Systems and Internet Infrastructure Security
Network and Security Research Center
Department of Computer Science and Engineering Pennsylvania State University, University Park PA
SCADA Security Measures
Paper 1:
The VIKING Project - Towards more
Secure SCADA Systems
Written by: Gunnar Björkman Presented by: Diana Koshy
Type of Paper
• Expository
• This paper discusses a future project aimed at
analyzing the security of SCADA systems.
The Problem:
Security on SCADA systems needs to
be improved…
The Problem
• SCADA systems need to be secure since a
problem with the system has dire consequences
• Security is non-trivial since the systems are very
complex and must perform under strict conditions
The Problem
• Risks come from insiders as well as new access points
opened by connecting the SCADA system to corporate networks, engineers, contractors, vendors, etc.
These risks have been somewhat mitigated by firewalls and Demilitarized Zones (DMZs)
• Risks also come from use of standardized protocols,
hardware and software
Communication protocols are becoming more standardized to allow different hardware to
The Solution
“The objective of the VIKING project is to develop, test and evaluate methodologies for the analysis, design and operation of resilient and secure
Background
Structure of a SCADA System
• Sensors
• Remote Terminal Units (RTUs) • Station Control Systems
• Central Control System
• Workstations
• Front-End Servers • SCADA Servers • Archive Servers
The Solution
• The VIKING project aims to take a holistic
The Solution: VIKING Goals
1. Assess security risk and (financial) consequences of an
attack on a SCADA system
2. Create a tool that can quantify security for comparison
across different systems
3. Use model-based system as IDS
4. Secure power system communication
5. Be able to identify vulnerable spots in a SCADA system
6. Create a system that can be used to test security solutions
The Solution: Method
Create 3 models:
1. power system model
- used to model the effects of an attack on electricity supply
2. society model
- used to gauge economic consequences of an attack
3. SCADA system models
(architectural and cyber-physical)
The Assumptions
None
The paper was just summarizing a proposed project.
Paper 2:
21 Steps to Improve Cyber Security of
SCADA Networks
Written by: US Department of Energy Presented by: Diana Koshy
Type of Paper
• Best-practices paper
• This paper proposes 21-steps to take in order to
alleviate the security problem inherent in current SCADA systems
The Problem
• SCADA systems were not designed with security
in mind
• Organizations using SCADA networks need to
The Solution
2 Categories:
1. Actions to Take to Increase Security
2. Management Actions to Establish Effective
The Solution: Actions to Take
1. Understand the risk, protection and necessity of
every connection to the SCADA network
2. Make the network as isolated as possible and
use safe methods for data transfer
3. Analyze and implement a strong security
strategy for all remaining connections
4. Remove or disable unused services provided by
The Solution: Actions to Take
5. Proprietary (obscure) protocols should not be
mistaken for secure protocols
6. Enable and configure all security features
already present and/or demand upgrades
7. Secure backdoors and vendor connections
8. Monitor for internal and external intrusions
The Solution: Actions to Take
9. Conduct audits of the system to find common
vulnerabilities
10.Check physical security of all remote sites that
communicate with the SCADA system
11.Put together a “Red Team” to come up with
The Solution: Management
12.Clearly define roles and responsibilities for all
organization personnel
13.Document the information security architecture
and its components
14.Identify risks and vulnerabilities and create an
ongoing risk management process
15.Base protection strategy on defense-in-depth
The Solution: Management
16.Create a clear, structured security program with
delineated requirements
17.Establish configuration management processes 18.Conduct routine self-assessments
19.Create system backups and disaster recovery
The Solution: Management
20.Establish an expectation for strong security for
all levels of personnel
21.Train personnel to prevent disclosure of
The Assumptions
None
Paper 3:
SCADA-specific Intrusion
Detection/Prevention Systems: A
Survey and Taxonomy
Written by: Bonnie Zhu and Shankar Sastry Presented by: Diana Koshy
Type of Paper
• Survey paper
• This paper discusses past work on
Classification and characteristics of attacks SCADA-specific IDS attempts
The Problem
• SCADA systems are vulnerable
Standardized protocols, software and hardware De-isolation of SCADA systems
The Problem
• Specific Vulnerabilities Listed:
HMI controller: Can falsify what operator sees sensor-HMI link: Can spy on what operator sees
actuator-controller link: Can see what actuators are told to do
sensor threshold values and settings: Can modify settings
The Problem
• Security research on SCADA systems is lacking Unrealistic testing environments
Poorly analyzed threat models
IDS implementations specific to different SCADA environments
Lack of analysis of false positives/false negatives of IDSs
The Problem
• 100% prevention of attacks is impossible Must combine prevention with detection
• Can’t use existing IDSs since SCADA is different It is a hard real-time system, which means
timeliness, freshness of data, and availability are
crucial
Its terminal devices have limited computing and memory resources
The Solution
• Create SCADA-specific IDS and security metrics • Ideal system should be able to:
detect and block intrusions in real time do so without interrupting performance
do so without extra burdens due to false positives do so despite normal noise
The Solution
Types of IDS:
• signature detection approach • anomaly detection approach • probabilistic approach
• specification-based approach • behavioral detection approach
The Solution
All of these can be applied to different parts of SCADA systems
The Solution: Past Work
Model-Based IDS for SCADA Using Modbus/TCP
• Uses the fact that network traffic on a SCADA
system is relatively constant to find anomalies
The Solution: Past Work
Anomaly-Based IDS
1. AutoAssociative Kernel Regression and
Statistical Probability Ratio Test
- monitor anomalous non-malicious activity to establish baseline - use baseline database to compare with new activity
2. Multi-Agent IDS Using Ant Clustering Approach
and Unsupervised Feature Extraction
-use multiple intelligent agents to perform IDS duties
-monitor agents capture packets, extract features and perform PCA
-decision agents perform clustering and notify of abnormalities -action agents respond to threats accordingly
The Solution: Past Work
Configurable Embedded Middleware-Level Detection
• put a detection system in the middle of the
communication channels
kind of like a firewall
• easiest to incorporate since few changes to
The Solution: Past Work
Intrusion Detection and Event Monitoring in SCADA Networks
• specific to SCADA power-grid and RTUs • automatically produce signatures for
unauthorized access
• store settings and details of each SCADA device
The Solution: Past Work
Model for Cyber-Physical Interaction
1. Power Plant interfacing Substations through
Probabilistic validation of attack-effect bindings
2. Workflow-based non-intrusive approach for
enhancing the survivability of
The Solution: Past Work
Model for Cyber-Physical Interaction
1. Power Plant interfacing Substations through
Probabilistic validation of attack-effect bindings
probabilistically build a profile of legitimate data flows and main characteristics of normal information
exchange
The Solution: Past Work
Model for Cyber-Physical Interaction
2. Workflow-based non-intrusive approach for
enhancing the survivability of
critical infrastructures in Cyber Environment
separate SCADA system into cyber, physical, and workflow layers
each physical component is a node in workflow layer model functionality and attack patterns
The Solution: Past Work
Modeling Flow Information and other Control Systems Behavior To Detect Anomalies
• analyzes flow on the network (so only good for
network layers)
• combine anomaly-, behavioral-, and
specification-based techniques to detect abnormal behavior
The Solution: Past Work
SHARP
• uses authentication and privilege escalation
protection to detect and block
The Assumptions
None