Pia -predict

Privacy Impact Assessment
for the

Protected Repository for the Defense of
Infrastructure Against Cyber Threats
(PREDICT)
February 25, 2008
Contact Point
Douglas Maughan
Cyber Security R&D
Science & Technology Directorate
(202) 254-6145
Reviewing Official
Hugo Teufel III
Chief Privacy Officer
Department of Homeland Security
(703) 235-0780

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 2

Abstract
The Science & Technology Directorate’s Protected Repository for the Defense of Infrastructure
Against Cyber Threats (PREDICT) system is a repository of test datasets of Internet traffic data that is made
available to approved researchers and managed by an outside contractor serving as the PREDICT
Coordination Center (PCC). The goal of PREDICT is to create a national research and development (R&D)
resource to bridge the gap between (a) the producers of security-relevant network operations data and (b)
technology developers and evaluators who can use this data to accelerate the design, production, and
evaluation of next-generation cyber security solutions, including commercial products. A key motivation
of PREDICT is to make these data sources more widely available to technology developers and evaluators,
who are currently forced to base the efficacy of their technical solutions on old, irrelevant traffic data,
anecdotal evidence, or small-scale test experiments, rather than on more comprehensive, real-world data
analysis.

Overview
PREDICT is sponsored by the Department of Homeland Security (DHS) Science & Technology
Directorate’s Cyber Security R&D Group (S&T/CCI). PREDICT represents an important three-way
partnership between government, critical information infrastructure providers, and the security
development community (both academic and commercial), all of whom seek technical solutions to protect
the public and private information infrastructure.
The purpose of PREDICT is to help accelerate the advancement of network-based cyber defense
research, product development, and evaluation. Specifically, PREDICT provides researchers, developers,
and evaluators with regularly updated network operations data sources relevant to cyber defense technology
development. PREDICT datasets will provide researchers and developers timely and detailed insight into
cyber attack phenomena occurring across the Internet.
PREDICT will facilitate the selection, indexing, and hosting of sets of data on Internet traffic from
approved network operators and make those datasets available to approved researchers for studies on cyber
security. Internet traffic data used in PREDICT does not include any content (e.g.: the text of email
messages or webpages). Traffic data consists of datasets containing, for example, topology measurement
data, blackhole address space data, packet traces and headers, firewall logs, border gateway protocol update
messages and routing table dumps, and VOIP measurement data. DHS selected the data providers and
approved the data that they would provide to the PREDICT project. DHS also contracted with data
providers or other entities to store and host the datasets and provide access to them to approved
researchers. DHS has contracted with a not-for-profit organization to serve as the PCC. The PCC will
maintain the list of approved researchers, data providers, and data hosts, as well as, an index of the sets of
data that are available. The PCC will not maintain any of the data sets, but rather will act as a library for
researchers to review a catalog of metadata and determine which datasets will be of use to their particular
research projects.
The concept of dataset repositories, created by research communities for the purpose of advancing
the depth and quality of analysis techniques or to better understand the characteristics of operational
systems, is not unique. Similar repositories have been developed in other types of technical fields for R&D
activities. By creating a community of data providers and an operational structure for disseminating critical

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 3

data sources to those seeking to field the next generation of cyber defense technologies, PREDICT has the
potential to dramatically accelerate the R&D community’s ability to develop effective and timely cyber
defense technologies. It will also enable researchers to more closely monitor the emerging trends and
patterns of attacks that propagate across the Internet.
The entire PREDICT operational structure is handled by contractors and a volunteer community of
data providers. These datasets will be made available to approved researchers through the PCC, in
accordance with established policies and procedures, for purposes of conducting cyber security research
and development. Researchers will be bound by legal agreement regarding the use and security of the
datasets. Researchers must establish and maintain the appropriate administrative, technical, and physical
safeguards to protect access to the data to only authorized and approved persons.
In the operation of PREDICT, the PCC will establish and maintain administrative records on
researchers, sponsoring institutions, data providers, data hosts, an advisory board, an application review
board (ARB), and a publication review board (PRB). These records will contain personally identifiable
information (PII) and will be for the internal administration of PREDICT. This data will not be accessed by
anyone other than the approved and authorized administrators of the PCC and persons involved in PCC
operations.
The administrative data collected by PCC is collected on behalf of DHS, and as such is covered by
the DHS System of Records Notice 002 published December 6, 2004 69 Fed Reg 70460 DHS/ALL 002. As
part of the user agreement individual users provide consent for the sharing that occurs between the data
provider, the PCC, and the data hosts. As a rule, DHS does not have access to any of the administrative
records or the data sets.
This PIA analyzes the privacy risks presented by (a) datasets used in PREDICT, and (b) the
administrative records. It describes the steps that S&T/CCI, the contractors, and volunteer community have
taken to mitigate those risks. While many of the test datasets contain no PII, some do contain Internet
Protocol (IP) addresses. PREDICT minimizes privacy risks by anonymizing those elements of data in the
test datasets that are not already available to the public and which may be linked to PII, notably IP
addresses. The non-anonymized data does not include any direct PII, such as names, addresses, or
telephone numbers. These datasets are described more fully in Section 1.0 of this document pertaining to
“Characterization of the Information.” The anonymization of the datasets is discussed in Section 5.0,
“External Sharing and Disclosure.”
In preparation for this project, S&T/CCI (1) analyzed the privacy risks associated with each type of
dataset and the administrative records, (2) determined which test datasets are suitable for the PREDICT
project, (3) worked closely with the PREDICT volunteer data provider community and data hosts to
establish PREDICT operational rules and policies and procedures, and (4) determined privacy protections
applicable for each test dataset and the administrative records. Intermediaries, such as DHS, the PCC, or
advisory board do not have access to any of the datasets provided to PREDICT.
PREDICT seeks to facilitate new cyber defense solutions by making available traffic data sources that
would otherwise be very difficult to obtain by cyber defense researchers and developers. The complexity
of the Internet architecture and the sophistication of attacks require several different types of Internet data
be available to researchers. For example, Domain Name Server (DNS) root server data, Internet topology
measurement data, blackhole address space data, Border Gateway Protocol (BGP) update messages and BGP
routing table dumps, packet headers, worm data, and firewall logs are some of the types of data required

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 4

for effective cyber security R&D. Today, however, these datasets are not available to most of the research
community.
At present, large-scale attack detection analyses are driven primarily through opportunistic data
sources, such as firewall logs and intrusion detection alerts. While several studies have demonstrated the
efficacy of firewall logs and intrusion detection data to detect some classes of self-replicating malicious
code, these data sources also have inherent issues regarding their enormous overall size, data sensitivities,
and threat coverage limitations. These data sources are used today, primarily, because they are available,
not because they are the best data sources. The most widely used intrusion detection data source was
created in 1998 by the Defense Advanced Research Projects Agency (DARPA). Given the age of this dataset,
the value it offers current researchers is limited. Traffic data that is 10 years old cannot be used to
effectively analyze today’s attacks, viruses, malicious code, and traffic patterns. This necessarily limits the
research community’s ability to produce offensive and defensive privacy and security solutions and
products.
Significant collections of diverse network operations data and statistics will allow researchers to
discover alternative data sources that support the detection of large-scale attack phenomena with:
•

Greater precision (i.e., less prone to perturbations that may lead to false positives);

•

Greater efficiency than current data sources (more economical); and

•

Lower sensitivity of available information (e.g., current IDS alerts and firewall logs have the
potential to reveal the producing site’s security posture).

For example, the types of issues that would be addressed by collecting a broad range of data from a
diverse set of providers would allow researchers to:
•

Map and predict how viruses and worms spread;

•

Develop models of the velocity/acceleration of the spreading of viruses and worms;

•

Predict the onset of attacks; and

•

Develop models of network topologies and how they accelerate/decelerate an attack.

Research of this nature will significantly advance efforts to find technological solutions to protect
privacy and defend against attacks and other forms of security breaches.
PREDICT OPERATIONAL STRUCTURE
DHS interviewed volunteer data providers who were willing to contribute data to the PREDICT
project, selected datasets that were suitable, and selected the PCC. DHS also contracted with selected
entities to serve as hosts for the datasets. DHS worked collaboratively with the data providers, hosts, PCC,
and privacy experts to develop the PREDICT operational structure and associated policies and procedures.
The PCC serves as PREDICT’s operational body. In this role, the PCC:
•

Facilitates the data flow between PREDICT participants;

•

Processes applications from researchers for access to datasets or approval to publish research
results;

•

Develops and maintains a catalog of metadata about the datasets; and

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 5

•

Coordinates the development of protocols, such as anonymization of IP addresses and
restrictions on data (which are subject to DHS approval), to protect the confidentiality and
integrity of data and direct its proper usage.

The following nine types of organizations participate in PREDICT:
1. PREDICT Coordination Center (PCC)
2. Data Providers
3. Data Hosts
4. Researchers
5. Sponsoring Institutions
6. Application Review Board
7. Publication Review Board
8. Advisory Board
9. Department of Homeland Security
These entities work in a coordinated fashion to help ensure that the operational policies and
procedures are followed, privacy protections are effective, researchers are legitimate and sponsored by
worthy institutions, and the datasets are handled in a secure and responsible manner.
The PCC receives and catalogs metadata about the test datasets and makes the metadata catalog
available to approved researchers, subject to the terms and conditions set forth by DHS, PCC, and the data
providers and hosts in Memorandum of Agreements (MOAs) between the various parties. DHS and the
PCC do not store, maintain, or have access to any of the datasets; only the data hosts and approved
researchers have access to specific datasets. Upon reviewing the catalog of metadata, and submitting an
application to use specific datasets, approved researchers will obtain access information and will be able to
access the datasets directly from the data hosts.

Data providers include the operators of the Internet backbone, large and small Internet service
providers (ISPs), and large enterprise network operators. They provide datasets of Internet traffic data that
they own, or have permission to release, to PREDICT, subject to terms and conditions that they stipulate
and in accordance with privacy requirements, which are set forth in the MOA between the data provider
and the PCC. These terms and conditions are included in the MOA between the researcher and the PCC.
Data providers may also require researchers to sign a separate MOA with them regarding the use of their
data. The combination of data provider security requirements and PCC privacy requirements that are set
forth in MOAs between the researcher, the PCC, and the data provider, help ensure that any potentially
personally identifiable information in the datasets is anonymized 1 and that usage of the data is appropriate.
Data providers may select a data host to receive and host their datasets, or they may host their own data.
The MOA between the data provider and the PCC requires the provider to agree that:
•

1

The data provider will provide the PCC with metadata on the data they agree to make available

IP addresses that are not anonymized in datasets that are already available to the public through other sources are
not required to be anonymized for PREDICT.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 6

to data hosts for release to researchers approved by the PCC.
•

The data provider will make the data available to data hosts, for release to approved researchers
and no others, under the terms and conditions for access and use as specified by them and the
PCC.

•

The data provider may provide terms and conditions for access to and use of the data,
including identification requirements for the researcher; permitted uses and specific
restrictions; minimum safeguards to protect the data; procedures for receipt, handling, control,
dissemination, and return of data; and restrictions on publishing or releasing information
about the data (which is addressed below under Publication Review Board). The data
provider’s terms and conditions for access and use of their datasets will flow down to the
researcher and become part of the researcher’s application and Memorandum of Agreement to
use the dataset.

•

The data provider is responsible for the release of the data to PREDICT, and is solely
responsible for ensuring that any data it releases complies with all applicable statutes and
regulations of governing or regulating bodies and contractual agreements. The data provider
must also ensure that the release of the data is consistent with its privacy, security, or other
policies and procedures applicable to the data. The data provider must certify that the data
provided for use in the PREDICT program is in compliance with the foregoing and that the
data has been sanitized, de-identified, or cleaned of any and all information that would not be
in compliance or consistent with its policies and procedures and privacy requirements. The
data providers have been asked to ensure that their privacy policies are available to the
researchers.

•

Non compliance with these requirements may result in the data provider’s expulsion from the
PREDICT project, including removal of the provider’s datasets from the PREDICT project.

Data hosts provide computing infrastructure to store datasets received from the data providers and
provide methods by which researchers can access datasets after the researcher’s application for particular
dataset(s) has been approved.
Data hosts must enter into a MOA with PCC in which they agree to specific terms regarding access
to the datasets, including that:
•

The data hosts will accept data from approved data providers, for release to approved
researchers, subject to the terms and conditions set forth by the providers and hosts.

•

The data hosts will provide terms and conditions for access to, transfer, storage, and use of the
data as required by the data provider and PCC, as well as any other restrictions the host deems
necessary to accomplish efficient and secure access to the data.

•

The data hosts acknowledge that the data access approval given to a researcher in any
application will permit access to the requested data by that researcher, regardless of approval or
denial of access to that researcher in any other previous PREDICT dataset application.

•

The data hosts are solely responsible for ensuring that any data they release complies with the
host’s separate agreement with the data provider.

Researchers are members of the cyber defense R&D community who complete an official PCC
application, which serves as a MOA between the researcher and the PCC, requesting datasets from PREDICT
for use in their research and who are affiliated with a sponsoring institution.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 7

Researchers must agree that:
•

They will not use the data for purposes other than described in their application.

•

The researchers will not disclose the data to any persons other than those identified in their
application.

•

The researchers will establish and maintain the appropriate administrative, technical, and
physical safeguards to protect the confidentiality of the data and to prevent unauthorized use of
or access to the data.

•

The researchers will permit others to use the data only in accordance with the terms of the
MOA and the procedures in the researcher’s application.

•

If the researcher moves to a different institution, he/she must notify PCC and the sponsoring
institution in writing regarding the disposition of all copies of the data and follow PCC’s
directions and the sponsoring institution’s guidelines. A researcher may apply for a new
PREDICT account through his/her new institution and, once approved, submit a new
application for continued usage of the datasets.

•

No findings, analysis, or information derived from the data may be released if such findings
contain any combination of data elements that might allow for identification or the deduction
of a person’s or institution’s identity. The Publication Review Board and associated review of
documents submitted for publication is intended to help ensure compliance with privacy
requirements.

•

Any findings, results of analysis, or manuscripts proposed for public release, publication, or
any other type of disclosure to persons not listed and approved in this application (e.g.,
abstracts, presentations (oral or written), publications) must be submitted for review by a
Publication Review Board managed by PCC prior to publication or public release to assure that
data confidentiality is maintained, entities or individuals cannot be identified, and the terms
and conditions attached to the use of the data have been followed.

•

The researcher must immediately report to the PCC any use or disclosure of the Data other than
as permitted and will take all reasonable steps to mitigate the effects of such improper use or
disclosure, cooperating with all reasonable requests of PCC towards that end.

•

In the event PCC determines or has a reasonable belief that researcher has violated any terms of
the MOA, PCC may terminate the MOA and require the researcher to return the data and all
derivative files. PCC may also seek injunctive relief against the researcher or the sponsoring
institution to prevent any unauthorized disclosure of data. In addition, PCC will report any
misuse or improper disclosure of the data to the data provider and host and to appropriate
authorities as required by applicable Federal or state law.

•

The researcher will destroy all copies of the data when the MOA expires or take such action as
specified in the MOA and will certify such destruction or return by signing and providing to
PCC a Certification of Data Return or Destruction.

Sponsoring Institutions are organizations that are affiliated with or otherwise sponsor researchers
and validate their research and need for PREDICT data. Sponsoring Institutions could be corporations,
academic institutions, national laboratories, or government research laboratories.
The Application Review Board (ARB), in conjunction with the PCC and the data provider, reviews
and approves or rejects the application to use PREDICT datasets and forwards approved applications to data

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 8

hosts to enable the researcher to have access to and use the requested data. The Application Review Board
is chaired by the PCC and includes representation from data providers, data hosts, and a subject matter
expert in cyber security.
The Publication Review Board (PRB) reviews and approves or rejects applications from
researchers or sponsoring institutions to publish or otherwise release any information, study results, or
other summaries or descriptions on, about, or relating to data or metadata previously made available
through the PREDICT project. Publication constraints are defined by the data provider during the process
when the data is initially inserted into the PREDICT system. The Publication Review Board is comprised of
PCC personnel, the original data provider, and the hosting site for the provider data (if the host is not the
provider).
The Advisory Board advises and makes recommendations to the PCC on policy and issues relating
to privacy issues and the general direction of the PREDICT project.
The Department of Homeland Security (DHS) selects the data providers and approves the data
that they provide to the PREDICT project. DHS also contracts with data providers or other entities to store
and host the datasets and provide access to them to approved researchers. DHS has contracted with a notfor-profit organization to serve as the PCC. DHS reviews the operational procedures of PREDICT and the
security of the PCC computer system through its certification and accreditation process.
This PIA addresses both the administrative data collected by the PCC and the test datasets that are
available to approved researchers for cyber security R&D. The administrative data is a system of records and
requires a SORN and PIA. The administrative data, however, is not used by DHS, it is not accessed by DHS,
and it does not support any DHS operations. It is not shared except as outlined in this PIA. The approved
test datasets that are accepted into the PREDICT project and provided to approved researchers do not
contain any data subject to the Privacy Act of 1974 and are not subject to a SORN. As noted above, they
are included in this PIA as a voluntary measure by DHS to explain the privacy analysis that has been
conducted with respect to the datasets.

Section 1.0 Characterization of the Information
The following questions are intended to define the scope of the information requested
and/or collected as well as reasons for its collection as part of the program, system, rule, or
technology being developed.

1.1

What information is collected, used, disseminated, or
maintained in the system?

Contact and Information Related to PREDICT Datasets
The PCC collects administrative information from researchers, data providers, data hosts,
sponsoring institutions, advisory board members, ARB members, and PRB members. The PCC maintains a
catalog of metadata about the test datasets which is available to approved researchers, and it keeps an
operational record of activities conducted within the PREDICT project. These operational records will be for
the internal administration of the PREDICT project and this data will not be accessed by anyone other than
the approved and authorized administrators of the PCC.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 9

1.1.1 Administrative Information Collected from Researchers,
Data Providers, and Data Hosts
The PCC collects the following information from researchers, data providers, and data hosts:
•

First and last name

•

Address (street, city, state, zip code)

•

Telephone numbers (home, work, cell, and fax – only the work number is a required field)

•

Email address

•

Name of Sponsoring Institution

•

Address of Sponsoring Institution (street, city, state, zip code)

•

Name of authorized representative from Sponsoring Institution

•

Telephone number of authorized representative

•

Email address of authorized representative

•

Approved researchers are assigned a username and password to access the PREDICT portal. The
PCC also keeps a list of the datasets requested by Researcher and whether the request was
approved or disapproved, including the reason for disapproval.

•

The PCC also keeps record of requests from approved researchers to publish documents,
articles, or materials pertaining to their research and whether the request was approved by the
Publication Review Board or disapproved, including the reason for disapproval.

•

The PCC keeps a record of the approved datasets that are provided from each data provider and
which datasets the data hosts are hosting.

1.1.2 Application Review Board and Publication Review Board
Members
The PCC collects the following information from ARB and PRB members:
•

First and last name

•

Address (street, city, state, zip code)

•

Telephone number

•

Email address

•

Name of organization

1.1.3 Advisory Board Members
The PCC collects the following information from Advisory Board members:
•

First and last name

•

Address (street, city, state, zip code)

•

Telephone number

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 10

•

Email address

•

Name of organization

1.1.4 Sponsoring Institutions
The PCC collects the following information from Sponsoring Institutions:
•

Name of Organization

•

First and last name of representative of Sponsoring Institution

•

Title

•

Address (street, city, state, zip code)

•

Telephone number

•

Email address

1.1.5 Dataset Information
The datasets used in PREDICT contain real traffic data provided by data providers in accordance
with the terms of the MOA between the PCC and the data provider (see above section discussing Data
Provider MOA obligations). The data providers are not, however, collecting this information specifically
for PREDICT. The data providers are already collecting the information to support their own cyber security
responsibilities and are now choosing to share that same information with PREDICT. DHS will not have
access to any of the administrative or traffic data provided for PREDICT.
TECHNICAL ASPECTS OF THE DATA
Content sent over the Internet is broken into packets. The packets contain a header with the
sender’s and recipient’s Internet Protocol (IP) addresses. Generally, an IP address is randomly assigned to a
computer at the time the user logs onto the Internet. However, it is sometimes possible to trace IP
addresses, even if they are not static, back to individual users. Therefore, all IP addresses in PREDICT Phase
I datasets that are not already available to the public will be anonymized by the data provider according to
requirements in the MOA between the data provider and the PCC and any other data that could be used to
trace the data back to an individual will be eliminated from the dataset. The PIA will be modified to
include information on different types of datasets that may be added to PREDICT in future phases.
The PCC, or another contractor, will conduct periodic compliance checks of the PREDICT project to
ensure compliance with these privacy protections and other PREDICT operating procedures and contractual
provisions. The following types of datasets are being used in Phase I of the PREDICT project. The
described uses of these datasets are intended as examples. As technology advances, traffic changes, or the
innovative process advances, researchers may use the datasets differently than described below, however all
uses of this data by researchers will always remain within the scope of the governing MOA.

Topology Measurement Data
This data is obtained from computers that the data provider puts on the network in order to map
the network connections of the Internet connecting out from that point. The computers send out probe

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 11

packets with Time to Live (TTL). 2 The packet is owned by the data provider. It is sent out and comes back
with information about routing, but no data is transmitted in the process. The data provider makes an
Anonymous System (AS) core and ISP level map of Internet connectivity, based on the measurement of the
TTL. The IP addresses in this dataset are not anonymized because this dataset contains measurement traffic 3
and not actual sender-receiver communication. The data provider requests that researchers not probe
certain Internet Protocol (IP) addresses or disclose IP addresses to anyone else.

Blackhole Address Space Data
The data provider owns a large number of IP addresses. Traffic to legitimate addresses owned by
the provider is delivered, and the remainder goes back to the data provider because the traffic is targeting
unassigned IP addresses. Since this traffic was trying to saturate a target population, it is generally hitting
illegitimate IP addresses and is malicious traffic, such as scanners and worms. This traffic data can be useful
for studying backscatter from distributed denial of service (DDOS) attacks, worm spread (growth rates,
population size, and affected population), scanning and backdoor activity, and evaluating various honeypot
responders. In addition, all IP addresses will be anonymized.

Hi-Speed ISP Exchange Data from OC-48 operational network (packet traces)
This dataset is comprised of headers of traffic data that contain the source and destination IP
addresses, but contains no content data. Pursuant to the terms of the MOA between the data provider and
the PCC, the data provider must anonymize the IP addresses in headers. This dataset can be used for DDOS
attack detection and characterization, malicious behavior detection (port scans and intrusion attempts),
worm outbreak characterization, backscatter analysis, botnet distribution and characterization, analysis of
impact of attack on normal traffic, spectral analysis of attacks and background traffic, and intrusion
detection and firewall research and development.

Full Packet Headers
Full packet headers contain IP addresses for the sender and recipient. The port numbers identify
the application used (e.g., Internet browser or email). No packet content is included in this dataset,
although a hash of the data associated with each packet header may be provided. Pursuant to the MOA
between the data provider and the PCC, all IP addresses will be anonymized.

Domain Name Server (DNS) Root Server Data
This is root server data from the host of major DNS root servers. The data will identify the user by
IP address. It will contain the IP address for the requested site, but it will not indicate whether the person
associated with the IP address actually connected to that site. Generally, requests are aggregated by multiple
users, but some are not. All IP addresses will be anonymized. This data will be used for DNS root server
traffic analysis and characterization and DNS root server attack analysis and characterization.

2

The time to live (TTL) value can be thought of as an upper bound on the time that an IP datagram can exist
in an internet system. The TTL field is set by the sender of the datagram, and reduced by every host on the route to
its destination. The purpose of the TTL field is to avoid a situation in which an undeliverable datagram keeps
circulating on an internet system, and such a system eventually becoming swamped by such immortal datagrams.
3

Measurement traffic contains IP addresses of the machines that a packet touches as it travels along the
network before it is sent back to the sender, allowing measurement of Internet topology and routing.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 12

Internet Topology Data
Internet topology data is created by a program that tries to map the Internet. The program is able
to determine which routers are capable of talking to other routers. Internet topology data only shows
router connectivity within the Internet core and to external enterprise borders; it does not contain any
identifiable information or internal enterprise topology information. This dataset can be used for worm
outbreak modeling and simulation, worm containment and countermeasures, zombie distribution for
DDOS attacks, vulnerability assessments, longitudinal studies of the evolution of Internet topology and
address distribution, Internet topology and address map inference.

Address Space Allocation Data
This data is used to determine which parts of the IP address space are used. This information will
not identify persons or organizations. This dataset can be used for worm outbreak modeling and
simulation, worm containment and countermeasures, zombie distribution for DDOS attacks, vulnerability
assessments, longitudinal studies of the evolution of Internet topology and address distribution, Internet
topology and address map inference.

Enterprise Data
This data is internal traffic data from a large enterprise. It will consist of only headers with
anonymized IP addresses. No content will be included. Nevertheless, the operating procedures of PREDICT
and the PCC involves internal control points to ensure that the data provided has undergone an internal
review by the provider, such as reviews by its legal counsel, regarding whether anonymization or
notification to users is required. The principle security-oriented use of the enterprise datasets will be to
serve as a representative example of real network traffic. By providing a large amount of real network
traffic (versus developed test datasets), PREDICT’s goal is to provide a resource for researchers to use in
assessing the false positive rates and/or collateral damage of deploying proposed detection algorithms.
Prior to this effort, there have been no such enterprise background traces available, to the
significant detriment of researchers attempting to devise enterprise-level network security mechanisms that
will actually work soundly in practice. Such data is crucial in assessing security counter-measures since
networks are massively heterogeneous and are filled with large amounts of benign traffic that is not
malicious, but which could confuse attack detectors and falsely trigger defenses. More generally, there
have been virtually no studies of what enterprise traffic looks like for nearly fifteen years, and, thus, these
traces will provide a key first look at a hitherto unknown landscape. Since security breaches can be from
internal or external sources, a clearer understanding of internal traffic data will help advance the
development of cyber security technologies to counter insider attacks.

BGP Routing Table Data
This dataset captures “snapshots” of the topological state of the Internet by archiving Border
Gateway Protocol (BGP) routing tables from Internet routers in many locations around the world (these are
called Internet Exchange Points). Each routing table expresses the “view” of the Internet from that router’s
point in the overall topology and, taken together, all of these views provide a relatively complete roadmap
of the connectivity within the Internet Service Provider core of the Internet. This dataset contains only
backbone topology information; it does not contain any packet header information or information which
relates to individuals.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 13

BGP Routing Table Data is used by researchers who study the overall growth patterns of the
Internet over time, as well as those who are looking specifically at individual carriers, regions, or resources.
It shows historical trends in the utilization of the two principal Internet resources, IP addresses and
Autonomous System Numbers (ASN), and this presents the basic backdrop against which many other
trends are tracked. Several organizations provide complementary, partly-overlapping, and slightly
methodologically different partial views of a large and diverse environment.

BGP Update Messages & BGP Routing Table Dumps
The BGP Routing protocol is used to exchange routing information between different autonomous
systems in the Internet. Each BGP router sends information describing its own part of the Internet in the
form of BGP Update messages, which are sent to every BGP device. ISPs pay each other to carry packets.
Peering points are routing exchange points. Routing data is collected from several places on Internet.
Routing sensors are responsible for providing two types of information: an accurate picture of the local
routing topology and alerts on instability as viewed by the local network, indicating a potential resourceimpacting event. The routing sensors receive this info by peering with critical routers near blackhole
sensors. This data will contain raw BGP updates and routing table dumps (which are the same as the BGP
Routing Table Data described above). When various networks are added (or if networks lose connectivity),
the BGP protocol is responsible for propagating this information throughout the Internet. When a network
taken down, it watches to see how it heals itself. This information is already being made publicly available
by some institutions, such as http://www.routeviews.org. This dataset can be used to develop tools and
techniques such as beacons, fault injection, other measurement methodologies for reachability and stability,
and new routing protocols. It can also be used in the development of applications that will help determine
the stability of the topology (convergence, flapping), route hijacking, reachability, and interconnectedness.

VOIP Statistical Data (End-to-End Quality)
There are 2200 ISPs registered with data provider that collect data about the quality of VOIP
connections. This dataset contains only statistical information about VOIP calls. It does not contain phone
numbers or content of calls. It consists only of statistical information and a researcher would not be able to
identify who was talking, with whom, or about what subject. The dataset is comprised of end-to-end data
which characterizes the quality of the paths which Voice Over Internet Protocol (VOIP) telephone calls take
across the global Internet and contains anonymized Session Initiation Protocol (SIP) teardown messages
collected from both ends of the conversation. There are three primary pieces of information:
1. Prefix representing logical grouping
2. Autonomous system number of the ISPs at both ends of the connection.
3. Country code of Autonomous System Numbers (ASN).
All three pieces of information are currently publicly available.
This data is very useful to study the quality and security of VOIP calls. It is anticipated that the VOIP
statistical data will be used by researchers who wish to compare differential quality of service in similar and
dissimilar regions of the Internet, such as across different backbone carriers which utilize different
technology or capacity-planning methodologies. These comparisons of quality of service including
statistical analysis of the VOIP data (e.g., call duration, call frequency, termination endpoints) could
indicate potential security weaknesses within the VOIP network. The data can also be analyzed to determine

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 14

the availability of the VOIP network in the event of an attack to see if emergency communications can be
accomplished over the Internet using the VOIP network.

Firewall Logs
Firewalls detect distributed denial of service (DDOS) attacks and other malicious activity. Firewall
logs contain detailed information regarding the end point that is directing harmful activity towards the
network they are protecting. They contain the number of packets, origin, and where it went. All IP
addresses will be anonymized.

Traffic Flows via Netflow
This data consists of statistics regarding data collected from routers. It identifies two end points.
Raw NetFlow data will indicate not only traffic totals, but the application breakdowns at each peering
point. The individual flows will be stored in a method compatible with free analysis tools. All IP addresses
will be anonymized. This dataset can be used to develop tools and techniques that will lead to a better
understanding of the tradeoffs in netflow versus other techniques and in improving netflow and new
netflow versions. It can also be used in the development of anomaly detection and intrusion detection
applications and for traffic characterization applications, such as self similarity and bottleneck bandwidth
estimation applications.

Network Management Data
Network management data is generally represented in the form of messages that are transmitted
using Simple Network Management Protocol (SNMP). The messages convey gross aggregated statistical
information from various network devices. It is possible to study Internet behavior by analyzing these
messages. SNMP messages are also used to transmit information conveying the physical well-being of these
devices. They often reveal important causal information that is subsequently observed in other types of
data logs. They do not pose an individual privacy risk, but can reveal information regarding the operation
of the network. Anonymization will protect critical network operator information. The data provider
currently makes this data publicly available.

Intrusion Detection System (IDS) Logs
An intrusion detection system scans traffic to detect unauthorized or malicious activity. When it
detects an attack, it can trigger protective actions. It is essentially a sensor that is watching for malicious
activity. It is important to study IDS traffic in order to understand the evolution, rise, and decay of such
traffic. It is possible to identify the end point responsible for originating the suspicious activity.

Anonymized Internet Witty Worm Data
The data provider owns a large number of IP addresses. Due to its size, this network receives a
large volume of unsolicited traffic trying to spread random-scan Internet worms. Packet header traces and
tables summarizing infected host activity contain detailed information regarding the end point that is
directing harmful activity towards the network. They contain the truncated packets (copies of the worm,
no other data) and summarized information on the number of packets, country-level origin, and type of
bandwidth. All IP addresses will be anonymized.

Code-Red Worm Data

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 15

The data provider owns a large number of IP addresses. Due to its size, this network receives a
large volume of unsolicited traffic trying to spread random-scan Internet worms. Tables summarizing
infected host activity contain detailed information regarding the end point that is directing harmful activity
towards the network. They contain the number of packets, country-level origin, and type of bandwidth.
All IP addresses will be anonymized.

1.2

What are the sources of the information in the system?

•

Researchers

•

Data providers

•

Data hosts

•

ARB/PRB members

•

Advisory Board members

•

Sponsoring Institutions

•

Academic institutions and private sector entities provide traffic data. No PREDICT Phase I data
is provided by communications providers to the public.

1.3

Why is the information being collected, used,
disseminated, or maintained?

1.3.1 Researcher Information
The information collected about researchers is essential to maintaining operational records
indicating who has used specific PREDICT datasets, for what purpose, when, and what organization
sponsored them. Additional collected information pertains to whether the researcher’s application was
approved or disapproved and whether any publication requests were approved or disapproved. Reasons for
disapproval of requests for access to certain datasets or publication are kept by the PCC. This information is
very useful to the Application Review Board and Publication Review Board and helps them ensure fairness
and even treatment. PREDICT operational procedures ensure that the denial of access or publication
requests is not used against a researcher when making subsequent requests.

1.3.2 Data Provider Information
The information regarding data providers is important to the operations of the PCC. The data
providers must be notified when a researcher is requesting their data or requests permission to publish
regarding the research conducted using the datasets. Additionally, the data providers are active members of
the Application Review Board and Publication Review Board and set the terms and conditions for use of
their data. Thus, it is important that the PCC maintain accurate information for each data provider and
which datasets they are providing to the PCC.

1.3.3 Data Host information
The information collected and maintained about data hosts is also critical to PCC operations. The
data hosts must be notified when a researcher is approved to access data sets that they are hosting. The data
host must also provide any special terms or conditions that they want the researcher to comply with in

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 16

accessing the data. It is important that the PCC maintain accurate information for each data host and the
datasets which they are hosting for the PREDICT project to enable the PCC to operate correctly and conduct
audits of compliance with PCC policies and procedures.

1.3.4 Application Review Board and Publication Review Board
Member Information
The PCC must maintain contact information of the people who serve as Application Review Board
and Publication Review Board members so they can contact them for assistance and decisions regarding
applications and requests to publish.

1.3.5 Advisory Board Member Information
The PCC collects contact information from the people who serve on the Advisory Board so they can
contact them with information regarding upcoming meetings, minutes of past meetings, questions that
need to be considered, and the like.

1.3.

Sponsoring Institution Information

The PCC collects the contact information of the person representing the institution who is
sponsoring the research so it can contact the institution to verify the researcher’s application, if necessary,
resolve any questions that may arise, or conduct an audit of the institution’s compliance with PCC policies
and procedures and the terms of the MOAs.

1.3.7 Dataset Information
The datasets in the PREDICT project are collected in the form of test datasets to enable researchers
to use them in the research and development of cyber security technologies that will better protect the
security of networks and applications and the privacy of information.

1.4

How is the information collected?

PREDICT administrative information is collected through the PREDICT portal and via email and
facsimile. The PREDICT test datasets are either retained hosted by the provider or the provider sends its
approved test datasets to an approved data host.

1.5

How will the information be checked for accuracy?

PREDICT administrative information is provided directly by the person submitting it (e.g.,
researcher, data provider, data host, sponsoring institution, or advisory, ARB, or PRB members). The PCC
verifies the information that has been provided prior to it being entered into the system and, for
researchers, prior to assigning them an account. Communications between the PCC and these persons
helps ensure the information remains accurate.

1.6

What specific legal authorities, arrangements, and/or
agreements defined the collection of information?

PREDICT provides a Privacy Statement to users to inform them about how their personal
information will be used and the purpose for which it will be used. Appendix A is attached with a copy of

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 17

the Privacy Statement. The MOAs between the PCC and researchers, data providers and data hosts and
corresponding provisions are discussed in the Overview section above. S&T/CCI has established contracts
with the PCC for the administration of PREDICT and with the data hosts for the hosting of the test datasets.

1.7

Privacy Impact Analysis: Given the amount and type of
data collected, discuss the privacy risks identified and how
they were mitigated.

There is a risk associated with the collection of personally identifiable information. To mitigate this
risk, the PCC has integrated robust security into its risk management plans and routinely tests the security
of PREDICT. Further, the PCC has implemented best practices, such as auditing, to defend against misuse of
the data and to monitor those with access to the information. In addition, in accordance with the DHS
Sensitive Systems Handbook, PREDICT ensures effective security controls and authentication. By enforcing
system policies and settings and strong passwords, the PCC protects the privacy of data to promote or
permit public access to the PREDICT project and to protect the integrity of the data itself. Also, any contact
information collected by the PCC is not searchable. Contact information is only secondary data and is used
by PCC to validate the user requesting an account and access to PREDICT data or for administrative
purposes.
PREDICT test datasets have privacy protections built in through the review process and required
anonymization of certain fields. These protections are described for each dataset in 1.1.5 above regarding
“Dataset Information.”

1.7.1 Data About Researchers
The privacy of the data about researchers will be maintained through authentication and
authorization controls, regular backups, PREDICT policies and procedures (including change management),
and audits of PREDICT operations regarding compliance with policies and procedures and operational
terms and conditions.

1.7.2 Data About Data Providers
The privacy of the data about data providers will be maintained through authentication and
authorization controls, regular backups, PREDICT policies and procedures (including change management),
and audits of PREDICT operations regarding compliance with policies and procedures and operational
terms and conditions.

1.7.3 Data About Data Hosts
The privacy of data about data hosts will be maintained through authentication and authorization
controls, regular backups, PREDICT policies and procedures (including change management), and audits of
PREDICT operations regarding compliance with policies and procedures and operational terms and
conditions.

1.7.4 Data About Application Review Board Members
The privacy of the data about Application Review Board members will be maintained through
authentication and authorization controls, regular backups, PREDICT policies and procedures (including

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 18

change management), and audits of PREDICT operations regarding compliance with policies and
procedures and operational terms and conditions.

1.7.5 Data About Publication Review Board Members
The privacy of the data about Publication Review Board members will be maintained through
authentication and authorization controls, regular backups, PREDICT policies and procedures (including
change management), and audits of PREDICT operations regarding compliance with policies and
procedures and operational terms and conditions.

1.7.6 Data About Advisory Board Members:
The privacy of the data about Advisory Board members will be maintained through authentication
and authorization controls, regular backups, PREDICT policies and procedures (including change
management), and audits of PREDICT operations regarding compliance with policies and procedures and
operational terms and conditions.

1.7.7 Data About Sponsoring Institutions:
The privacy of the data about Sponsoring Institutions will be maintained through authentication
and authorization controls, regular backups, PREDICT policies and procedures (including change
management), and audits of PREDICT operations regarding compliance with policies and procedures and
operational terms and conditions.

1.7.8 Dataset Information:
The PREDICT datasets will be secured by their respective PREDICT data hosts and will only be
accessed by approved researchers who have been provided with terms and conditions for accessing the
datasets, downloading them, storing them, and destroying or returning them. The quality of the PREDICT
datasets will be maintained through authentication and authorization controls, regular backups, PREDICT
policies and procedures (including MOAs regarding the use and handling of the datasets), and audits of
PREDICT operations regarding compliance with policies and procedures and operational terms and
conditions.
The dataset will be downloaded to the sponsoring institution site where the researcher is working.
The researcher is bound contractually to use the data only as specified and within the bounds of the
research described in the application. In addition, the data provider may require an additional written
agreement between the provider and researcher setting forth specific rights and restrictions regarding the
use of the data.

Section 2.0 Uses of the Information
The following questions are intended to delineate clearly the use of information and the
accuracy of the data being used.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 19

2.1

Describe all the uses of information.

2.1.1 The Intended Use of Information Collected From Researchers,
Data Providers, Data Hosts:
The specific intended use of the information collected from researchers, data providers, and data
hosts is described above (see 1.3.1, 1.3.2, 1.3.3). This information will be used to manage PREDICT
operations and conduct periodic audits to ensure PREDICT policies and procedures and operational terms
and conditions are being followed.

2.1.2 The Intended Use of Information Collected From Application
Review Board, Publication Review Board, and Advisory Board
Members:
The specific intended use of the information collected from the Application Review Board,
Publication Review Board, and Advisory Board members is described above (see 1.3.4 and 1.3.5). This
information will be used to facilitate and manage PREDICT operations and may also be used from time to
time in periodic audits to check compliance with PREDICT policies and procedures and operational terms
and conditions.

2.1.3 The Intended Use of Information Collected from Sponsoring
Institutions:
The specific intended use of the information collected from Sponsoring Institutions is described
above (see 1.3.6). This information may also be used from time to time in periodic audits to check
compliance with PREDICT policies and procedures and operational terms and conditions.

2.1.4 The Intended Use of the Datasets:
The intended use of the datasets will be specified by the researcher in their application. The
application is reviewed by the Application Review Board, which consists of a PCC representative, a data
provider representative, a data host representative, and an external cyber security expert. The researcher
must use the data for the purposes stated in the application and in accordance with the restrictions specified
by the data provider. The Application Review Board operates under a Non-Disclosure Agreement to protect
the researchers’ privacy and confidential/proprietary information contained in the researchers’ applications.

2.2

What types of tools are used to analyze data and what type
of data may be produced?

There are no tools, such as data mining tools, being used to analyze PREDICT information, and no
data is produced.

2.3

If the system uses commercial or publicly available data
please explain why and how it is used.

As noted in 1.1.5, “Dataset Information,” some PREDICT datasets are already publicly available and
some are provided by commercial entities, non-profit organizations, or academic institutions.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 20

2.4

Privacy Impact Analysis: Describe any types of controls
that may be in place to ensure that information is handled
in accordance with the above described uses.

Information collected by the PCC is secured within PCC operations and will only be accessed by
PCC personnel with a valid need-to-know for the information. The PCC is comprised of a small staff that
has clear delineations in responsibility for managing the information collected from these sources. The PCC,
in accordance with the DHS Sensitive Systems Handbook, ensures effective security controls and
authentication mechanisms are implemented and working as intended.
2.4.1 Administrative Data
As noted in 1.7 above, the PCC has implemented auditing checkpoints, tested controls, and
deployed technical safeguards to protect administrative data. With respect to test datasets, PREDICT privacy
requirements are embedded in the MOA between the data provider and the PCC. Likewise, terms for
access, special handling and privacy/security that are set by the data providers are embedded in the MOA
between the researcher and the PCC. These terms are also embedded in the MOA between the data host
and the PCC.
2.4.2 Dataset Information
The PREDICT project has analyzed each dataset and has directed data providers to anonymize any
data that has the potential to be personally identifiable information, such as IP addresses. Anonymization of
such data fields must comply with requirements set forth by the PCC and DHS. Researchers must agree to
establish and maintain the appropriate administrative, technical, and physical safeguards to protect the
privacy and confidentiality of the data and to prevent unauthorized use or access to the data. This
specifically includes the use of locked storage facilities and strong passwords.
Additional security
provisions for access to the host system and the dataset may be included as terms of agreement in the
researcher’s application (which becomes the MOA between the researcher and the PCC). In addition, the
data provider and data host may require a separate agreement with the researcher regarding security and
use of the data. To ensure the long-term integrity of the datasets and efficacy of PREDICT policies and
procedures, the PCC has established an advisory board that provides input on an ongoing basis regarding
PREDICT operations, the security and integrity of the datasets, and privacy issues and concerns. In
addition, periodic compliance checks of provider and researcher compliance will be conducted to ensure all
parties are in compliance with PREDICT policies and procedures and the terms of the MOAs. As the
PREDICT program expands, supplemental PIAs will be developed to accommodate new datasets and to
maintain transparency of operations. The Certification and Accreditation (C&A) has been completed by the
S&T CIO effective September 30, 2007.

Section 3.0 Retention
The following questions are intended to outline how long information will be retained after the
initial collection.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 21

3.1

How long is information retained?

3.1.1 Data Collected from Researchers, Data Providers, Data Hosts,
ARB and PRB Members, Advisory Board Members, and
Sponsoring Institutions:
The PCC maintains all data collected from researchers, data providers, data hosts, ARB and PRB
members, Advisory Board members, and sponsoring institutions and this information is retained for six
years from the time they are no longer associated with the PREDICT project. This data retention is done
offline as is specified in the System Security Plan (SSP).

3.1.2 Datasets
Upon termination of the right to use the data (when the researcher completes his/her research;
when the researcher terminates his/her right to use the data; or when the PCC, data provider or data host
terminates the researcher’s right to use the data), the researcher must destroy all copies of the data or,
where directed by PCC, return the data to the data host or destroy it per its instructions. The researcher
must certify such destruction or return by signing and providing to PCC a Certification of Data Return or
Destruction document.
The PREDICT data hosts maintain the datasets they host until such time that the data provider, data
host, or the PCC removes them from the PREDICT project.

3.2

Has the retention schedule been approved by the
component records officer and the National Archives and
Records Administration (NARA)?

The retention schedule for the PREDICT administrative information has not been approved by the
National Archives and Records Administration, but S&T is working with NARA to obtain a retention
schedule.

3.3

Privacy Impact Analysis: Please discuss the risks
associated with the length of time data is retained and how
those risks are mitigated.

Retention of administrative records is per DHS policies and procedures as laid out in FISMA
Certification and Accreditation.

Section 4.0 Internal Sharing and Disclosure
The following questions are intended to define the scope of sharing within the Department of
Homeland Security.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 22

4.1

With which internal organization(s) is the information
shared, what information is shared and for what purpose?

Neither PREDICT administrative information nor the test datasets are within the Department of
Homeland Security. DHS does not have access to this data.

4.2

How is the information transmitted or disclosed?

Not applicable as DHS does not have access to PREDICT data.

4.3

Privacy Impact Analysis: Considering the extent of internal
information sharing, discuss the privacy risks associated
with the sharing and how they were mitigated.

Not applicable as DHS does not have access to PREDICT data.

Section 5.0 External Sharing and Disclosure
The following questions are intended to define the content, scope, and authority for
information sharing external to DHS which includes Federal, state and local government, and the
private sector.

5.1

With which external organization(s) is the information
shared, what information is shared, and for what purpose?

5.1.1 Information About Researchers:
The information collected about researchers will be shared with PCC personnel. Portions of it may
be shared on an as-needed basis with Application Review Board and Publication Review Board members;
data providers of the requested datasets; data hosts that host the requested datasets; personnel outside of
PCC who may be conducting audits of PCC operations and compliance with PCC policies and procedures;
and contractors working on PREDICT in a consulting capacity regarding PCC operations. In rare
circumstances, the information about a particular researcher may be shared with the PREDICT project
manager in the Department of Homeland Security.

5.1.2 Information About Data Providers:
Information collected from data providers will be shared on an as-needed basis with the PCC
personnel; Application Review Board and the Publication Review Board members; data hosts who will be
hosting the data provider’s data; other data providers; personnel outside of PCC who may be conducting
audits of PCC operations and compliance with PCC policies and procedures; contractors working on
PREDICT in a consulting capacity regarding PCC operations; and the PREDICT project manager in the
Department of Homeland Security.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 23

5.1.3 Information About Data Hosts:
Information collected from data hosts will be shared on an as-needed basis with the PCC personnel;
Application Review Board; data providers who will be using the data host to host their data; other data
hosts; personnel outside of PCC who may be conducting audits of PCC operations and compliance with
PCC policies and procedures; contractors working on PREDICT in a consulting capacity regarding PCC
operations; and the PREDICT project manager in the Department of Homeland Security.

5.1.4 Information About Application Review Board Members:
Information collected from Application Review Board members will be shared on an as-needed
basis with the PCC personnel; other Application Review Board members; data providers who need the
information to interact with the ARB; personnel outside of PCC who may be conducting audits of PCC
operations and compliance with PCC policies and procedures; contractors working on PREDICT in a
consulting capacity regarding PCC operations; and the PREDICT project manager in the Department of
Homeland Security.

5.1.5 Information About Publication Review Board Members:
Information collected from Publication Review Board members will be shared on an as-needed
basis with the PCC personnel; other Publication Review Board members; data providers who need the
information to interact with the PRB; personnel outside of PCC who may be conducting audits of PCC
operations and compliance with PCC policies and procedures; contractors working on PREDICT in a
consulting capacity regarding PCC operations; and the PREDICT project manager in the Department of
Homeland Security.

5.1.6 Information About Advisory Board Members:
Information collected from Advisory Board members will be shared on an as-needed basis with the
PCC personnel; other Advisory Board members; personnel outside of PCC who may be conducting audits of
PCC operations and compliance with PCC policies and procedures; contractors working on PREDICT in a
consulting capacity regarding PCC operations; and the PREDICT project manager in the Department of
Homeland Security.

5.1.7 Information About Sponsoring Institutions:
Information collected from Sponsoring Institutions will be shared on an as-needed basis with the
PCC personnel; Application Review Board and Publication Review Board members; data providers who
need the information to interact with the ARB or PRB; personnel outside of PCC who may be conducting
audits of PCC operations and compliance with PCC policies and procedures; contractors working on
PREDICT in a consulting capacity regarding PCC operations; and the PREDICT project manager in the
Department of Homeland Security.

5.1.8 Dataset Information:
The PREDICT datasets will be shared with approved researchers who are affiliated with a
sponsoring institution and who have completed and signed an application (which also serves as an MOA)
setting forth the research they will be conducting and agreeing to the terms of use of the dataset as set forth
by the data provider and data host and the terms of PREDICT. If a researcher leaves their sponsoring

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 24

institution, the rights to use the dataset are terminated. 4 The researcher may, at that time, submit a new
application through the new sponsoring institution to continue using the data. Once the new application is
approved, the researcher will be allowed to continue using the data. At no time will DHS or PCC have
access to the data.

5.2

Is the sharing of personally identifiable information outside
the Department compatible with the original collection? If
so, is it covered by an appropriate routine use in a SORN?
If so, please describe. If not, please describe under what
legal mechanism the program or system is allowed to
share the personally identifiable information outside of
DHS.

The PREDICT administrative information is never inside DHS. It is collected and maintained solely
by the PCC. As noted in the Overview, this data collection is covered by the DHS System of Records Notice
002 published December 6, 2004 69 Fed Reg 70460 DHS/ALL 002.

5.3

How is the information shared outside the Department and
what security measures safeguard its transmission?

How PCC administrative information is shared outside the PCC is set forth in 5.1.1 through 5.1.7.
Only the minimum amount of information is shared to facilitate PREDICT operations. The ARB, PRB, and
Advisory board members sign non-disclosure agreements with the PCC. Internal and external audits also
test control points to monitor the effectiveness of the controls and the security of shared information.

5.4

Privacy Impact Analysis: Given the external sharing,
explain the privacy risks identified and describe how they
were mitigated.

The risks to PREDICT PII were identified through a review and analysis of process flow and
controls conducted by PCC personnel, audit personnel, PREDICT contractors, and the DHS PREDICT project
manager. In addition, in accordance with the DHS Sensitive Systems Handbook, PREDICT ensures effective
security controls and authentication. By enforcing system policies, settings, and strong passwords, PCC
mitigates risks of disclosure of PII when shared.

4

If the researcher is also the sponsoring institution (e.g., a corporation who is acting as both sponsoring
institution and researcher) and a researcher leaves the organization, that organization continues to have the right to
use the data and may submit the name of another researcher.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 25

Section 6.0 Notice
The following questions are directed at notice to the individual of the scope of
information collected, the right to consent to uses of said information, and the right to decline to
provide information.

6.1

Was notice provided to the individual prior to collection of
information?

General notice of this collection is covered by the DHS System of Records Notice 002 published
December 6, 2004 69 Fed Reg 70460 DHS/ALL 002. Additionally, The PREDICT portal provides a Terms
of Use and Privacy Statements for general users to review prior to registration for access. This user
agreement allows for the individual data users to consent to the sharing of their information with the data
providers. The PREDICT portal also contains the PREDICT Operations Manual, which details PREDICT
operations and how information is used.

6.2

Do individuals have the opportunity and/or right to decline
to provide information?

All individuals requesting access to PREDICT information have the opportunity and right to decline
to provide personally identifiable information, however with their declination they will not have access to
PREDICT and its information.

6.2.1 Researchers, Data Providers, and Data Hosts
All researchers, data providers, and data hosts must have a PREDICT account, which requires them
to complete all the requested information listed in 1.1, except the home, cell, and fax telephone numbers
are optional. As part of the user agreement, the researcher consents to their information being shared with
the data providers and the data hosts so as to facilitate the information flow. If all of the required fields are
not provided, they may not participate in the PREDICT project. The information collected is used only for
PREDICT operations.

6.2.2 Application Review Board, Publication Review Board, and
Advisory Board Members
All Application Review Board, Publication Review Board, and Advisory Board members must
provide the information set forth above in 1.1 or they may not participate in the PREDICT project. There is
no procedure to enable any person associated with one of these Boards to decline to provide information or
consent to only particular uses of the information.

6.2.3 Sponsoring Institutions
All Sponsoring Institutions must provide the requested information set forth above in 1.1 or their
researchers may not participate in PREDICT. There is no procedure for a sponsoring institution to decline
to provide the required information or to consent to only particular uses of the information.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 26

6.2.4 Datasets
The data provider must consent to its data being used in PREDICT. DHS must also consent to the
inclusion of any dataset in the PREDICT project. Prior to DHS granting such consent, it undertakes a review
of data provider privacy policies and privacy risks associated with the dataset. Once a dataset has been
accepted by DHS, no consent beyond that of the Application Review Board is required in order for
researchers to use the datasets. Each data provider is responsible for the release of their data to the PREDICT
project and is solely responsible for ensuring that any data it releases complies with all applicable statutes
and regulations of applicable governing or regulating bodies and contractual agreements, and is consistent
with the data provider’s privacy, security, or other applicable policies and procedures. The only data being
provided is non-content data and all personally identifiable forms of data are being anonymized, except for
topology and measurement data that does not encompass sender-receiver communication. The data
provider also agrees through the MOA that the data provided for use in PREDICT has been sanitized, deidentified, or cleaned of any and all information that would not be in compliance or consistent with any
policies, procedures, and understandings applicable to the data, whether explicit or implicit.

6.3

Do individuals have the right to consent to particular uses
of the information? If so, how does the individual exercise
the right?

By submitting their information, the individual agrees that their information can be used in
accordance with the PREDICT Privacy Statement, Operations Manual, and policies and procedures. All
persons participating in the PREDICT initiative must abide by all of the Terms of Use and Privacy statement
as well as PREDICT policies and procedures.

6.4

Privacy Impact Analysis: Describe how notice is provided
to individuals, and how the risks associated with
individuals being unaware of the collection are mitigated.

The PREDICT portal provides a Terms of Use and Privacy Statement for general users to review
prior to registration for access and on every page after entry to the portal. The end user must also agree to
the following every time they access the PREDICT portal:
WARNING This system is for the use of authorized users only. Individuals using
this computer system without authority, or in excess of their authority, are subject
to having all of their activities on this system monitored and recorded by system
personnel. In the course of monitoring individuals improperly using this system,
or in the course of system maintenance, the activities of authorized users may also
be monitored. Anyone using this system expressly consents to such monitoring
and is advised that if such monitoring reveals possible evidence of criminal
activity, system personnel may provide the evidence of such monitoring to law
enforcement officials.
Access to PREDICT is voluntary. Users must agree to follow the Terms of use, Privacy Statements,
and PREDICT policies and procedures each time they log into the portal.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 27

Section 7.0 Access, Redress and Correction
The following questions are directed at an individual’s ability to ensure the accuracy of
the information collected about them.

7.1

What are the procedures that allow individuals to gain
access to their information?

Registered users have a user profile. The profile contains contact information such as name,
address, phone numbers, etc. Users have limited access to their profile - they can view and update the
contact information but cannot change the roles they belong to. They can not access the profile of any
other user. Users update their contact via a web form; the input is protected by an input validation scheme.
All user-input text is tested for suspicious entries such as SQL keywords and scripting. Offending text is
rejected and discarded.
PREDICT does not allow an individual to access or correct any data other than the data the
individual has submitted him/herself.
General users (i.e., those who are not registered users) do not have access to any data and no
information is collected about them until they submit an account request to become a registered user.

7.2

What are the procedures for correcting inaccurate or
erroneous information?

An individual can submit a written request to amend his or her record. After the receipt of the
amendment request, the PCC acknowledges in writing that it has either: (a) corrected any information
which the individual asserts is not accurate, relevant, timely, or complete; or (b) informs the individual of
our refusal to amend in accordance with the request, the reason for refusal, and the procedures for
administrative appeal. If the individual disagrees with the PCC’s refusal to amend the information and
requests a review, the PCC will comply with requirements for a review pursuant to 5 U.S.C. §§ 552a(c)
(4), (d) (3), and (d) (4).

7.3

How are individuals notified of the procedures for
correcting their information?

Individuals are notified via e-mail that the request for Access to PREDICT and or the information
hosted by PREDICT, has been denied. Denial of access to PREDICT and the information hosted by PREDICT
can result from and is not limited to the following:
Incorrect:
•

First and last name

•

Address (street, city, state, zip code)

•

Telephone number

•

Email address

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 28

•

Name of organization

•

Title

•

Researcher does not have the right criteria for requested data set.

In these cases the user is requested to provide updated/corrected information and re-submit
his/her request.

7.4

If no formal redress is provided, what alternatives are
available to the individual?

PREDICT provides redress to its users, see 7.1 and 7.3.

7.5

Privacy Impact Analysis: Please discuss the privacy risks
associated with the redress available to individuals and
how those risks are mitigated.

Privacy risks are mitigated through the opportunities to access, correct and seek redress regarding
information in PREDICT.

Section 8.0 Technical Access and Security
The following questions are intended to describe technical safeguards and security
measures.

8.1

What procedures are in place to determine which users
may access the system and are they documented?

The PREDICT Operations Manual is available on the PREDICT portal. It documents the operational
process and procedures of the PREDICT initiative, including access to datasets. The process flow for
PREDICT operations has been documented and reviewed by PCC personnel, consultants, data providers, and
the DHS PREDICT project manager.

8.2

Will Department contractors have access to the system?

Yes, the entire PREDICT operational structure is handled by contractors and a volunteer community
of data providers. PCC staff provides operation and maintenance of PREDICT. Contractors to DHS host the
PREDICT datasets. Other contractors provide expertise regarding privacy, procedural, and audit issues.
Only those contractors that have a valid need for access will be granted access to the PREDICT portal.

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 29

8.3

Describe what privacy training is provided to users either
generally or specifically relevant to the program or
system?

Personal information is gathered and maintained in accordance with pertinent Federal rules and
regulations, including the Privacy Act of 1974, the E-Government Act of 2002, the Government Paperwork
Reduction Act, and other relevant rules and regulations. Questions relating to privacy are forwarded from
the PREDICT Information Systems Security Officer to the DHS S&T liaison to the DHS Privacy Office.
Training is being conducted to ensure that individuals working with personal identifiable information are
educated on its proper handling and use.

8.4

Has Certification & Accreditation been completed for the
system or systems supporting the program?

Yes. The Certification and Accreditation (C&A) has been completed by the S&T CIO effective
September 30, 2007.

8.5

What auditing measures and technical safeguards are in
place to prevent misuse of data?

PREDICT project staff ensures audit trails and audit logs are recorded and retained in accordance
with the Homeland Security Department Records and Information Management Program. In addition,
auditing and intrusion detection capabilities are provided through firewalls and server logs, which alert the
PCC staff. Proactive scanning and monitoring of logs and events on a daily basis assists in identifying
incidents as early as possible to mitigate potential issues. An audit trail is maintained and stored to facilitate
investigation of incidents. Incident reporting policies will include not only the DHS Information System
Security Manager (ISSM), but also the DHS Systems Operations Center (SOC) and US-CERT where
appropriate.

8.6

Privacy Impact Analysis: Given the sensitivity and scope of
the information collected, as well as any information
sharing conducted on the system, what privacy risks were
identified and how do the security controls mitigate them?

8.6.1 Administrative Data
There is the risk that inadvertent access to PREDICT administrative data containing PII may occur.
To mitigate this risk, and in accordance with the DHS Sensitive Systems Handbook, the PCC ensures
effective security controls and authentication. By enforcing system policies and settings and strong
passwords, PREDICT project staff protects the privacy of PII and the integrity of the data itself.
Requirements Identification: PREDICT project staff will include the following policies, practices, guidance,
and legal requirements for this process: DHS 4300 - IT Systems Security - Sensitive Systems Pub - Vol I Part
A; Federal Information Security Management Act of 2002; OMB Circular A-130 Appendix III, Security of
Federal Automated Information Systems; Computer Security Act of 1987; OMB Circular A-11, Preparation

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 30

and Submission of Budget Estimates; Presidential Decision Memorandum (PDD-63), Critical Infrastructure
Protection; National Institute of Standards and Technology (NIST) Special Publication (SP) 800-55, Security
Metrics Guide for Information Technology Systems.

8.6.2 Datasets
As noted in 1.7.8, the PREDICT datasets will be secured by their respective PCC data hosts and will
only be accessed by approved researchers who have been provided with terms and conditions for accessing
the datasets, downloading them, storing them, and destroying or returning them. The quality of the
PREDICT datasets will be maintained through authentication and authorization controls, regular backups,
PREDICT policies and procedures (including MOAs regarding the use and handling of the datasets), and
audits of PREDICT operations and compliance with policies and procedures and operational terms and
conditions.
The dataset will be downloaded to the sponsoring institution site where the researcher is working.
The researcher is bound contractually to use the data only as specified and within the bounds of the
research described in the application. In addition, the data provider may require an additional written
agreement between the provider and researcher setting forth specific rights and restrictions regarding the
use of the data.

Section 9.0 Technology
The following questions are directed at critically analyzing the selection process for any
technologies utilized by the system, including system hardware, RFID, biometrics and other technology.

9.1

What type of project is the program or system?

The PREDICT portal is a unique, custom web-based system developed by the PCC. The portal also
includes COTS products as part of the system. The security of all of these products has been assessed during
the system C&A.

9.2

What stage of development is the system in and what
project development lifecycle was used?

The PREDICT system is operational.

9.3

Does the project employ technology which may raise
privacy concerns? If so please discuss their
implementation.

PREDICT uses the latest in technology to ensure that personal information, along with the rest of
the system's data, is secured against unauthorized access. All communications with PREDICT are through
encrypted means (SSL). There is no part of PREDICT that is accessible without first authenticating with a
username and password. Physical access to the room where the PREDICT server is housed is controlled
through key card access. All access to PREDICT is logged and maintained for future reference if necessary.
The use of secure encryption protocols were chosen not only for personal information protection but also

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 31

to ensure that all information within PREDICT is secured while being transmitted. Access to user
information is restricted to very few personnel and logging of all access to that information is conducted
within the servers.

Approval Signature Page

Original signed and on file with the DHS Privacy Office
Hugo Teufel III
Chief Privacy Officer
Department of Homeland Security

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 32

Appendix A – Privacy Statement
PREDICT is committed to protecting your privacy and developing technology that gives you the most powerful and
safe online experience. This Statement of Privacy applies to the PREDICT Web site and governs data collection and
usage of that site. By using the PREDICT website, you consent to the data practices described in this statement.
Collection of your Personal Information
PREDICT collects personally identifiable information, such as your e-mail address, name, home or work address or
telephone number.
There is also information about your computer hardware and software that is automatically collected by PREDICT.
This information can include: your IP address, user name, browser type, and access times. This information is used
by PREDICT for the operation of the service, to maintain quality of the service, and to provide general statistics
regarding use of the PREDICT Web site.
PREDICT encourages you to review the privacy statements of Web sites that you may choose to link to from
PREDICT so that you can understand how those Web sites collect, use and share your information. PREDICT is not
responsible for the privacy statements or other content on Web sites outside of the PREDICT web site.
Use of your Personal Information
PREDICT collects and uses your personal information to operate the PREDICT Web site and deliver the services you
have requested. PREDICT may also contact you via surveys to conduct research about your opinion of current
services or of potential new services that may be offered.
PREDICT does not sell, rent or lease data it collects to third parties. PREDICT may share data with data hosts, data
providers, and review board members of the PREDICT community to help us deliver requested PREDICT services.
These parties are prohibited from using your personal information except to provide these requested services, and
they are required to maintain the confidentiality of your information.
PREDICT will disclose your personal information, without notice, only if required to do so by law or in the good faith
belief that such action is necessary to: (a) conform to the edicts of the law or comply with legal processes served on
PREDICT or the site; (b) protect and defend the rights or property of PREDICT; and, (c) act under exigent
circumstances to protect the personal safety of users of PREDICT, or the public.
Use of Cookies
The PREDICT Web site requires the use of non-persistent (session-based) "cookies" during an active session for
proper operation of pages for registered PREDICT users during that one session. The purpose of the cookies is to
identify a user to the PREDICT web server only when accessing content within the PREDICT domain. PREDICT does
not use “ad” cookies to track user activity. PREDICT does not retain any cookie data. You have the ability to accept
or decline cookies. Most Web browsers automatically accept cookies, but you can usually modify your browser setting
to decline cookies if you prefer. If you choose to decline cookies, you may not be able to fully experience the
interactive features of the PREDICT services.
Security of your Personal Information
PREDICT secures your personal information from unauthorized access, use or disclosure. PREDICT secures the
personally identifiable information you provide on computer servers in a controlled, secure environment, protected
from unauthorized access, use or disclosure.
Changes to this Statement

Privacy Impact Assessment
S&T/CCI, PREDICT
Page 33

PREDICT will occasionally update this Statement of Privacy to reflect feedback or changes in operations or policy.
PREDICT encourages you to periodically review this Statement to be informed of how PREDICT is protecting your
information.
Contact Information
PREDICT welcomes your comments regarding this Statement of Privacy. If you believe that PREDICT has not
adhered to this Statement, please contact PREDICT at [email protected]. We will use commercially
reasonable efforts to promptly determine and remedy the problem.