💾 Archived View for gemini.bortzmeyer.org › rfc-mirror › rfc8192.txt captured on 2023-09-08 at 17:12:17.
⬅️ Previous capture (2021-11-30)
-=-=-=-=-=-=-
Internet Engineering Task Force (IETF) S. Hares
Request for Comments: 8192 Huawei
Category: Informational D. Lopez
ISSN: 2070-1721 Telefonica I+D
M. Zarny
vArmour
C. Jacquenet
France Telecom
R. Kumar
Juniper Networks
J. Jeong
Sungkyunkwan University
July 2017
Interface to Network Security Functions (I2NSF):
Problem Statement and Use Cases
Abstract
This document sets out the problem statement for Interface to Network
Security Functions (I2NSF) and outlines some companion use cases.
Status of This Memo
This document is not an Internet Standards Track specification; it is
published for informational purposes.
This document is a product of the Internet Engineering Task Force
(IETF). It represents the consensus of the IETF community. It has
received public review and has been approved for publication by the
Internet Engineering Steering Group (IESG). Not all documents
approved by the IESG are a candidate for any level of Internet
Standard; see Section 2 of RFC 7841.
Information about the current status of this document, any errata,
and how to provide feedback on it may be obtained at
http://www.rfc-editor.org/info/rfc8192.
Hares, et al. Informational [Page 1]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Copyright Notice
Copyright (c) 2017 IETF Trust and the persons identified as the
document authors. All rights reserved.
This document is subject to BCP 78 and the IETF Trust's Legal
Provisions Relating to IETF Documents
(http://trustee.ietf.org/license-info) in effect on the date of
publication of this document. Please review these documents
carefully, as they describe your rights and restrictions with respect
to this document. Code Components extracted from this document must
include Simplified BSD License text as described in Section 4.e of
the Trust Legal Provisions and are provided without warranty as
described in the Simplified BSD License.
Hares, et al. Informational [Page 2]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Table of Contents
1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . 4
2. Terminology . . . . . . . . . . . . . . . . . . . . . . . . . 5
3. Problem Space . . . . . . . . . . . . . . . . . . . . . . . . 6
3.1. Challenges Facing Security Service Providers . . . . . . 6
3.1.1. Diverse Types of Security Functions . . . . . . . . . 6
3.1.2. Diverse Interfaces to Control and Monitor NSFs . . . 8
3.1.3. More Distributed NSFs and vNSFs . . . . . . . . . . . 8
3.1.4. More Demand to Control NSFs Dynamically . . . . . . . 9
3.1.5. Demand for Multi-tenancy to Control and Monitor NSFs 9
3.1.6. Lack of Characterization of NSFs and Capability
Exchange . . . . . . . . . . . . . . . . . . . . . . 9
3.1.7. Lack of Mechanism for NSFs to Utilize External
Profiles . . . . . . . . . . . . . . . . . . . . . . 10
3.1.8. Lack of Mechanisms to Accept External Alerts to
Trigger Automatic Rule and Configuration Changes . . 10
3.1.9. Lack of Mechanism for Dynamic Key Distribution to
NSFs . . . . . . . . . . . . . . . . . . . . . . . . 10
3.2. Challenges Facing Customers . . . . . . . . . . . . . . . 12
3.2.1. NSFs from Heterogeneous Administrative Domains . . . 12
3.2.2. Today's Vendor-Specific Control Requests . . . . . . 13
3.2.3. Difficulty for Customers to Monitor the Execution of
Desired Policies . . . . . . . . . . . . . . . . . . 14
3.3. Lack of Standard Interface to Inject Feedback to NSF . . 15
3.4. Lack of Standard Interface for Capability Negotiation . . 15
3.5. Difficulty in Validating Policies across Multiple Domains 15
3.6. Software-Defined Networks . . . . . . . . . . . . . . . . 16
4. Use Cases . . . . . . . . . . . . . . . . . . . . . . . . . . 17
4.1. Basic Framework . . . . . . . . . . . . . . . . . . . . . 17
4.2. Access Networks . . . . . . . . . . . . . . . . . . . . . 18
4.3. Cloud Data Center Scenario . . . . . . . . . . . . . . . 21
4.3.1. On-Demand Virtual Firewall Deployment . . . . . . . . 21
4.3.2. Firewall Policy Deployment Automation . . . . . . . . 22
4.3.3. Client-Specific Security Policy in Cloud VPNs . . . . 22
4.3.4. Internal Network Monitoring . . . . . . . . . . . . . 23
4.4. Preventing DDoS, Malware, and Botnet Attacks . . . . . . 23
4.5. Regulatory and Compliance Security Policies . . . . . . . 24
5. Management Considerations . . . . . . . . . . . . . . . . . . 24
6. IANA Considerations . . . . . . . . . . . . . . . . . . . . . 24
7. Security Considerations . . . . . . . . . . . . . . . . . . . 24
8. Informative References . . . . . . . . . . . . . . . . . . . 25
Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 27
Contributors . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . 28
Hares, et al. Informational [Page 3]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
1. Introduction
This document sets out the problem statement for Interface to Network
Security Functions (I2NSF) and outlines some use cases. A summary of
the state of the art in the industry and IETF that is relevant to
I2NSF work is documented in [I2NSF-ANALYSIS].
The growing challenges and complexity in maintaining a secure
infrastructure, complying with regulatory requirements, and
controlling costs are enticing enterprises into consuming network
security functions hosted by service providers. The hosted security
service is especially attractive to small- and medium-size
enterprises which suffer from a lack of security experts to
continuously monitor networks, acquire new skills, and propose
immediate mitigations to ever increasing sets of security attacks.
According to [Gartner], the demand for hosted (or cloud-based)
security services is growing. Small- and medium-size businesses
(SMBs) are increasingly adopting cloud-based security services to
replace on-premises security tools, while larger enterprises are
deploying a mix of traditional and cloud-based security services.
To meet the demand, more and more service providers are providing
hosted security solutions to deliver cost-effective managed security
services to enterprise customers. The hosted security services are
primarily targeted at enterprises (especially small and medium ones)
but could also be provided to any kind of mass-market customer. As a
result, the Network Security Functions (NSFs) are provided and
consumed in a large variety of environments. Users of NSFs may
consume network security services hosted by one or more providers,
which may be their own enterprise, service providers, or a
combination of both.
This document also briefly describes the following use cases
summarized by [I2NSF-USECASES]:
o I2NSF Access Use Cases [OAM-USECASE],
o I2NSF Data Center Use Cases [DC-USECASE], and
o Integrated Security with Access Network Use Case [ACCESS-USECASE].
Hares, et al. Informational [Page 4]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
2. Terminology
AAA: Authentication, Authorization, and Accounting [RFC2904]
ACL: Access Control List
Bespoke security management: Security management that is made to fit
a particular customer.
DC: Data Center
FW: Firewall
IDS: Intrusion Detection System
IPS: Intrusion Protection System
I2NSF: Interface to Network Security Functions
NSF: Network Security Function. An NSF is a function that is used
to ensure integrity, confidentiality, or availability of network
communication; to detect unwanted network activity; or to block,
or at least mitigate, the effects of unwanted activity.
Flow-based NSF: An NSF that inspects network flows according to a
security policy. Flow-based security also means that packets are
inspected in the order they are received and without altering
packets due to the inspection process (e.g., Medium Access Control
(MAC) rewrites, TTL decrement action, or NAT inspection or
changes). (Note: Some existing firewalls store packets and look
at the packets in logical order, which is not the order these are
received in time. This document restricts flow-based NSF to this
definition.)
Security service provider: A provider of security services to the
customers (end users or enterprises) using NSF equipment purchased
from vendors or created by the service provider.
SDN: Software-Defined Networking. (See [RFC7426] for architecture
and terminology or [RFC7149] for a service provider view.)
vCPE: virtual Customer Premises Equipment
vEPC: virtual Evolved Packet Core [EPC-3GPP]
vNSF: Virtual NSF. An NSF that is deployed as a distributed virtual
resource.
Hares, et al. Informational [Page 5]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
vPE: virtual Provider Edge
VPN: Virtual Private Network
3. Problem Space
The following sub-sections describe the problems and challenges
facing customers and security service providers when some or all of
the security functions are no longer physically hosted by the
customer's administrative domain.
Security service providers can be internal or external to the
company. For example, an internal IT security group within a large
enterprise could act as a security service provider for the
enterprise. In contrast, an enterprise could outsource all security
services to an external security service provider. In this document,
the security service provider function, whether it is internal or
external, will be denoted as "service provider".
The "Customer-Provider" relationship may be between any two parties.
The parties can be in different organizations or different domains of
the same organization. Contractual agreements may be required in
such contexts to formally document the customer's security
requirements and the provider's guarantees to fulfill those
requirements. Such agreements may detail protection levels,
escalation procedures, alarms reporting, etc. There is currently no
standard mechanism to capture those requirements.
A service provider may be a customer of another service provider.
It is the objective of the I2NSF work to address these problems and
challenges.
3.1. Challenges Facing Security Service Providers
3.1.1. Diverse Types of Security Functions
There are many types of NSFs. NSFs by different vendors can have
different features and interfaces. NSFs can be deployed in multiple
locations in a given network and perhaps have different roles.
Below are a few examples of security functions and locations or
contexts in which they are often deployed:
External Intrusion and Attack Protection: Examples of this function
are firewall/ACL authentication, IPS, IDS, and endpoint
protection.
Hares, et al. Informational [Page 6]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Security Functions in a Demilitarized Zone (DMZ): Examples of this
function are firewall/ACLs, IDS/IPS, one or all of AAA services,
NAT, forwarding proxies, and application filtering. These
functions may be physically on-premise in a server provider's
network at the DMZ spots or located in a "virtual" DMZ.
Centralized or Distributed Security Functions: The security
functions could be deployed in a centralized fashion for ease of
management and network design or in a distributed fashion for
scaled requirement. No matter how a security function is deployed
and provisioned, it is desirable to have the same interface to
provision security policies; otherwise, the job of security
administration is more complex, requiring knowledge of firewall
and network design.
Internal Security Analysis and Reporting: Examples of this function
are security logs, event correlation, and forensic analysis.
Internal Data and Content Protection: Examples of this function are
encryption, authorization, and public/private key management for
internal databases.
Security Gateways and VPN Concentrators: Examples of these functions
are IPsec gateways, secure VPN concentrators that handle bridging
secure VPNs, and secure VPN controllers for data flows.
Given the diversity of security functions, the contexts in which
these functions can be deployed, and the constant evolution of these
functions, standardizing all aspects of security functions is
challenging and probably not feasible. Fortunately, it is not
necessary to standardize all aspects. For example, from an I2NSF
perspective, there is no need to standardize how every firewall's
filtering is created or applied. Some features in a specific
vendor's filtering may be unique to the vendor's product, so it is
not necessary to standardize these features.
What is needed is a standardized interface to control and monitor the
rule sets that NSFs use to treat packets traversing through these
NSFs. Thus, standardizing interfaces will provide an impetus for
standardizing established security functions.
I2NSF may specify some filters, but these filters will be linked to
specific common functionality developed by I2NSF in information
models or data models.
Hares, et al. Informational [Page 7]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
3.1.2. Diverse Interfaces to Control and Monitor NSFs
To provide effective and competitive solutions and services, security
service providers may need to utilize multiple security functions
from various vendors to enforce the security policies desired by
their customers.
Since no widely accepted industry standard interface to NSFs exists
today, management of NSFs (device and policy provisioning,
monitoring, etc.) tends to be custom-made security management offered
by product vendors. As a result, automation of such services, if it
exists at all, is also custom made. Thus, even in the traditional
way of deploying security features, there is a gap that needs to be
filled; this would require coordination among implementations from
distinct vendors.
A challenge for monitoring prior to mitigation of a security
intrusion is that an NSF cannot monitor what it cannot view. For
example, enabling a security function to mitigate an intrusion (e.g.,
firewall [FIREWALLS]) must include a mechanism to provide monitoring
feedback in order to determine the intrusion has been stopped.
Therefore, it is necessary to have a mechanism to monitor and provide
execution status of NSFs to security and compliance management tools.
Such mechanisms exist in vendor-specific network security interfaces
for forensics and troubleshooting, but an industry standard interface
could provide monitoring across a variety of NSFs.
3.1.3. More Distributed NSFs and vNSFs
The security functions that are invoked to enforce a security policy
can be located in different equipment and network locations.
The European Telecommunications Standards Institute (ETSI) Network
Functions Virtualization (NFV) initiative [ETSI-NFV] creates new
management challenges for security policies to be enforced by
distributed vNSFs.
A vNSF has higher risk of changes to the state of network connection,
interfaces, or traffic, as their hosting Virtual Machines (VMs) are
being created, moved, or decommissioned.
Hares, et al. Informational [Page 8]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
3.1.4. More Demand to Control NSFs Dynamically
In the advent of Software-Defined Networking (SDN) (see
[SDN-SECURITY]), more clients, applications, or application
controllers need to dynamically update their security policies that
are enforced by NSFs. The security service providers have to
dynamically update their decision-making process (e.g., in terms of
NSF resource allocation and invocation) upon receiving security-
related requests from their clients.
3.1.5. Demand for Multi-tenancy to Control and Monitor NSFs
Service providers may need to deploy several NSF controllers to
control and monitor the NSFs, especially when NSFs become distributed
and virtualized.
3.1.6. Lack of Characterization of NSFs and Capability Exchange
To offer effective security services, service providers need to
activate various security functions in NSFs or vNSFs manufactured by
multiple vendors. Even within one product category (e.g., firewall),
security functions provided by different vendors can have different
features and capabilities. For example, filters that can be designed
and activated by a firewall may or may not support IPv6 depending on
the firewall technology.
The service provider's management system (or controller) needs a way
to retrieve the capabilities of service functions by different
vendors so that it can build an effective security solution. These
service function capabilities can be documented in a static manner
(e.g., a file) or via an interface that accesses a repository of
security function capabilities that the NSF vendors dynamically
update.
A dynamic capability registration is useful for automation because
security functions may be subject to software and hardware updates.
These updates may have implications on the policies enforced by the
NSFs.
Today, there is no standard method for vendors to describe the
capabilities of their security functions. Without a common technical
framework to describe the capabilities of security functions, service
providers cannot automate the process of selecting NSFs by different
vendors to accommodate customers' security requirements.
The I2NSF work will focus on developing a standard method to describe
capabilities of security functions.
Hares, et al. Informational [Page 9]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
3.1.7. Lack of Mechanism for NSFs to Utilize External Profiles
Many security functions depend on signature files or profiles (e.g.,
IPS/IDS signatures and DDoS Open Threat Signaling (DOTS) filters).
Different policies might need different signatures or profiles.
Today, blacklist databases can be a beneficial strategy for all
parties involved (except the attackers), but in the future, there
might be open-source signatures and profiles distributed as part of
IDS systems (e.g., by Snort, Suricata, Bro, and Kismet).
There is a need to have a standard envelope (i.e., a message format)
to allow NSFs to use external profiles.
3.1.8. Lack of Mechanisms to Accept External Alerts to Trigger
Automatic Rule and Configuration Changes
NSFs can ask the I2NSF security controller to alter specific rules
and/or configurations. For example, a Distributed Denial of Service
(DDoS) alert could trigger a change to the routing system to send
traffic to a traffic scrubbing service to mitigate the DDoS.
The DDoS protection has two parts: a) the configuration of signaling
of open threats and b) DDoS mitigation. The DOTS controller manages
the signaling part of DDoS. I2NSF controller(s) would control any
changes to affected policies (e.g., forwarding and routing,
filtering, etc.). By monitoring the network alerts regarding DDoS
attacks (e.g., from DOTS servers or clients), the I2NSF controller(s)
can feed an alerts analytics engine that could recognize attacks so
the I2NSF can enforce the appropriate policies.
DDoS mitigation is enhanced if the provider's network security
controller can monitor, analyze, and investigate the abnormal events
and provide information to the customer or change the network
configuration automatically.
[CAP-INTERFACE] provides details on how monitoring aspects of the
flow-based Network Security Functions (NSFs) can use the I2NSF
interfaces to receive traffic reports and enforce appropriate
policies.
3.1.9. Lack of Mechanism for Dynamic Key Distribution to NSFs
There is a need for a controller to create, manage, and distribute
various keys to distributed NSFs. While there are many key
management methods and cryptographic suites (e.g., encryption
algorithms, key derivation functions, etc.) and other functions,
there is a lack of a standard interface to provision and manage
security associations.
Hares, et al. Informational [Page 10]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
The keys may be used for message authentication and integrity in
order to protect data flows. In addition, keys may be used to secure
the protocols and messages in the core routing infrastructure (see
[RFC4948]).
As of now, there is not much focus on an abstraction for keying
information that describes the interface between protocols,
operators, and automated key management.
An example of a solution may provide some insight into why the lack
of a mechanism is a problem. If a device had an abstract key table
maintained by security services, it could use these keys for routing
and security devices.
What does this take?
Conceptually, there must be an interface defined for routing/
signaling protocols that can a) make requests for automated key
management when it is being used and b) notify the protocols when
keys become available in the key table. One potential use of such an
interface is to manage IPsec security associations on Software-
Defined Networks.
An abstract key service will work under the following conditions:
1. I2NSF needs to design the key table abstraction, the interface
between key management protocols and routing/other protocols, and
possibly security protocols at other layers.
2. For each routing/other protocol, I2NSF needs to define the
mapping between how the protocol represents key material and the
protocol-independent key table abstraction. If several protocols
share common mechanisms for authentication (e.g., TCP
Authentication Option [RFC5925]), then the same mapping may be
used for all usages of that mechanism.
3. Automated key management needs to support both pairwise keys and
group keys via the abstract key service provided by items 1 and
2. I2NSF controllers within the NSF that are required to
exchange data with NSFs may exchange data with individual NSFs
using individual pairwise keys or with a group of NSFs
simultaneously using an IP group address secured by a group
security key(s).
Hares, et al. Informational [Page 11]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
3.2. Challenges Facing Customers
When customers invoke hosted security services, their security
policies may be enforced by a collection of security functions hosted
in different domains. Customers may not have the security skills to
express sufficiently precise requirements or security policies.
Usually, these customers express the expectations of their security
requirements or the intent of their security policies. These
expectations can be considered customer-level security expectations.
Customers may also desire to express guidelines for security
management. Examples of such guidelines include:
o which critical communications are to be preserved during critical
events and which hosts will continue services over the network,
o what signaling information is passed to a controller during a DDoS
in order to ask for mitigation services (within the scope of the
DOTS Working Group),
o reporting of attacks to CERT (within the scope of the MILE Working
Group), and
o managing network connectivity of systems out of compliance (within
the scope of the SACM Working Group).
3.2.1. NSFs from Heterogeneous Administrative Domains
Many medium and large enterprises have deployed various on-premises
security functions that they want to continue to use. These
enterprises want to combine local security functions with remote
hosted security functions to achieve more efficient and immediate
countermeasures to attacks originating on both the Internet and
enterprise networks.
Some enterprises may only need the hosted security services for their
remote branch offices where minimal security infrastructures/
capabilities exist. The security solution will consist of deploying
NSFs on customer networks and on service provider networks.
Hares, et al. Informational [Page 12]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
3.2.2. Today's Vendor-Specific Control Requests
Customers may utilize NSFs provided by multiple service providers.
Customers need to express their security requirements, guidelines,
and expectations to the service providers. In turn, the service
providers must translate this customer information into customer
security policies and associated configuration tasks for the set of
security functions in their network. Without a standardized
interface that provides a clear technical characterization, the
service provider faces many challenges:
No standard technical characterization, APIs, or interface(s):
Even for the most common security services, there is no standard
technical characterization, APIs, or interface(s). Most security
services are accessible only through disparate, proprietary
interfaces (e.g., portals or APIs) in whatever format vendors
choose to offer. The service provider must process the customer's
input with these widely varying interfaces and differing
configuration models for security devices and security policy.
Without a standard interface, new innovative security products
find a large barrier to entry into the market.
Lack of immediate feedback: Customers may also require a mechanism
to easily update/modify their security requirements with immediate
effect in the underlying involved NSFs.
Lack of explicit invocation request: While security agreements are
in place, security functions may be solicited without requiring an
explicit invocation means. Nevertheless, some explicit invocation
means may be required to interact with a service function.
Managing by scripts du jour: The current practices rely upon the use
of scripts that generate other scripts, which automatically run to
upload or download configuration changes, log information, and
other things. These scripts have to be adjusted each time an
implementation from a different vendor technology is enabled by a
provider.
To see how standard interfaces could help achieve faster
implementation time cycles, let us consider a customer who would like
to dynamically allow an encrypted flow with a specific port, src/dst
addresses, or protocol type through the firewall/IPS to enable an
encrypted video conferencing call only during the time of the call.
With no commonly accepted interface in place, as shown in Figure 1,
the customer would have to learn about the particular provider's
firewall/IPS interface and send the request in the provider's
required format.
Hares, et al. Informational [Page 13]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
+------------+
| Security |
| Management |
| System |
+----||------+
|| Proprietary
|| or I2NSF Standard
Video: ||
Port 10 +--------+
--------| FW/IPS |-------------
Encrypted +--------+
Video Flow
Figure 1: Example of Non-standard vs. Standard Interface
In contrast, as Figure 1 shows, if a firewall/IPS interface standard
exists, the customer would be able to send the request to a security
management system, and the security management would send it via a
I2NSF standard interface. Service providers could now utilize the
same standard interface to represent firewall/IPS services
implemented using products from many vendors.
3.2.3. Difficulty for Customers to Monitor the Execution of Desired
Policies
How a policy is translated into technology-specific actions is hidden
from the customers. However, customers still need ways to monitor
the delivered security service that results from the execution of
their desired security requirements, guidelines, and expectations.
Customers want to monitor existing policies to determine such things
as which policies are in effect, how many security attacks are being
prevented, and how much bandwidth efficiency does security
enforcement cost.
Today, there is no standard way for customers to get these details
from the security service. As a consequence, there is no way to
assure customers that their specified security policies are properly
enforced by the security functions located in the provider domain.
Customers also want this monitoring information from the security
system in order to plan for the future using "what-if" scenarios with
real data. A tight loop between the data gathered from security
systems and the "what-if" scenario planning can reduce the time to
design and deploy workable security policies that deal with new
threats.
It is the objective of the I2NSF work to provide a standard way to
get the information that security service assurance systems need to
Hares, et al. Informational [Page 14]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
provide customers an evaluation about the current security systems
and to quickly plan for future security policies using "what-if"
scenarios based on today's information.
3.3. Lack of Standard Interface to Inject Feedback to NSF
Today, many security functions in the NSF, such as IPS, IDS, DDoS
mitigation, and antivirus, depend heavily on the associated profiles.
NSF devices can perform more effective protection if these NSF
devices have the up-to-date profiles for these functions. Today,
there is no standard interface to provide these security profiles for
the NSF.
As more sophisticated threats arise, protection will depend on
enterprises, vendors, and service providers being able to cooperate
to develop optimal profiles; one example of this cooperation is the
Cyber Threat Alliance [CTA]. The standard interface to provide
security profiles to the NSF should interwork with the formats that
exchange security profiles between organizations.
One objective of the I2NSF work is to provide this type of standard
interface to security profiles.
3.4. Lack of Standard Interface for Capability Negotiation
There could be situations when the selected NSFs cannot perform the
policies requested by the security controller due to resource
constraints. The customer and security service provider should
negotiate the appropriate resource constraints before the security
service begins. However, unexpected events may happen that cause the
NSF to exhaust those negotiated resources. At this point, the NSF
should inform the security controller that the allotted resources
have been exhausted. To support the automatic control in the SDN
era, it is necessary to have a set of messages for proper
notification (and a response to that notification) between the
security controller and the NSFs.
3.5. Difficulty in Validating Policies across Multiple Domains
As discussed in the previous four sub-sections, both service
providers and customers have need to express policies and profiles,
monitor systems, verify security policy has been installed in NSFs
within a security domain, and establish limits for services NSFs can
safely perform. This sub-section and the next sub-section
(Section 3.6) examine what happens in two specific network scenarios:
a) multi-domain control of security devices hosted on virtual and
non-virtual NSFs and b) Software-Defined Networking.
Hares, et al. Informational [Page 15]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Hosted security service may instantiate NSFs in virtual machines that
are sometimes widely distributed in the network and sometimes are
combined together in one device to perform a set of tasks for
delivering a service. Hosted security services may be connected
within a single service provider or via multiple service providers.
Ensuring that the security service purchased by the customer adheres
to customer policy requires that the central controller(s) for this
service monitor and validate this service across multiple networks on
NSFs (some of which may be virtual networks on virtual machines). To
set up this cross-domain service, the security controller must be
able to communicate with NSFs and/or controllers within its domain
and across domains to negotiate for the services needed.
Without standard interfaces and security policy data models, the
enforcement of a customer-driven security policy remains challenging
because of the inherent complexity created by combining the
invocation of several vendor-specific security functions into a
multi-vendor, heterogeneous environment across multiple domains.
Each vendor-specific function may require specific configuration
procedures and operational tasks.
Ensuring the consistent enforcement of the policies at various
domains is also challenging. Standard data models are likely to
contribute to solving that issue.
3.6. Software-Defined Networks
Software-Defined Networks have changed the landscape of data-center
designs by introducing overlay networks deployed over Top-of-Rack
(ToR) switches that connect to a hypervisor. SDN techniques are
meant to improve the flexibility of workload management without
affecting applications and how they work. Workload can thus be
easily and seamlessly managed across private and public clouds. SDN
techniques optimize resource usage and are now being deployed in
various networking environments besides cloud infrastructures. Yet,
such SDN-conferred agility may raise specific security issues. For
example, a security administrator must make sure that a security
policy can be enforced regardless of the location of the workload,
thereby raising concerns about the ability of SDN computation logic
to send security policy-provisioning information to the participating
NSFs. A second example is workload migration to a public cloud
infrastructure, which may raise additional security requirements
during the migration.
Hares, et al. Informational [Page 16]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
4. Use Cases
Standard interfaces for monitoring and controlling the behavior of
NSFs are essential building blocks for security service providers and
enterprises to automate the use of different NSFs from multiple
vendors by their security management entities. I2NSF may be invoked
by any (authorized) client. Examples of authorized clients are
upstream applications (controllers), orchestration systems, and
security portals.
4.1. Basic Framework
Users request security services through specific clients (e.g., a
customer application, the Business Support Systems / Operations
Support Systems (BSSs/OSSs) of Network Service Providers (NSPs), or a
management platform), and the appropriate NSP network entity will
invoke the (v)NSFs according to the user service request. This
network entity is denoted as the security controller in this
document. The interaction between the entities discussed above
(client, security controller, and NSF) is shown in Figure 2:
+----------+
+-------+ | | +-------+
| | Interface 1 |Security | Interface 2 | NSF(s)|
|Client <--------------> <------------------> |
| | |Controller| | |
+-------+ | | +-------+
+----------+
Figure 2: Interaction between Entities
Interface 1 is used for receiving security requirements from a client
and translating them into commands that NSFs can understand and
execute. The security controller also passes back NSF security
reports (e.g., statistics) to the client that the security controller
has gathered from NSFs. Interface 2 is used for interacting with
NSFs according to commands (e.g., enact/revoke a security policy or
distribute a policy) and collecting status information about NSFs.
Client devices or applications can require the security controller to
add, delete, or update rules in the security service function for
their specific traffic.
When users want to get the executing status of a security service,
they can request NSF status from the client. The security controller
will collect NSF information through Interface 2, consolidate it, and
give feedback to the client through Interface 1. This interface can
Hares, et al. Informational [Page 17]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
be used to collect not only individual service information, but also
aggregated data suitable for tasks like infrastructure security
assessment.
Customers may require validating NSF availability, provenance, and
execution. This validation process, especially relevant to vNSFs,
includes at least:
Integrity of the NSF: Ensuring that the NSF is not compromised;
Isolation: Ensuring the execution of the NSF is self-contained for
privacy requirements in multi-tenancy scenarios; and
Provenance of the NSF: Customers may need to be provided with strict
guarantees about the origin of the NSF, its status (e.g.,
available, idle, down, and others), and feedback mechanisms so
that a customer may be able to check that a given NSF or set of
NSFs properly conform to the customer's requirements and
subsequent configuration tasks.
In order to achieve this, the security controller may collect
security measurements and share them with an independent and trusted
third party (via Interface 1) in order to allow for attestation of
NSF functions using the third-party added information.
This implies that there may be the following two types of clients
using Interface 1: the end user and the trusted, independent third
party. The I2NSF work may determine that Interface 1 creates two
sub-interfaces to support these two types of clients.
4.2. Access Networks
This scenario describes use cases for users (e.g., residential user,
enterprise user, mobile user, and management system) that request and
manage security services hosted in the NSP infrastructure. Given
that NSP customers are essentially users of their access networks,
the scenario is essentially associated with their characteristics as
well as with the use of vNSFs. Figure 3 shows how different types of
customers connect through virtual access nodes (vCPE, vPE, and vEPC)
to an NSF.
The vCPE described in use case #7 in [NFVUC] requires a model of
access virtualization that includes mobile and residential access
networks where the operator may offload security services from the
customer's local environment (e.g., device or terminal) to its own
infrastructure.
Hares, et al. Informational [Page 18]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
These use cases define the interaction between the operator and the
vNSFs through automated interfaces that support the business
communications between customer and provider or between two business
entities.
Customer + Access + PoP / Data Center
| | +--------+
| ,-----+--. |Network |
| ,' | `-|Operator|
+-------------+ | /+----+ | |Mgmt Sys|
| Residential |-+------/-+vCPE+----+ +--------+
+-------------+ | / +----+ | \ | :
| / | \ | |
+----------+ | ; +----+ | +----+ |
|Enterprise|---+---+----+ vPE+--+----+ NSF| |
+----------+ | : +----+ | +----+ |
| : | / |
+--------+ | : +----+ | / ;
| Mobile |-+-----\--+vEPC+----+ /
+--------+ | \ +----+ | Service /
| `--. | Provider /
| `----+---------..
+ + ^^
||
Service Provider
encompasses
everything
in circle
vCPE - virtual customer premises equipment
vPE - virtual provider edge
vEPC - virtual evolved packet core
PoP - point of presence
Figure 3: NSF and Actors
Different access clients may have different service requests:
Residential: service requests for parental control, content
management, and threat management.
Threat content management may include identifying and blocking
malicious activities from web contents, mail, or files downloaded.
Threat management may include identifying and blocking botnets or
malware.
Enterprise: service requests for enterprise flow security policies
and managed security services.
Hares, et al. Informational [Page 19]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Flow security policies identify and block malicious activities
during access to (or isolation from) web sites or social media
applications. Managed security services for an enterprise may
include detection and mitigation of external and internal threats.
External threats can include application or phishing attacks,
malware, botnet, DDoS, and others.
Service Provider: service requests for policies that protect service
provider networks against various threats (including DDoS,
botnets, and malware). Such policies are meant to securely and
reliably deliver contents (e.g., data, voice, and video) to
various customers, including residential, mobile, and corporate
customers. These security policies are also enforced to guarantee
isolation between multiple tenants, regardless of the nature of
the corresponding connectivity services.
Mobile: service requests from interfaces that monitor and ensure
user quality of experience, content management, parental controls,
and external threat management.
Content management for the mobile device includes identifying and
blocking malicious activities from web contents, mail, and files
uploaded/downloaded. Threat management for infrastructure
includes detecting and removing malicious programs such as botnet,
malware, and other programs that create DDoS attacks).
Some access customers may not care about which NSFs are utilized to
achieve the services they requested. In this case, provider network
orchestration systems can internally select the NSFs (or vNSFs) to
enforce the security policies requested by the clients.
Other access customers, especially some enterprise customers, may
want to contract separately for dedicated NSFs (most likely vNSFs)
for direct control purposes. In this case, here are the steps to
associate vNSFs to specific customers:
vNSF Deployment: The deployment process consists of instantiating an
NSF on an NFV Infrastructure (NFVI), within the NSP administrative
domain(s) or with other external domain(s). This is a required
step before a customer can subscribe to a security service
supported in the vNSF.
vNSF Customer Provisioning: Once a vNSF is deployed, any customer
can subscribe to it. The provisioning life cycle includes the
following:
* Customer enrollment and cancellation of the subscription to a
vNSF.
Hares, et al. Informational [Page 20]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
* Configuration of the vNSF, based on specific configurations or
derived from common security policies defined by the NSP.
* Retrieval of the vNSF functionalities, extracted from a
manifest or a descriptor. The NSP management systems can
demand this information to offer detailed information through
the commercial channels to the customer.
4.3. Cloud Data Center Scenario
In a data center, network security mechanisms such as firewalls may
need to be dynamically added or removed for a number of reasons.
These changes may be explicitly requested by the user or triggered by
a pre-agreed-upon demand level in the Service Level Agreement (SLA)
between the user and the provider of the service. For example, the
service provider may be required to add more firewall capacity within
a set of time frames whenever the bandwidth utilization hits a
certain threshold for a specified period. This capacity expansion
could result in adding new instances of firewalls on existing
machines or provisioning a completely new firewall instance in a
different machine.
The on-demand, dynamic nature of security service delivery
essentially encourages that the network security "devices" be in
software or virtual forms rather than in a physical appliance form.
This requirement is a provider-side concern. Users of the firewall
service are agnostic (as they should be) as to whether or not the
firewall service is run on a VM or any other form factor. Indeed,
they may not even be aware that their traffic traverses firewalls.
Furthermore, new firewall instances need to be placed in the "right
zone" (domain). The issue applies not only to multi-tenant
environments where getting the tenant in the right domain is of
paramount importance, but also in environments owned and operated by
a single organization with its own service segregation policies. For
example, an enterprise may mandate that firewalls serving Internet
traffic within the organization be separated from inter-organization
traffic. Another example is IPS/IDS services that split investment
banking traffic from other data traffic to comply with regulatory
restrictions for transfer of investment banking information.
4.3.1. On-Demand Virtual Firewall Deployment
A cloud data center operated by a service provider could serve tens
of thousands of clients. Clients' compute servers are typically
hosted on VMs, which could be deployed across different server racks
located in different parts of the data center. It is often not
technically and/or financially feasible to deploy dedicated physical
Hares, et al. Informational [Page 21]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
firewalls to suit each client's security policy requirements, which
can be numerous. What is needed is the ability to dynamically deploy
virtual firewalls for each client's set of servers based on
established security policies and underlying network topologies.
Figure 4 shows an example topology of virtual firewalls within a data
center.
---+-----------------------------+-----
| |
+---+ +-+-+
|vFW| |vFW|
+---+ +-+-+
| Client #1 | Client #2
---+-------+--- ---+-------+---
+-+-+ +-+-+ +-+-+ +-+-+
|VM | |VM | |VM | |VM |
+---+ +---+ +---+ +---+
Figure 4: NSF in Data Centers
4.3.2. Firewall Policy Deployment Automation
Firewall rules apply to traffic usually identified with addresses and
ports. It becomes far more complex in provider-owned cloud networks
that serve myriads of customers.
Firewall rules today are highly tied with ports and addresses that
identify traffic. This makes it very difficult for clients of cloud
data centers to construct rules for their own traffic, as the clients
only see the virtual networks and the virtual addresses. The
customer-visible virtual networks and addresses may be different from
the actual packets traversing the firewalls.
Even though most vendors support similar firewall features, the
specific rule configuration keywords are different from vendor to
vendor, making it difficult for automation. Automation works best
when it can leverage a common set of standards that will work across
NSFs by multiple vendors and utilize dynamic key management.
4.3.3. Client-Specific Security Policy in Cloud VPNs
Clients of cloud data centers operated by a service provider need to
secure Virtual Private Networks (VPNs) and virtual security functions
that apply the clients' security policies. The security policies may
govern communication within the clients' own virtual networks as well
as communication with external networks. For example, VPN service
providers may need to provide firewall and other security services to
their VPN clients. Today, it is generally not possible for clients
Hares, et al. Informational [Page 22]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
to dynamically view (let alone change) what, where, and how security
policies are implemented on their provider-operated clouds. Indeed,
no standards-based framework exists to allow clients to retrieve/
manage security policies in a consistent manner across different
providers.
As described above, the dynamic key management is critical for
securing the VPN and the distribution of policies.
4.3.4. Internal Network Monitoring
There are many types of internal traffic monitors that may be managed
by a security controller. This includes the class of services
referred to as Data Loss Prevention (DLP) or Reputation Protection
Services (RPS). Depending on the class of event, alerts may go to
internal administrators or external services.
4.4. Preventing DDoS, Malware, and Botnet Attacks
On the Internet, where everything is connected, preventing unwanted
traffic that may cause a DoS attack or a DDoS attack has become a
challenge. Similarly, a network could be exposed to malware attacks
and become an attack vector that may jeopardize the operation of
other networks, by means of remote commands for example. Many
networks that carry groups of information (such as Internet of Things
(IoT) networks, Information-Centric Networks (ICNs), Content Delivery
Networks (CDNs), Voice over IP (VoIP) packet networks, and Voice over
LTE (VoLTE)) are also exposed to such remote attacks. There are many
examples of remote attacks on these networks, but the following
examples will illustrate the issues. A malware attack on an IoT
network that carries sensor readings and instructions may attempt to
alter the sensor instructions in order to disable a key sensor. A
malware attack on VoIP or VoLTE networks involves software that
attempts to place unauthorized long-distance calls. Botnets may
overwhelm nodes in ICNs and CDNs so that the networks cannot pass
critical data.
In order for organizations to better secure their networks against
these kind of attacks, the I2NSF framework should provide a client-
side interface that is use case independent and technology agnostic.
Technology agnostic is defined to be generic, technology independent,
and able to support multiple protocols and data models. For example,
such an I2NSF interface could be used to provision security policy
configuration information that looks for specific malware signatures.
Similarly, botnet attacks could be easily prevented by provisioning
security policies using the I2NSF client-side interface that prevents
access to botnet command and control servers.
Hares, et al. Informational [Page 23]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
4.5. Regulatory and Compliance Security Policies
Organizations must protect their networks against attacks and must
also adhere to various industry regulations: any organization that
falls under a specific regulation, like the Payment Card Industry -
Data Security Standard (PCI-DSS) [PCI-DSS] for the payment industry
or the Health Insurance Portability and Accountability Act [HIPAA]
for the healthcare industry, must be able to isolate various kinds of
traffic. They must also show records of their security policies
whenever audited.
The I2NSF client-side interface could be used to provision regulatory
and compliance-related security policies. The security controller
would keep track of when and where a specific policy is applied and
if there is any policy violation; this information can be provided in
the event of an audit as proof that traffic is isolated between
specific endpoints, in full compliance with the required regulations.
5. Management Considerations
Management of NSFs usually include the following:
o Life-cycle management and resource management of NSFs,
o Device configuration, such as address configuration, device
internal attributes configuration, etc.,
o Signaling of events, notifications, and changes, and
o Policy rule provisioning.
I2NSF will only focus on the policy provisioning part of NSF
management.
6. IANA Considerations
This document does not require any IANA actions.
7. Security Considerations
Having secure access to control and monitor NSFs is crucial for
hosted security services. An I2NSF security controller raises new
security threats. It needs to be resilient to attacks and quickly
recover from them. Therefore, proper secure communication channels
have to be carefully specified for carrying, controlling, and
monitoring traffic between the NSFs and their management entity (or
entities).
Hares, et al. Informational [Page 24]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
The traffic flow security policies specified by customers can
conflict with providers' internal traffic flow security policies.
This conflict can be resolved in one of two ways: a) installed
policies can restrict traffic if either the customer traffic flow
security policies or the provider's internal security policies
restrict traffic, or b) installed policies can only restrict traffic
if both the customer traffic flow security policies and the
provider's internal traffic flow security policies restrict data.
Either choice could cause potential problems. It is crucial for the
management system to flag these conflicts to the customers and to the
service provider.
It is important to proper AAA [RFC2904] to authorize access to the
network and access to the I2NSF management stream.
Enforcing the appropriate privacy is key to all IETF protocols (see
[RFC6973]) and is especially important for IETF security management
protocols since they are deployed to protect the network. In some
circumstances, security management protocols may be utilized to
protect an individual's home, phone, or other personal data. In this
case, any solution should carefully consider whether combining
management streams abides by the recommendations of [RFC6973] for
data minimization, user participation, and security.
8. Informative References
[ACCESS-USECASE]
Wang, K. and X. Zhuang, "Integrated Security with Access
Network Use Case", Work in Progress,
draft-qi-i2nsf-access-network-usecase-02, March 2015.
[CAP-INTERFACE]
Zhou, C., Xia, L., Boucadair, M., and J. Xiong, "The
Capability Interface for Monitoring Network Security
Functions (NSF) in I2NSF", Work in Progress,
draft-zhou-i2nsf-capability-interface-monitoring-00,
October 2015.
[CTA] "Cyber Threat Alliance", <http://cyberthreatalliance.org>.
[DC-USECASE]
Zarny, M., Majee, S., Leymann, N., and L. Dunbar, "I2NSF
Data Center Use Cases", Work in Progress,
draft-zarny-i2nsf-data-center-use-cases-00, October 2014.
[EPC-3GPP] Firmin, F., "The Evolved Packet Core", January 2017.
Hares, et al. Informational [Page 25]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
[ETSI-NFV] ETSI, "Network Functions Virtualisation (NFV);
Architectural Framework", ETSI GS NFV 002 V1.2.1, December
2014.
[FIREWALLS]
Baker, F. and P. Hoffman, "On Firewalls in Internet
Security", Work in Progress,
draft-ietf-opsawg-firewalls-01, October 2012.
[Gartner] Messmer, E., "Gartner: Cloud-based security as a service
set to take off", October 2013.
[HIPAA] US Congress, "Health Insurance Portability and
Accountability Act of 1996 (Public Law 104-191)", August
1996, <https://www.hhs.gov/hipaa/>.
[I2NSF-ANALYSIS]
Hares, S., Moskowitz, R., and D. Zhang, "Analysis of
Existing work for I2NSF", Work in Progress,
draft-ietf-i2nsf-gap-analysis-03, March 2017.
[I2NSF-USECASES]
Pastor, A., Lopez, D., Wang, K., Zhuang, X., Qi, M.,
Zarny, M., Majee, S., Leymann, N., Dunbar, L., and M.
Georgiades, "Use Cases and Requirements for an Interface
to Network Security Functions", Work in Progress,
draft-pastor-i2nsf-merged-use-cases-00, June 2015.
[NFVUC] ETSI, "Network Functions Virtualization (NFV); Use Cases",
ETSI GR NFV 001 V1.2.1, May 2017.
[OAM-USECASE]
Pastor, A. and D. Lopez, "Access Use Cases for an Open OAM
Interface to Virtualized Security Services", Work in
Progress, draft-pastor-i2nsf-access-usecases-00, October
2014.
[PCI-DSS] PCI Security Standards Council, "Payment Card Industry
(PCI) Data Security Standard -- Requirements and Security
Assessment Procedures", PCS DSS v3.2, April 2016,
<https://www.pcisecuritystandards.org/pci_security/>.
[RFC2904] Vollbrecht, J., Calhoun, P., Farrell, S., Gommans, L.,
Gross, G., de Bruijn, B., de Laat, C., Holdrege, M., and
D. Spence, "AAA Authorization Framework", RFC 2904,
DOI 10.17487/RFC2904, August 2000,
<http://www.rfc-editor.org/info/rfc2904>.
Hares, et al. Informational [Page 26]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
[RFC4948] Andersson, L., Davies, E., and L. Zhang, "Report from the
IAB workshop on Unwanted Traffic March 9-10, 2006",
RFC 4948, DOI 10.17487/RFC4948, August 2007,
<http://www.rfc-editor.org/info/rfc4948>.
[RFC5925] Touch, J., Mankin, A., and R. Bonica, "The TCP
Authentication Option", RFC 5925, DOI 10.17487/RFC5925,
June 2010, <http://www.rfc-editor.org/info/rfc5925>.
[RFC6973] Cooper, A., Tschofenig, H., Aboba, B., Peterson, J.,
Morris, J., Hansen, M., and R. Smith, "Privacy
Considerations for Internet Protocols", RFC 6973,
DOI 10.17487/RFC6973, July 2013,
<http://www.rfc-editor.org/info/rfc6973>.
[RFC7149] Boucadair, M. and C. Jacquenet, "Software-Defined
Networking: A Perspective from within a Service Provider
Environment", RFC 7149, DOI 10.17487/RFC7149, March 2014,
<http://www.rfc-editor.org/info/rfc7149>.
[RFC7426] Haleplidis, E., Ed., Pentikousis, K., Ed., Denazis, S.,
Hadi Salim, J., Meyer, D., and O. Koufopavlou, "Software-
Defined Networking (SDN): Layers and Architecture
Terminology", RFC 7426, DOI 10.17487/RFC7426, January
2015, <http://www.rfc-editor.org/info/rfc7426>.
[SDN-SECURITY]
Jeong, J., Kim, H., Park, J., Ahn, T., and S. Lee,
"Software-Defined Networking Based Security Services using
Interface to Network Security Functions", Work in
Progress, draft-jeong-i2nsf-sdn-security-services-05, July
2016.
Acknowledgments
This document was supported by the Institute for Information &
Communications Technology Promotion (IITP), which is funded by the
Ministry of Science, ICT & Future Planning (MSIP) (R0166-15-1041,
Standard Development of Network Security based SDN).
Hares, et al. Informational [Page 27]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Contributors
I2NSF is a group effort. The following people actively contributed
to the initial use case text: Xiaojun Zhuang (China Mobile), Sumandra
Majee (F5), Ed Lopez (Curveball Networks), and Robert Moskowitz
(Huawei).
I2NSF has had a number of contributing authors. The following are
considered co-authors:
o Linda Dunbar (Huawei)
o Antonio Pastur (Telefonica I+D)
o Mohamed Boucadair (France Telecom)
o Michael Georgiades (Prime Tel)
o Minpeng Qi (China Mobile)
o Shaibal Chakrabarty (US Ignite)
o Nic Leymann (Deutsche Telekom)
o Anil Lohiya (Juniper)
o David Qi (Bloomberg)
o Hyoungshick Kim (Sungkyunkwan University)
o Jung-Soo Park (ETRI)
o Tae-Jin Ahn (Korea Telecom)
o Se-Hui Lee (Korea Telecom)
Authors' Addresses
Susan Hares
Huawei
7453 Hickory Hill
Saline, MI 48176
United States of America
Phone: +1-734-604-0332
Email: shares@ndzh.com
Hares, et al. Informational [Page 28]
RFC 8192 I2NSF Problem Statement & Use Cases July 2017
Diego R. Lopez
Telefonica I+D
Don Ramon de la Cruz, 82
Madrid 28006
Spain
Email: diego.r.lopez@telefonica.com
Myo Zarny
vArmour
800 El Camino Real, Suite 3000
Mountain View, CA 94040
United States of America
Email: myo@varmour.com
Christian Jacquenet
France Telecom
Rennes, 35000
France
Email: Christian.jacquenet@orange.com
Rakesh Kumar
Juniper Networks
1133 Innovation Way
Sunnyvale, CA 94089
United States of America
Email: rakeshkumarcloud@gmail.com
Jaehoon Paul Jeong
Department of Software
Sungkyunkwan University
2066 Seobu-Ro, Jangan-Gu
Suwon, Gyeonggi-Do 16419
Republic of Korea
Phone: +82 31 299 4957
Fax: +82 31 290 7996
Email: pauljeong@skku.edu
URI: http://iotlab.skku.edu/people-jaehoon-jeong.php
Hares, et al. Informational [Page 29]