://docs.python.org/2.7/library/configparser.html?highlight=configuration
