250 likes | 509 Views
Outline. TerminologyClassificationGeneral ModelComplicationsSome Current SystemsResources. Terminology. What is an intrusion?Examples includeBreaching a locked doorPassword guessingEscalation of privilegeDenial of Service attacksRoot kits, worms, viruses, etc.Working definition: A violation, or attempted violation, of a security policy.
E N D
1. An Introduction to Intrusion Detection Dr. James P. Early
Department of
Computer Sciences
2. Outline Terminology
Classification
General Model
Complications
Some Current Systems
Resources
3. Terminology What is an intrusion?
Examples include
Breaching a locked door
Password guessing
Escalation of privilege
Denial of Service attacks
Root kits, worms, viruses, etc.
Working definition: A violation, or attempted violation, of a security policy
4. Terminology What is intrusion detection?
Def: The process of observing events for evidence of security policy violations or attempts
Some important operational questions
What is observed?
How is a security violation determined?
What is the response?
5. Classifications Based on operational behavior
Host vs. Network
Defines the event source
Signature vs. Anomaly-based
Defines model for assessing policy violations
Active vs. Passive
Probing versus monitoring
6. Classifications Centralized vs. Distributed
Location of analysis
Real Time vs. Interval
Determines when notification takes place
7. Host-based IDS Data sources commonly found on a host
Examples:
System call sequences
Audit logs files
File attributes
May be integral to OS (e.g., Solaris BSM)
Lacks “big picture”; correlation
8. Network-based IDS Observes packets and connections
Visibility to potentially many hosts
Identify violations before it impacts the intended victim (hopefully)
Can be difficult to assess impact on individual hosts (more on this later)
9. Signature-based IDS Use models of “bad” behavior
Each “signature” is an observed policy violation
Examples: Buffer overflow strings, SQL injection attacks, virus definitions
Detection occurs when bad behavior is observed
List of signatures must be kept current
10. Anomaly-based IDS Uses model of “good” behavior
Detection occurs when observed behavior deviates from “good” behavior
Useful for detecting novel attacks
May generate excessive false positives
11. Active vs. Passive IDS Active IDS
Probe systems to uncover attack artifacts
May take corrective/preventive action
Lockout a user ID
Terminate a network connection and update a firewall rule
Passive IDS
Monitor (do not alter) event stream
Alert the user; user responsible for response
12. Centralized vs. Distributed Centralized
Monitoring, analysis, and detection are performed by a single system
Can we keep up with the event stream?
Distributed
Many monitoring points or agents contribute to the process
How do we communicate securely among entities?
13. Real Time vs. Interval Real Time
Detection and response occur before intrusion can take place (hopefully)
Necessary for autonomous response
Interval
Analysis and detection are reported over some time interval (e.g., once per day)
User is responsible for response
14. Let’s build an IDS! Assume a general model (not specific to host or network)
Assume we will use both signature and anomaly-based detection
What functional components do we need?
15. General Model
16. Important Dates in IDS History 1980
Anderson introduces idea of anomaly detection based on accounting logs
1986
D. Denning (Purdue grad!) formalizes definition of anomaly detection IDS (IDES)
1988
Morris Worm released. Spafford (Purdue prof!) discovers immunization process
17. Important Dates in IDS History 1992
G. Kim (Purdue undergrad!) and Spafford create Tripwire
1994
S. Forrest, et. al propose IDS based on system call sequences
1998
W. Lee uses data mining to build anomaly profiles
18. Complications Feature Extraction
Evasion
Performance
Confidence with detection
19. Feature Extraction What do we measure in order to identify intrusions?
A fundamental question originally posed by D. Denning
Much current research still focuses on this question
20. Evasion Techniques Techniques employed by an attacker with knowledge of the system
Designed to render IDS ineffective
Examples
Inject/drop/alter events
Flood system with events
Cause misclassification of events
21. Evasion Example Described by Ptacek and Newsham (1997)
Manipulate IP TTL field to cause packet drops
22. Performance Issues Are models current/relevant?
Keeping up with events
False positives
False negatives
Vulnerability of the IDS
Fail open versus fail closed
23. Confidence in Detection How confident are we?
Given confidence, what is an appropriate response?
Can attacker exploit response to create a denial of service attack?
24. Some Current Systems SNORT
Open source signature-based network intrusion detection
Also has SPADE, an anomaly detection component
Bro
Network IDS with a specification language
25. Resources “Intrusion Detection”, Rebecca Bace, Macmillan Technical Publishing, 2000
Intrusion Detection FAQ, SANS Institute, http://www.sans.org/resources/idfaq/
Distributed Intrusion Detection System, http://www.dshield.com/
26. Thank you! Any questions?