TRUSTWORTHY CYBER INFRASTRUCTURE FOR THE POWER GRID | TCIPG.ORG

UNIVERSITY OF ILLINOIS | DARTMOUTH COLLEGE | UC DAVIS | WASHINGTON STATE UNIVERSITY

FUNDING SUPPORT PROVIDED BY DOE-OE AND DHS S&T

GOALS

• Design is “physics-aware” Network Intrusion Detection System

(NIDS) for process control.

• Integrate NIDS cyber-physical state analytics within the process

data historian in EMS.

• Control environments include physical systems, switches, and control

programs.

• CONTROL: receive data from field devices process decide

issue switching commands.

• The combination of the safe operations of the protective schemes and

the physical assets can be described by a Hybrid Automaton model.

• Basic question: Can we use such models as the baseline for “safe”

behavior and use any set of message and command that is

inconsistent with that as the indication of an attack/anomaly?

FUNDAMENTAL QUESTIONS/CHALLENGES

• Validation of Hybrid Control NIDS (HC-NIDS).

– We developed an experimental framework to test HC-NIDS that

combines simulated physical and control environments interacting

with actual logic controllers (Siemens PLC using Modbus TCP).

• Integration with Data Management Services (OSIsoft case study).

– We are collaborating with OSIsoft, one of the industry leaders in

ICS data management systems, to implement inclusion of sensor

tags for appropriately located network taps.

– HC-NIDS rules are then implemented as analytics/queries of the

OSIsoft database.

RESEARCH PLAN

BROADER IMPACT

• Operators are made aware of Cyber-Physical State.

FUTURE EFFORTS

• Blind HC-NIDS: Learn the rules by analyzing traffic.

• Integrate OSIsoft with Wireshark so that it can leverage the extensive

literature.

Cyber-Physical Data Analytics Based on “Hybrid Control”

Network Intrusion DetectionGeorgia Koutsandria, Masood Parvania, Reinhard Gentz, Mehdi Jamei, Vishak Muthukumar

Researchers: Masood Parvania, Sean Peisert, Chuck McParland, and Anna Scaglione

Functional Security Enhancements for Existing SCADA Systems

INTERACTION WITH OTHER PROJECTS

• TCIPG Specification-based IDS for the DNP3 Protocol.

• CEDS project with Lawrence Berkeley National Lab (LBNL).

• Design methodology for Hybrid Control NIDS (HC-NIDS).

– Each hybrid state corresponds to specific values for the switches

and specific ranges for the current, voltage, temperature, etc.

– Transitions between hybrid states are triggered by physical changes

and commands.

– Network packets, flowing between field devices and central

controllers, should only produce “allowed” transitions and “allowed”

hybrid states.

– HC-NIDS continuously monitors and analyzes the network traffic

exchanged by field devices that are used to activate the protection

schemes.

– HC-NIDS Rule generation commands and information

exchanged must be consistent with the protection hybrid automaton

model.

1. Simulink model:

simulation of the

physical

application.

2. C MEX S-function:

allows

communication

through the

Modbus TCP

protocol.

3. Emulation of the

protection function

in Ladder logic.

Validation Testbed

Example: Overcurrent Protection

q0

CB1 = 0CB2 = 0

I < I p

M = 1

N < 1

q1

CB1 = 0CB2 = 0

I < I p

M = 0S = 1

N < 2

N = 1

q3

N = 2

CB1 = 0CB2 = 0

I < I p

M = 1S = 0S = 0

N < 3

CB1 = 0CB2 = 0

I < I pM = 0S = 1

q5

M = 1

N < 1

M = 0S = 1

N < 2

N = 1 N = 2

M = 1S = 0S = 0

N < 3

M = 0S = 1

q6 q7

q2

N < 3& I < I p

CB1 = 1 CB1 = 1 CB1 = 1 CB1 = 1CB2 = 1CB2 = 1CB2 = 1CB2 = 1

I = 0 I = 0 I = 0 I = 0

N < 3& I = 0q4

N < 4

N < 4

N = 3& I ≥ I p

N = 3& I < I p

N = 3& I = 0

N = 3&0 < I < I p

N = 4/ r eset

N = 4/ r eset

Simulink model Hybrid Automaton

."."." ."."."

Injected network t raffic

Normal network trafficNormal network

t rafficNormal network

t rafficNormal network

t raffic

• The different data items of the

different controllers have different

colors.

• Source ID/Destination ID,

function code, register, and

value range (set) are different.

• The normal sequence is green-

light green-blue-turquoise.

• Red packets are not part of it, so

they are anomalies.

Arrows indicate phenomena that

can be identified as attacks, since

the switches’ state (CB) and

current are not in the right

combination.

Cyber-Physical Analytics

Network tap

HC-NIDS

values

Cyber-Physical

Process Control

Data Analytics

Physical values

Historian

RESEARCH RESULTS (CON’T)

RESEARCH RESULTS