SlideShare a Scribd company logo
34	 IT Pro July/August 2013	 P u b l i s h e d by t h e I E E E C o m p u t e r S o c i e t y	 1520-9202/13/$31.00 © 2013 IEEE
Jeffery Payne, Mark Fenner, and Jonathan Kauffman, Coveros
The authors use system event monitoring to distinguish between the
behavioral characteristics of normal and anomalous computer system
users. Identifying anomalous behavior at the system event level
diminishes privacy concerns and supports the identification of
cross-application behavioral patterns.
A
nomaly detection with system events
is used in a variety of intrusion-
detection systems.1–4,8
Our project
for DARPA’s Active Authentication
program uses system event monitoring to dis-
tinguish between the behavioral characteristics
of normal and anomalous computer system us-
ers. Identifying anomalous behavior in system
events has advantages over doing so at other lev-
els. First, there are fewer privacy concerns with
the information collected, because system event
streams don’t include the acted-on content. Sec-
ond, information over all applications and pro-
cesses is available to identify cross-application
behavioral patterns.
Our Approach
We perform active authentication by detecting
anomalies in a stream of operating system events
generated by a computer user. Figure 1 shows
how our system operates.
The process in Figure 1a begins when a comput-
er user generates system events as he or she per-
forms application-level operations. These events
are captured using either Microsoft’s Procmon or
Detours utilities. Procmon monitors a variety of
system events, including file system events, reg-
istry access, network events, and process/thread
activity. Detours can intercept Windows system
events individually. System events are filtered
based on their process type and event class. Initial
System Event
Monitoring
for Active
Authentication
Security: DARPA
computer.org/ITPro 	 35
experimentation led us to remove
system events from background
tasks. We filter events by class to
keep similar events together so
that detectors can evaluate a single
type of event.
Once filtered, system events
are analyzed using one or more
anomaly-detection learning sys-
tems—including one-class sup-
port vector machine (1SVM),
replicator neural network (RNN),
probabilistic similarity (PS), and
similarity matching. Anomaly-
detection learning systems5
dif-
fer from standard classification
learning systems, because they’re
trained on instances of a single
class and then determine wheth-
er new instances are members of
that original class. 1SVM is a spe-
cific implementation of SVMs to
handle anomaly detection.6
RNNs
use backpropagation training cou-
pled with a specific architecture
to reconstruct input examples as
output.7
The error in reconstruc-
tion is interpreted as an anomaly
score. PS, based on an approach
first proposed by Salvatore Stolfo
and his colleagues,7
is an anomaly
detector that identifies anomalies
based on the marginal distribu-
tions of tuples in a training set of
features. Similarity matching is a
baseline algorithm that matches
system event streams with previ-
ously seen streams.
In our approach, we break the
raw system event stream into
fixed-time chunks called blocks.
Each block provides a sequence
of overlapping subsequences, called windows, of
system events. The learning algorithms are applied
directly to these windows. Each window is labeled
normal or anomalous. We then compare the percent
of anomalous windows in the larger block against
a threshold and render an accept or reject decision
for that block. The results of multiple detectors are
combined using Boolean or counting operations.
At both the window and block levels, we can apply
a leaky bucket algorithm to smooth the detection
stream. When the system recognizes an anomaly,
it either sends an alert to a centralized monitoring
system or requires a manual re-authentication.
Figure 1b shows how we evaluated the accu-
racy of our anomaly-detection approach. After
performing training on a training block, we fed
our system both authorized user data (additional
system events from the same training user) and
Figure 1. A high-level depiction of the system-event-monitoring
approach: the (a) system-event-monitoring process and (b) approach for
determining false accept and false reject rates.
A computer user generates system calls. The system calls are captured and filtered.
Then the calls are analyzed
Procmon/Detours
Subsets
noise
Multilayer
perceptron
or support
vector
machine
Noise
reduction
rating
identity
Leaky bucket
thresholds
voting
Login prompt
centralized monitor
Authorized user Unauthorized user
Training block Self-test block Other-test block
Windows
Blocks
Streams
Anomaly
detector
(learn)
Anomaly
detector
(detect)
Anomaly
detector
(detect)
Threshold Compare number of
anomalous windows
to threshold
Compare number of
anomalous windows
to threshold
False reject Accept Reject False accept
(a)
(b)
to make a decision whether to re-authenitcate.
36	 IT Pro July/August 2013
Security: DARPA
unauthorized user data (system events from a dif-
ferent user). If a block from the training user was
rejected, the result was a false reject. If a block from
a different user was accepted, the result was a false
accept. Otherwise, the result was true (correct).
Empirical Results
To identify a set of system parameters that ex-
hibits good performance for a variety of users,
we developed an experimentation framework in
Python that evaluates various sets of parameters
for event streams produced by multiple users. To
collect these streams, we deployed Procmon on
seven test subjects’ workstations and collected
two hours of system events during their normal,
daily computer usage. Their typical tasks includ-
ed communication, software development and
configuration, and document creation.
Each participant produced an event stream that
was 2–4 Gbytes in size and comprised 3–8 million
records. We split each of these event streams into
two sets. We used the first set to tune the system
parameters. The experimentation framework uses
this first set as normal data for a particular partici-
pant while using the second set of each participant’s
event stream to represent anomalous behavior.
During initial experimentation, we explored
a variety of parameter settings to measure their
­effect on system performance. For most of the
system parameters, we didn’t observe a significant
change in performance as we varied their values.
However, using the registry subset of events, as
well as enhancing event representation to include
both the elapsed time since the previous event and
the name of the process that invoked the system
event, noticeably improved system performance.
Figure 2 presents our best results using a detec-
tion error trade-off (DET) curve for both individ-
ual users and all users as a whole. We produced
the graph by applying cross-validation to the first
dataset. DET curves show how the performance
of a single system changes as the threshold for
a block of anomalies varies. Each point on the
graph represents system performance, both in
terms of the false reject rate (FRR) and false ac-
cept rate (FAR) at a particular threshold value.
The unconnected points of the same shape show
the DET curve for each individual user, while the
thick line shows the DET curve for all users. To
quantify a DET curve, we measure its area under
the curve (AUC). A DET curve that’s pushed to-
ward the origin will have a low AUC measure, in-
dicating that the system is more accurate.
Our best results were obtained by aggregating
the verdicts produced by two separate detectors.
We tried several methods of combining verdicts and
found that a method based on our relative trust of
the two detectors produced the best results. In this
approach, one detector more reliably accepts an au-
thorized user, and another more reliably rejects an
unauthorized user. When one detector produces a
trusted verdict and the other produces an untrusted
verdict, we use the trusted verdict as the aggregate.
In the case where both detectors produce a
trusted verdict, we choose the verdict from the
detector we trust (a priori) the most. The detec-
tor we trust the most in this scenario is the one
that gave the most accurate verdicts during indi-
vidual testing of each detector. When both detec-
tors produce an untrusted verdict, we declare the
result as unknown. Depending on the threshold
value, this means that anywhere from 0–20 per-
cent of the testing results were removed. By con-
sidering the curves for individual users, we reach
an FRR of 0 percent and a FAR of approximately
10–15 percent for several users.
O
ur future research will focus on improv-
ing these results when applied to differ-
ent types of users. Experimentation will
examine the degree to which normal behavior
can be distinguished from malicious behavior as
opposed to simply normal versus abnormal use.
Figure 2. Results from multisensor analysis. The
unconnected points of the same shape show the
detection error trade-off (DET) curve for each
individual user, while the thick line shows the
DET curve for all users.
100
80
60
40
20
0
0 20 40 60 80
FRR
FAR
100
computer.org/ITPro 	 37
We’re currently collecting additional field test
data so we can compare it to malicious use-case
scenarios developed from real-world attacks. We
also plan on researching mechanisms for dynam-
ically modifying threshold values on the basis of
system performance.
References
	1.	 A.K. Ghosh, J. Wanken, and F. Charron, “Detect-
ing Anomalous and Unknown Intrusions Against
Programs,” Proc. Computer Security Applications Conf.,
IEEE, 1999, pp. 259–267.
	 2.	 A.K. Ghosh, A. Schwartzbard, and M. Schatz, “Using
Program Behavior Profiles for Intrusion Detection,”
Proc. SANS Conf. and Workshop on Intrusion Detection and
Response, 1999.
	3.	A.K. Ghosh, A. Schwartzbard, and M. Schatz,
“Learning Program Behavior Profiles for Intrusion
Detection,” Proc. 1st Workshop on Intrusion Detection and
Network Monitoring (ID 99), Usenix, 1999, p. 6.
	4.	 V. Chandola, A. Banerjee, and V. Kumar. “Anomaly
Detection: A Survey,” ACM Computing Surveys, vol. 41,
no. 3, 2009, article 15.
	5.	 S. Hawkins et al., “Outlier Detection Using Replica-
tor Neural Networks,” Proc. 4th Int’l Conf. Data Ware-
housing and Knowledge Discovery (DaWaK 00), Springer,
2002, pp. 170–180.
	6.	 B. Scholkopf et al., “Estimating the Support of a
High-Dimensional Distribution,” Neural Computation,
vol. 13, no. 7, 2001, pp. 1443–1471.
	 7.	 S. Stolfo et al., “A Comparative Evaluation of Two Al-
gorithms for Windows Registry Anomaly Detection,”
J. Computer Security, vol. 13, no. 4, 2005, pp. 659–693.
	 8.	 S. Forrest et al.,“A Sense of Self for Unix Processes,” Proc.
IEEE Symp. Security and Privacy, IEEE, 1996, pp. 120–128.
Jeffery Payne is CEO and founder of Coveros and is a princi-
pal investigator on an Active Authentication project ­sponsored
by DARPA. Contact him at jeff.payne@coveros.com.
Mark E. Fenner is a research scientist at Coveros. Contact
him at mark.fenner@coveros.com.
Jonathan Kauffman is a consultant at Coveros, where
he’s performing research in Active Authentication. Contact
him at jonathan.kauffman@coveros.com.
handles the details
so you don’t have to!
Professional management and production of your publication
Inclusion into the IEEE Xplore and CSDL Digital Libraries
Access to CPS Online: Our Online Collaborative Publishing System
Choose the product media type that works for your conference:
Books, CDs/DVDs, USB Flash Drives, SD Cards, and Web-only delivery!
www.computer.org/cps or cps@computer.org
Contact CPS for a Quote Today!

More Related Content

What's hot

Code Quality - Security
Code Quality - SecurityCode Quality - Security
Code Quality - Security
sedukull
 
You Build It, You Secure It: Introduction to DevSecOps
You Build It, You Secure It: Introduction to DevSecOpsYou Build It, You Secure It: Introduction to DevSecOps
You Build It, You Secure It: Introduction to DevSecOps
Sumo Logic
 
DevSecOps: What Why and How : Blackhat 2019
DevSecOps: What Why and How : Blackhat 2019DevSecOps: What Why and How : Blackhat 2019
DevSecOps: What Why and How : Blackhat 2019
NotSoSecure Global Services
 
Tests your pipeline might be missing
Tests your pipeline might be missingTests your pipeline might be missing
Tests your pipeline might be missing
Gene Gotimer
 
Connect Ops and Security with Flexible Web App and API Protection
Connect Ops and Security with Flexible Web App and API ProtectionConnect Ops and Security with Flexible Web App and API Protection
Connect Ops and Security with Flexible Web App and API Protection
DevOps.com
 
Sec4dev 2021 - Catch Me If You can : Continuous Delivery vs. Security Assurance
Sec4dev 2021  - Catch Me If You can : Continuous Delivery vs. Security AssuranceSec4dev 2021  - Catch Me If You can : Continuous Delivery vs. Security Assurance
Sec4dev 2021 - Catch Me If You can : Continuous Delivery vs. Security Assurance
Abdessamad TEMMAR
 
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
Christian Schneider
 
Rapid software testing and conformance with static code analysis
Rapid software testing and conformance with static code analysisRapid software testing and conformance with static code analysis
Rapid software testing and conformance with static code analysis
Rogue Wave Software
 
Building Security in Using CI
Building Security in Using CIBuilding Security in Using CI
Building Security in Using CI
Coveros, Inc.
 
Integrating security into Continuous Delivery
Integrating security into Continuous DeliveryIntegrating security into Continuous Delivery
Integrating security into Continuous Delivery
Tom Stiehm
 
The Future of Security and Productivity in Our Newly Remote World
The Future of Security and Productivity in Our Newly Remote WorldThe Future of Security and Productivity in Our Newly Remote World
The Future of Security and Productivity in Our Newly Remote World
DevOps.com
 
DevSecOps: Taking a DevOps Approach to Security
DevSecOps: Taking a DevOps Approach to SecurityDevSecOps: Taking a DevOps Approach to Security
DevSecOps: Taking a DevOps Approach to Security
Alert Logic
 
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly DavidoffDevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
DevSecCon
 
Devops security-An Insight into Secure-SDLC
Devops security-An Insight into Secure-SDLCDevops security-An Insight into Secure-SDLC
Devops security-An Insight into Secure-SDLC
Suman Sourav
 
Test parallelization using Jenkins
Test parallelization using JenkinsTest parallelization using Jenkins
Test parallelization using Jenkins
Rogue Wave Software
 
How to adapt the SDLC to the era of DevSecOps
How to adapt the SDLC to the era of DevSecOpsHow to adapt the SDLC to the era of DevSecOps
How to adapt the SDLC to the era of DevSecOps
Zane Lackey
 
Proactive Security AppSec Case Study
Proactive Security AppSec Case StudyProactive Security AppSec Case Study
Proactive Security AppSec Case Study
Andy Hoernecke
 
The Joy of Proactive Security
The Joy of Proactive SecurityThe Joy of Proactive Security
The Joy of Proactive Security
Andy Hoernecke
 
SecDevOps: The New Black of IT
SecDevOps: The New Black of ITSecDevOps: The New Black of IT
SecDevOps: The New Black of IT
CloudPassage
 

What's hot (19)

Code Quality - Security
Code Quality - SecurityCode Quality - Security
Code Quality - Security
 
You Build It, You Secure It: Introduction to DevSecOps
You Build It, You Secure It: Introduction to DevSecOpsYou Build It, You Secure It: Introduction to DevSecOps
You Build It, You Secure It: Introduction to DevSecOps
 
DevSecOps: What Why and How : Blackhat 2019
DevSecOps: What Why and How : Blackhat 2019DevSecOps: What Why and How : Blackhat 2019
DevSecOps: What Why and How : Blackhat 2019
 
Tests your pipeline might be missing
Tests your pipeline might be missingTests your pipeline might be missing
Tests your pipeline might be missing
 
Connect Ops and Security with Flexible Web App and API Protection
Connect Ops and Security with Flexible Web App and API ProtectionConnect Ops and Security with Flexible Web App and API Protection
Connect Ops and Security with Flexible Web App and API Protection
 
Sec4dev 2021 - Catch Me If You can : Continuous Delivery vs. Security Assurance
Sec4dev 2021  - Catch Me If You can : Continuous Delivery vs. Security AssuranceSec4dev 2021  - Catch Me If You can : Continuous Delivery vs. Security Assurance
Sec4dev 2021 - Catch Me If You can : Continuous Delivery vs. Security Assurance
 
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
Security DevOps - Staying secure in agile projects // OWASP AppSecEU 2015 - A...
 
Rapid software testing and conformance with static code analysis
Rapid software testing and conformance with static code analysisRapid software testing and conformance with static code analysis
Rapid software testing and conformance with static code analysis
 
Building Security in Using CI
Building Security in Using CIBuilding Security in Using CI
Building Security in Using CI
 
Integrating security into Continuous Delivery
Integrating security into Continuous DeliveryIntegrating security into Continuous Delivery
Integrating security into Continuous Delivery
 
The Future of Security and Productivity in Our Newly Remote World
The Future of Security and Productivity in Our Newly Remote WorldThe Future of Security and Productivity in Our Newly Remote World
The Future of Security and Productivity in Our Newly Remote World
 
DevSecOps: Taking a DevOps Approach to Security
DevSecOps: Taking a DevOps Approach to SecurityDevSecOps: Taking a DevOps Approach to Security
DevSecOps: Taking a DevOps Approach to Security
 
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly DavidoffDevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
DevSecCon Tel Aviv 2018 - End2End containers SSDLC by Vitaly Davidoff
 
Devops security-An Insight into Secure-SDLC
Devops security-An Insight into Secure-SDLCDevops security-An Insight into Secure-SDLC
Devops security-An Insight into Secure-SDLC
 
Test parallelization using Jenkins
Test parallelization using JenkinsTest parallelization using Jenkins
Test parallelization using Jenkins
 
How to adapt the SDLC to the era of DevSecOps
How to adapt the SDLC to the era of DevSecOpsHow to adapt the SDLC to the era of DevSecOps
How to adapt the SDLC to the era of DevSecOps
 
Proactive Security AppSec Case Study
Proactive Security AppSec Case StudyProactive Security AppSec Case Study
Proactive Security AppSec Case Study
 
The Joy of Proactive Security
The Joy of Proactive SecurityThe Joy of Proactive Security
The Joy of Proactive Security
 
SecDevOps: The New Black of IT
SecDevOps: The New Black of ITSecDevOps: The New Black of IT
SecDevOps: The New Black of IT
 

Similar to System Event Monitoring for Active Authentication

Ids 014 anomaly detection
Ids 014 anomaly detectionIds 014 anomaly detection
Ids 014 anomaly detection
jyoti_lakhani
 
Improving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approachImproving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approach
IJERA Editor
 
Ids 013 detection approaches
Ids 013 detection approachesIds 013 detection approaches
Ids 013 detection approaches
jyoti_lakhani
 
Intrusion detection system based on web usage mining
Intrusion detection system based on web usage miningIntrusion detection system based on web usage mining
Intrusion detection system based on web usage mining
IJCSEA Journal
 
Anomalous payload based network intrusion detection
Anomalous payload based network intrusion detectionAnomalous payload based network intrusion detection
Anomalous payload based network intrusion detectionUltraUploader
 
M41028892
M41028892M41028892
M41028892
IJERA Editor
 
A sense of self for unix processes
A sense of self for unix processesA sense of self for unix processes
A sense of self for unix processesUltraUploader
 
Building a usage profile for anomaly detection on computer networks
Building a usage profile for anomaly detection on computer networksBuilding a usage profile for anomaly detection on computer networks
Building a usage profile for anomaly detection on computer networks
Nathanael Asaam
 
Personam Solution - How it Works Brief
Personam Solution - How it Works BriefPersonam Solution - How it Works Brief
Personam Solution - How it Works BriefSunny Geo
 
Personam Solution - How it Works Brief
Personam Solution - How it Works BriefPersonam Solution - How it Works Brief
Personam Solution - How it Works BriefSunny Geo
 
P47 Eait06
P47 Eait06P47 Eait06
DB-OLS: An Approach for IDS1
DB-OLS: An Approach for IDS1DB-OLS: An Approach for IDS1
DB-OLS: An Approach for IDS1
IJITE
 
Application of genetic algorithm in intrusion detection system
Application of genetic algorithm in intrusion detection systemApplication of genetic algorithm in intrusion detection system
Application of genetic algorithm in intrusion detection system
Alexander Decker
 
IRJET- Genetic Algorithm based Intrusion Detection-Survey
IRJET- Genetic Algorithm based Intrusion Detection-SurveyIRJET- Genetic Algorithm based Intrusion Detection-Survey
IRJET- Genetic Algorithm based Intrusion Detection-Survey
IRJET Journal
 
Intrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern miningIntrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern mining
eSAT Journals
 
Intrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern miningIntrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern mining
eSAT Journals
 
Reveelium Technical Overview - Datasheet EN
Reveelium Technical Overview - Datasheet ENReveelium Technical Overview - Datasheet EN
Reveelium Technical Overview - Datasheet EN
ITrust - Cybersecurity as a Service
 
A Survey On Intrusion Detection Systems
A Survey On Intrusion Detection SystemsA Survey On Intrusion Detection Systems
A Survey On Intrusion Detection Systems
Mary Calkins
 
Automatically generated win32 heuristic virus detection
Automatically generated win32 heuristic virus detectionAutomatically generated win32 heuristic virus detection
Automatically generated win32 heuristic virus detectionUltraUploader
 

Similar to System Event Monitoring for Active Authentication (20)

Ids 014 anomaly detection
Ids 014 anomaly detectionIds 014 anomaly detection
Ids 014 anomaly detection
 
Improving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approachImproving the accuracy of fingerprinting system using multibiometric approach
Improving the accuracy of fingerprinting system using multibiometric approach
 
Ids 013 detection approaches
Ids 013 detection approachesIds 013 detection approaches
Ids 013 detection approaches
 
Intrusion detection system based on web usage mining
Intrusion detection system based on web usage miningIntrusion detection system based on web usage mining
Intrusion detection system based on web usage mining
 
Anomalous payload based network intrusion detection
Anomalous payload based network intrusion detectionAnomalous payload based network intrusion detection
Anomalous payload based network intrusion detection
 
M41028892
M41028892M41028892
M41028892
 
A sense of self for unix processes
A sense of self for unix processesA sense of self for unix processes
A sense of self for unix processes
 
Building a usage profile for anomaly detection on computer networks
Building a usage profile for anomaly detection on computer networksBuilding a usage profile for anomaly detection on computer networks
Building a usage profile for anomaly detection on computer networks
 
Personam Solution - How it Works Brief
Personam Solution - How it Works BriefPersonam Solution - How it Works Brief
Personam Solution - How it Works Brief
 
Personam Solution - How it Works Brief
Personam Solution - How it Works BriefPersonam Solution - How it Works Brief
Personam Solution - How it Works Brief
 
P47 Eait06
P47 Eait06P47 Eait06
P47 Eait06
 
WS97-07-013
WS97-07-013WS97-07-013
WS97-07-013
 
DB-OLS: An Approach for IDS1
DB-OLS: An Approach for IDS1DB-OLS: An Approach for IDS1
DB-OLS: An Approach for IDS1
 
Application of genetic algorithm in intrusion detection system
Application of genetic algorithm in intrusion detection systemApplication of genetic algorithm in intrusion detection system
Application of genetic algorithm in intrusion detection system
 
IRJET- Genetic Algorithm based Intrusion Detection-Survey
IRJET- Genetic Algorithm based Intrusion Detection-SurveyIRJET- Genetic Algorithm based Intrusion Detection-Survey
IRJET- Genetic Algorithm based Intrusion Detection-Survey
 
Intrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern miningIntrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern mining
 
Intrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern miningIntrusion detection and anomaly detection system using sequential pattern mining
Intrusion detection and anomaly detection system using sequential pattern mining
 
Reveelium Technical Overview - Datasheet EN
Reveelium Technical Overview - Datasheet ENReveelium Technical Overview - Datasheet EN
Reveelium Technical Overview - Datasheet EN
 
A Survey On Intrusion Detection Systems
A Survey On Intrusion Detection SystemsA Survey On Intrusion Detection Systems
A Survey On Intrusion Detection Systems
 
Automatically generated win32 heuristic virus detection
Automatically generated win32 heuristic virus detectionAutomatically generated win32 heuristic virus detection
Automatically generated win32 heuristic virus detection
 

More from Coveros, Inc.

Which Development Metrics Should I Watch?
Which Development Metrics Should I Watch?Which Development Metrics Should I Watch?
Which Development Metrics Should I Watch?
Coveros, Inc.
 
10 Things You Might Not Know: Continuous Integration
10 Things You Might Not Know: Continuous Integration10 Things You Might Not Know: Continuous Integration
10 Things You Might Not Know: Continuous Integration
Coveros, Inc.
 
Increasing Quality with DevOps
Increasing Quality with DevOpsIncreasing Quality with DevOps
Increasing Quality with DevOps
Coveros, Inc.
 
Better Security Testing: Using the Cloud and Continuous Delivery
Better Security Testing: Using the Cloud and Continuous Delivery Better Security Testing: Using the Cloud and Continuous Delivery
Better Security Testing: Using the Cloud and Continuous Delivery
Coveros, Inc.
 
Create Disposable Test Environments with Vagrant and Puppet
Create Disposable Test Environments with Vagrant and Puppet Create Disposable Test Environments with Vagrant and Puppet
Create Disposable Test Environments with Vagrant and Puppet
Coveros, Inc.
 
Continuous Delivery in a Legacy Shop - One Step at a Time
Continuous Delivery in a Legacy Shop - One Step at a Time Continuous Delivery in a Legacy Shop - One Step at a Time
Continuous Delivery in a Legacy Shop - One Step at a Time
Coveros, Inc.
 
DevOps in a Regulated and Embedded Environment (AgileDC)
DevOps in a Regulated and Embedded Environment (AgileDC) DevOps in a Regulated and Embedded Environment (AgileDC)
DevOps in a Regulated and Embedded Environment (AgileDC)
Coveros, Inc.
 
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
Coveros, Inc.
 
Tests Your Pipeline Might Be Missing
 Tests Your Pipeline Might Be Missing Tests Your Pipeline Might Be Missing
Tests Your Pipeline Might Be Missing
Coveros, Inc.
 
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
Coveros, Inc.
 
Web Application Security Testing: Kali Linux Is the Way to Go
Web Application Security Testing: Kali Linux Is the Way to Go Web Application Security Testing: Kali Linux Is the Way to Go
Web Application Security Testing: Kali Linux Is the Way to Go
Coveros, Inc.
 

More from Coveros, Inc. (11)

Which Development Metrics Should I Watch?
Which Development Metrics Should I Watch?Which Development Metrics Should I Watch?
Which Development Metrics Should I Watch?
 
10 Things You Might Not Know: Continuous Integration
10 Things You Might Not Know: Continuous Integration10 Things You Might Not Know: Continuous Integration
10 Things You Might Not Know: Continuous Integration
 
Increasing Quality with DevOps
Increasing Quality with DevOpsIncreasing Quality with DevOps
Increasing Quality with DevOps
 
Better Security Testing: Using the Cloud and Continuous Delivery
Better Security Testing: Using the Cloud and Continuous Delivery Better Security Testing: Using the Cloud and Continuous Delivery
Better Security Testing: Using the Cloud and Continuous Delivery
 
Create Disposable Test Environments with Vagrant and Puppet
Create Disposable Test Environments with Vagrant and Puppet Create Disposable Test Environments with Vagrant and Puppet
Create Disposable Test Environments with Vagrant and Puppet
 
Continuous Delivery in a Legacy Shop - One Step at a Time
Continuous Delivery in a Legacy Shop - One Step at a Time Continuous Delivery in a Legacy Shop - One Step at a Time
Continuous Delivery in a Legacy Shop - One Step at a Time
 
DevOps in a Regulated and Embedded Environment (AgileDC)
DevOps in a Regulated and Embedded Environment (AgileDC) DevOps in a Regulated and Embedded Environment (AgileDC)
DevOps in a Regulated and Embedded Environment (AgileDC)
 
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
Compatibility Testing of Your Web Apps - Tips and Tricks for Debugging Locall...
 
Tests Your Pipeline Might Be Missing
 Tests Your Pipeline Might Be Missing Tests Your Pipeline Might Be Missing
Tests Your Pipeline Might Be Missing
 
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
Testing in a Continuous Delivery Pipeline - Better, Faster, Cheaper
 
Web Application Security Testing: Kali Linux Is the Way to Go
Web Application Security Testing: Kali Linux Is the Way to Go Web Application Security Testing: Kali Linux Is the Way to Go
Web Application Security Testing: Kali Linux Is the Way to Go
 

Recently uploaded

Top Features to Include in Your Winzo Clone App for Business Growth (4).pptx
Top Features to Include in Your Winzo Clone App for Business Growth (4).pptxTop Features to Include in Your Winzo Clone App for Business Growth (4).pptx
Top Features to Include in Your Winzo Clone App for Business Growth (4).pptx
rickgrimesss22
 
Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604
Fermin Galan
 
OpenMetadata Community Meeting - 5th June 2024
OpenMetadata Community Meeting - 5th June 2024OpenMetadata Community Meeting - 5th June 2024
OpenMetadata Community Meeting - 5th June 2024
OpenMetadata
 
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing SuiteAI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
Google
 
A Study of Variable-Role-based Feature Enrichment in Neural Models of Code
A Study of Variable-Role-based Feature Enrichment in Neural Models of CodeA Study of Variable-Role-based Feature Enrichment in Neural Models of Code
A Study of Variable-Role-based Feature Enrichment in Neural Models of Code
Aftab Hussain
 
May Marketo Masterclass, London MUG May 22 2024.pdf
May Marketo Masterclass, London MUG May 22 2024.pdfMay Marketo Masterclass, London MUG May 22 2024.pdf
May Marketo Masterclass, London MUG May 22 2024.pdf
Adele Miller
 
Launch Your Streaming Platforms in Minutes
Launch Your Streaming Platforms in MinutesLaunch Your Streaming Platforms in Minutes
Launch Your Streaming Platforms in Minutes
Roshan Dwivedi
 
Fundamentals of Programming and Language Processors
Fundamentals of Programming and Language ProcessorsFundamentals of Programming and Language Processors
Fundamentals of Programming and Language Processors
Rakesh Kumar R
 
Cracking the code review at SpringIO 2024
Cracking the code review at SpringIO 2024Cracking the code review at SpringIO 2024
Cracking the code review at SpringIO 2024
Paco van Beckhoven
 
Transform Your Communication with Cloud-Based IVR Solutions
Transform Your Communication with Cloud-Based IVR SolutionsTransform Your Communication with Cloud-Based IVR Solutions
Transform Your Communication with Cloud-Based IVR Solutions
TheSMSPoint
 
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket ManagementUtilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
Utilocate
 
Graspan: A Big Data System for Big Code Analysis
Graspan: A Big Data System for Big Code AnalysisGraspan: A Big Data System for Big Code Analysis
Graspan: A Big Data System for Big Code Analysis
Aftab Hussain
 
Using Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional SafetyUsing Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional Safety
Ayan Halder
 
Artificia Intellicence and XPath Extension Functions
Artificia Intellicence and XPath Extension FunctionsArtificia Intellicence and XPath Extension Functions
Artificia Intellicence and XPath Extension Functions
Octavian Nadolu
 
GraphSummit Paris - The art of the possible with Graph Technology
GraphSummit Paris - The art of the possible with Graph TechnologyGraphSummit Paris - The art of the possible with Graph Technology
GraphSummit Paris - The art of the possible with Graph Technology
Neo4j
 
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI AppAI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
Google
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOMLORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
lorraineandreiamcidl
 
Quarkus Hidden and Forbidden Extensions
Quarkus Hidden and Forbidden ExtensionsQuarkus Hidden and Forbidden Extensions
Quarkus Hidden and Forbidden Extensions
Max Andersen
 
Enterprise Resource Planning System in Telangana
Enterprise Resource Planning System in TelanganaEnterprise Resource Planning System in Telangana
Enterprise Resource Planning System in Telangana
NYGGS Automation Suite
 
E-commerce Application Development Company.pdf
E-commerce Application Development Company.pdfE-commerce Application Development Company.pdf
E-commerce Application Development Company.pdf
Hornet Dynamics
 

Recently uploaded (20)

Top Features to Include in Your Winzo Clone App for Business Growth (4).pptx
Top Features to Include in Your Winzo Clone App for Business Growth (4).pptxTop Features to Include in Your Winzo Clone App for Business Growth (4).pptx
Top Features to Include in Your Winzo Clone App for Business Growth (4).pptx
 
Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604Orion Context Broker introduction 20240604
Orion Context Broker introduction 20240604
 
OpenMetadata Community Meeting - 5th June 2024
OpenMetadata Community Meeting - 5th June 2024OpenMetadata Community Meeting - 5th June 2024
OpenMetadata Community Meeting - 5th June 2024
 
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing SuiteAI Pilot Review: The World’s First Virtual Assistant Marketing Suite
AI Pilot Review: The World’s First Virtual Assistant Marketing Suite
 
A Study of Variable-Role-based Feature Enrichment in Neural Models of Code
A Study of Variable-Role-based Feature Enrichment in Neural Models of CodeA Study of Variable-Role-based Feature Enrichment in Neural Models of Code
A Study of Variable-Role-based Feature Enrichment in Neural Models of Code
 
May Marketo Masterclass, London MUG May 22 2024.pdf
May Marketo Masterclass, London MUG May 22 2024.pdfMay Marketo Masterclass, London MUG May 22 2024.pdf
May Marketo Masterclass, London MUG May 22 2024.pdf
 
Launch Your Streaming Platforms in Minutes
Launch Your Streaming Platforms in MinutesLaunch Your Streaming Platforms in Minutes
Launch Your Streaming Platforms in Minutes
 
Fundamentals of Programming and Language Processors
Fundamentals of Programming and Language ProcessorsFundamentals of Programming and Language Processors
Fundamentals of Programming and Language Processors
 
Cracking the code review at SpringIO 2024
Cracking the code review at SpringIO 2024Cracking the code review at SpringIO 2024
Cracking the code review at SpringIO 2024
 
Transform Your Communication with Cloud-Based IVR Solutions
Transform Your Communication with Cloud-Based IVR SolutionsTransform Your Communication with Cloud-Based IVR Solutions
Transform Your Communication with Cloud-Based IVR Solutions
 
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket ManagementUtilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
Utilocate provides Smarter, Better, Faster, Safer Locate Ticket Management
 
Graspan: A Big Data System for Big Code Analysis
Graspan: A Big Data System for Big Code AnalysisGraspan: A Big Data System for Big Code Analysis
Graspan: A Big Data System for Big Code Analysis
 
Using Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional SafetyUsing Xen Hypervisor for Functional Safety
Using Xen Hypervisor for Functional Safety
 
Artificia Intellicence and XPath Extension Functions
Artificia Intellicence and XPath Extension FunctionsArtificia Intellicence and XPath Extension Functions
Artificia Intellicence and XPath Extension Functions
 
GraphSummit Paris - The art of the possible with Graph Technology
GraphSummit Paris - The art of the possible with Graph TechnologyGraphSummit Paris - The art of the possible with Graph Technology
GraphSummit Paris - The art of the possible with Graph Technology
 
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI AppAI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
AI Fusion Buddy Review: Brand New, Groundbreaking Gemini-Powered AI App
 
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOMLORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
LORRAINE ANDREI_LEQUIGAN_HOW TO USE ZOOM
 
Quarkus Hidden and Forbidden Extensions
Quarkus Hidden and Forbidden ExtensionsQuarkus Hidden and Forbidden Extensions
Quarkus Hidden and Forbidden Extensions
 
Enterprise Resource Planning System in Telangana
Enterprise Resource Planning System in TelanganaEnterprise Resource Planning System in Telangana
Enterprise Resource Planning System in Telangana
 
E-commerce Application Development Company.pdf
E-commerce Application Development Company.pdfE-commerce Application Development Company.pdf
E-commerce Application Development Company.pdf
 

System Event Monitoring for Active Authentication

  • 1. 34 IT Pro July/August 2013 P u b l i s h e d by t h e I E E E C o m p u t e r S o c i e t y 1520-9202/13/$31.00 © 2013 IEEE Jeffery Payne, Mark Fenner, and Jonathan Kauffman, Coveros The authors use system event monitoring to distinguish between the behavioral characteristics of normal and anomalous computer system users. Identifying anomalous behavior at the system event level diminishes privacy concerns and supports the identification of cross-application behavioral patterns. A nomaly detection with system events is used in a variety of intrusion- detection systems.1–4,8 Our project for DARPA’s Active Authentication program uses system event monitoring to dis- tinguish between the behavioral characteristics of normal and anomalous computer system us- ers. Identifying anomalous behavior in system events has advantages over doing so at other lev- els. First, there are fewer privacy concerns with the information collected, because system event streams don’t include the acted-on content. Sec- ond, information over all applications and pro- cesses is available to identify cross-application behavioral patterns. Our Approach We perform active authentication by detecting anomalies in a stream of operating system events generated by a computer user. Figure 1 shows how our system operates. The process in Figure 1a begins when a comput- er user generates system events as he or she per- forms application-level operations. These events are captured using either Microsoft’s Procmon or Detours utilities. Procmon monitors a variety of system events, including file system events, reg- istry access, network events, and process/thread activity. Detours can intercept Windows system events individually. System events are filtered based on their process type and event class. Initial System Event Monitoring for Active Authentication Security: DARPA
  • 2. computer.org/ITPro 35 experimentation led us to remove system events from background tasks. We filter events by class to keep similar events together so that detectors can evaluate a single type of event. Once filtered, system events are analyzed using one or more anomaly-detection learning sys- tems—including one-class sup- port vector machine (1SVM), replicator neural network (RNN), probabilistic similarity (PS), and similarity matching. Anomaly- detection learning systems5 dif- fer from standard classification learning systems, because they’re trained on instances of a single class and then determine wheth- er new instances are members of that original class. 1SVM is a spe- cific implementation of SVMs to handle anomaly detection.6 RNNs use backpropagation training cou- pled with a specific architecture to reconstruct input examples as output.7 The error in reconstruc- tion is interpreted as an anomaly score. PS, based on an approach first proposed by Salvatore Stolfo and his colleagues,7 is an anomaly detector that identifies anomalies based on the marginal distribu- tions of tuples in a training set of features. Similarity matching is a baseline algorithm that matches system event streams with previ- ously seen streams. In our approach, we break the raw system event stream into fixed-time chunks called blocks. Each block provides a sequence of overlapping subsequences, called windows, of system events. The learning algorithms are applied directly to these windows. Each window is labeled normal or anomalous. We then compare the percent of anomalous windows in the larger block against a threshold and render an accept or reject decision for that block. The results of multiple detectors are combined using Boolean or counting operations. At both the window and block levels, we can apply a leaky bucket algorithm to smooth the detection stream. When the system recognizes an anomaly, it either sends an alert to a centralized monitoring system or requires a manual re-authentication. Figure 1b shows how we evaluated the accu- racy of our anomaly-detection approach. After performing training on a training block, we fed our system both authorized user data (additional system events from the same training user) and Figure 1. A high-level depiction of the system-event-monitoring approach: the (a) system-event-monitoring process and (b) approach for determining false accept and false reject rates. A computer user generates system calls. The system calls are captured and filtered. Then the calls are analyzed Procmon/Detours Subsets noise Multilayer perceptron or support vector machine Noise reduction rating identity Leaky bucket thresholds voting Login prompt centralized monitor Authorized user Unauthorized user Training block Self-test block Other-test block Windows Blocks Streams Anomaly detector (learn) Anomaly detector (detect) Anomaly detector (detect) Threshold Compare number of anomalous windows to threshold Compare number of anomalous windows to threshold False reject Accept Reject False accept (a) (b) to make a decision whether to re-authenitcate.
  • 3. 36 IT Pro July/August 2013 Security: DARPA unauthorized user data (system events from a dif- ferent user). If a block from the training user was rejected, the result was a false reject. If a block from a different user was accepted, the result was a false accept. Otherwise, the result was true (correct). Empirical Results To identify a set of system parameters that ex- hibits good performance for a variety of users, we developed an experimentation framework in Python that evaluates various sets of parameters for event streams produced by multiple users. To collect these streams, we deployed Procmon on seven test subjects’ workstations and collected two hours of system events during their normal, daily computer usage. Their typical tasks includ- ed communication, software development and configuration, and document creation. Each participant produced an event stream that was 2–4 Gbytes in size and comprised 3–8 million records. We split each of these event streams into two sets. We used the first set to tune the system parameters. The experimentation framework uses this first set as normal data for a particular partici- pant while using the second set of each participant’s event stream to represent anomalous behavior. During initial experimentation, we explored a variety of parameter settings to measure their ­effect on system performance. For most of the system parameters, we didn’t observe a significant change in performance as we varied their values. However, using the registry subset of events, as well as enhancing event representation to include both the elapsed time since the previous event and the name of the process that invoked the system event, noticeably improved system performance. Figure 2 presents our best results using a detec- tion error trade-off (DET) curve for both individ- ual users and all users as a whole. We produced the graph by applying cross-validation to the first dataset. DET curves show how the performance of a single system changes as the threshold for a block of anomalies varies. Each point on the graph represents system performance, both in terms of the false reject rate (FRR) and false ac- cept rate (FAR) at a particular threshold value. The unconnected points of the same shape show the DET curve for each individual user, while the thick line shows the DET curve for all users. To quantify a DET curve, we measure its area under the curve (AUC). A DET curve that’s pushed to- ward the origin will have a low AUC measure, in- dicating that the system is more accurate. Our best results were obtained by aggregating the verdicts produced by two separate detectors. We tried several methods of combining verdicts and found that a method based on our relative trust of the two detectors produced the best results. In this approach, one detector more reliably accepts an au- thorized user, and another more reliably rejects an unauthorized user. When one detector produces a trusted verdict and the other produces an untrusted verdict, we use the trusted verdict as the aggregate. In the case where both detectors produce a trusted verdict, we choose the verdict from the detector we trust (a priori) the most. The detec- tor we trust the most in this scenario is the one that gave the most accurate verdicts during indi- vidual testing of each detector. When both detec- tors produce an untrusted verdict, we declare the result as unknown. Depending on the threshold value, this means that anywhere from 0–20 per- cent of the testing results were removed. By con- sidering the curves for individual users, we reach an FRR of 0 percent and a FAR of approximately 10–15 percent for several users. O ur future research will focus on improv- ing these results when applied to differ- ent types of users. Experimentation will examine the degree to which normal behavior can be distinguished from malicious behavior as opposed to simply normal versus abnormal use. Figure 2. Results from multisensor analysis. The unconnected points of the same shape show the detection error trade-off (DET) curve for each individual user, while the thick line shows the DET curve for all users. 100 80 60 40 20 0 0 20 40 60 80 FRR FAR 100
  • 4. computer.org/ITPro 37 We’re currently collecting additional field test data so we can compare it to malicious use-case scenarios developed from real-world attacks. We also plan on researching mechanisms for dynam- ically modifying threshold values on the basis of system performance. References 1. A.K. Ghosh, J. Wanken, and F. Charron, “Detect- ing Anomalous and Unknown Intrusions Against Programs,” Proc. Computer Security Applications Conf., IEEE, 1999, pp. 259–267. 2. A.K. Ghosh, A. Schwartzbard, and M. Schatz, “Using Program Behavior Profiles for Intrusion Detection,” Proc. SANS Conf. and Workshop on Intrusion Detection and Response, 1999. 3. A.K. Ghosh, A. Schwartzbard, and M. Schatz, “Learning Program Behavior Profiles for Intrusion Detection,” Proc. 1st Workshop on Intrusion Detection and Network Monitoring (ID 99), Usenix, 1999, p. 6. 4. V. Chandola, A. Banerjee, and V. Kumar. “Anomaly Detection: A Survey,” ACM Computing Surveys, vol. 41, no. 3, 2009, article 15. 5. S. Hawkins et al., “Outlier Detection Using Replica- tor Neural Networks,” Proc. 4th Int’l Conf. Data Ware- housing and Knowledge Discovery (DaWaK 00), Springer, 2002, pp. 170–180. 6. B. Scholkopf et al., “Estimating the Support of a High-Dimensional Distribution,” Neural Computation, vol. 13, no. 7, 2001, pp. 1443–1471. 7. S. Stolfo et al., “A Comparative Evaluation of Two Al- gorithms for Windows Registry Anomaly Detection,” J. Computer Security, vol. 13, no. 4, 2005, pp. 659–693. 8. S. Forrest et al.,“A Sense of Self for Unix Processes,” Proc. IEEE Symp. Security and Privacy, IEEE, 1996, pp. 120–128. Jeffery Payne is CEO and founder of Coveros and is a princi- pal investigator on an Active Authentication project ­sponsored by DARPA. Contact him at jeff.payne@coveros.com. Mark E. Fenner is a research scientist at Coveros. Contact him at mark.fenner@coveros.com. Jonathan Kauffman is a consultant at Coveros, where he’s performing research in Active Authentication. Contact him at jonathan.kauffman@coveros.com. handles the details so you don’t have to! Professional management and production of your publication Inclusion into the IEEE Xplore and CSDL Digital Libraries Access to CPS Online: Our Online Collaborative Publishing System Choose the product media type that works for your conference: Books, CDs/DVDs, USB Flash Drives, SD Cards, and Web-only delivery! www.computer.org/cps or cps@computer.org Contact CPS for a Quote Today!