Sunteți pe pagina 1din 6

Cybersecurity Applications & Technology Conference For Homeland Security

Virtual Private Groups for Protecting Critical Infrastructure Networks∗

Richard C. O’Brien and Charles N. Payne, Jr.


Adventium Labs LLC
Minneapolis, MN 55082
richard.obrien@adventiumlabs.org
charles.payne@adventiumlabs.org

1. Introduction however, the landscape has changed. Proprietary network


protocols are being replaced by TCP/IP. DCS networks that
In an era when critical infrastructure networks are in- used to be isolated are now frequently connected to com-
creasingly less isolated and more accessible from open net- pany business LANs, perhaps via a firewall, perhaps not.
works, including the Internet, the air-gap security that these These business LANs are, in turn, connected to the Inter-
critical networks once enjoyed no longer exists. Malicious net as shown in Figure 1. SCADA networks that used to
individuals can exploit this network connectivity, in con- run on leased lines, now run over the Internet. As a result,
junction with security weaknesses in widely used, homoge- many critical infrastructure networks that used to be pro-
neous, COTS (commercial off-the-shelf) products, to pene- tected from the outside via an air-gap are now susceptible
trate deep within an organization’s critical networks. Such to remote attacks.
an attack on SCADA (Supervisory Control And Data Ac-
quisition) and Process Control networks could have dev-
astating consequences. This paper describes an approach,
Virtual Private Groups (VPGs), for creating and managing
a virtual air-gap between these networks and the environ-
ments in which they may operate.
After a brief description of the security issues that con-
front these networks, we describe our approach for address-
ing them. Many of the ideas presented here are the result
of work done while implementing a version of VPGs di-
rected towards critical infrastructure networks. In the pro-
cess of doing that work we made a number of advances in
managing policy for VPG and related mechanisms. These
advances are identified in Section 5 and Section 6.

2. Critical infrastructure networks Figure 1. A process control network

The critical infrastructure networks that we focus on in While in many ways SCADA and DCS networks are be-
this paper are SCADA and DCS (Distributed Control Sys- coming more like other Ethernet networks, in some impor-
tems) networks used in industrial process control. Histori- tant ways, they remain very different. Because of their spe-
cally, these networks originated as proprietary, isolated lo- cific purpose and critical nature, they usually include a va-
cal area networks (LANs). They ran their own proprietary riety of specialized computers and devices that have been
network protocols and remote sites were connected via pri- developed and certified to perform very specific processing
vate networks. As such they were not vulnerable to at- with a high-degree of reliability. Many of these devices are
tacks from outside their networks and security of the net- legacy systems that cannot be easily modified. In addition,
work was not a major concern. With the advent of the Inter- the vendors of these devices are usually more concerned
net and the trend towards ubiquitous network connectivity, with reliability than security, since security has not been an
∗ This work was funded primarily under DHS contract #FA8750-05-C- issue in the past, and so may be reluctant to add features to
0144 and DARPA contract #FA8750-07-C-0017. their systems that might affect availability or certification.

978-0-7695-3568-5/09 $25.00 © 2009 IEEE 118


DOI 10.1109/CATCH.2009.14
Reliability is also often supported by the use of redundant and as a result uses a model where compromise of a key on
network connections, so that failure of a network compo- one system only affects that system’s security.
nent (e.g. NIC card or switch) does not prevent the system The network model for SCADA and DCS networks we
from continuing to function. are considering, however, has some additional features that
can be leveraged to provide a simpler solution.
3. Protecting Critical infrastructure networks • The network is managed by a single organization and
so does not require the full generality of a PKI infras-
A promising approach to increasing the security of these tructure.
networks is to incorporate security features in unobtrusive
bump-in-the-wire devices such as a network interface card • The network devices used to provide the firewalling
or a network appliance. Such devices can protect the net- and authentication/encryption are special devices that
work, without requiring any changes to the operation or cer- can be hardened so that they are not easily accessible.
tification of the application nodes they protect, by a combi- As a consequence they can be trusted to protect any
nation of firewalling each critical node and authenticating key material they store.
(and possibly encrypting) traffic between critical nodes. All
critical network traffic travels authenticated over the net- To take advantage of these network features, we propose
work and only authenticated traffic is accepted by a criti- using the concept of virtual private groups.
cal node. In DCS networks that are monitoring and man-
aging local process control systems, authentication of the 4. Virtual Private Groups
traffic and its source, with replay protection, may be suffi-
cient since the actual data that is transmitted would be of
The key assumption behind the VPG approach is that
little use to an attacker. In instances, such as large scale
when the endpoints are isolated devices that can be hard-
SCADA systems, where the data is being sent over the In-
ened because we have control over them, then they can be
ternet, encryption might also be used to limit the visibility
trusted to adequately protect the key material with which
of an attacker into the state of the system.
they are entrusted, and as a result key management can be
One possibility for providing the authentication and en-
simplified. For SCADA and DCS networks that are pro-
cryption needed with this approach is to use standard IPsec.
tected by bump-in-the-wire-network devices, this assump-
The maturity of public key infrastructures (PKI) has encour-
tion is appropriate so the VPG trust model fits.
aged the widespread adoption of certificate-based virtual
VPGs are based on the use of a centralized management
private networks (VPNs), even when the target networks do
component, the policy server, that creates and distributes
not require that level of assurance, so this often means set-
shared keys used by groups of nodes to communicate se-
ting up a Public Key Infrastructure to create and distribute
curely. VPG keys are created, distributed and revoked by
private keys and certificates for each device. Peer-to-peer
the policy server using secure management channels be-
security associations are then negotiated between any two
tween the policy server and each trusted endpoint. This en-
nodes that need to communicate with each other. While this
ables simpler key administration for dynamically managing
approach is workable, it has some significant disadvantages.
groups of critical nodes and does not require a supporting
It uses a PKI that, in itself, is an additional management bur-
PKI.
den which may be prohibitive for smaller DCS networks.
The additional overhead of key negotiation between nodes A Virtual Private Group (VPG) distinguishes itself from
may also not be desirable, since a node that serves a large a Virtual Private Network (VPN) in the following ways:
number of clients might spend an unacceptable amount of 1. VPGs use a single encryption key for all members of
time just negotiating keys with them. Moreover, because the same group, rather than a separate key for each pair
there is a unique key for each host, the model does not scale as with VPNs.
well to large networks, and redundant communication paths
only exacerbate this problem. Finally, Intrusion Detection 2. A VPG key can also be shared with a third party, such
Systems (IDSs) are blinded to any encrypted traffic since as a network intrusion detection device, that is not one
they do not have access to the keys used to do the encryp- of the protected endpoints. This is very difficult to
tion. achieve in a VPN when the key is negotiated between
The underlying reason that IPsec does not fit the pro- those endpoints.
posed approach better is that it is based on a more open
network model where communications might be between 3. A VPG can be defined along any practical criteria that
different organizations with different PKI infrastructures. It could define the group, such as a network service in
cannot assume anything about the security of the endpoints which the hosts may engage.

119
4. Since VPGs are hardware based, they are independent VPGs were implemented using the hardware crypto-
of the host and cannot be easily turned off from a com- graphic support on the 3Com EFW NICs. VPGs were im-
promised host, as can software based systems, and any plemented on the NIC and all encryption/decryption was
attempt at bypassing the security by removing the de- done on the NIC via crypto hardware on packets as they are
vice results in the host no longer being able to partici- processed. The decision on which packets to encrypt and
pate in any encrypted communications. which keys to use to encrypt them was determined by the
packet filtering rules. A new key value field was added to
The concept of virtual private groups was first pro- each packet filter rule. If the rule that the packet matched
posed by Carney, Hanzlik and Markham [1] and imple- had an associated key, then the packet was encrypted with
mented in the Autonomic Distributed Firewall (ADF) [6, that key using triple DES. IPsec tunnel mode was used,
5], a research prototype of the 3Com Embedded Firewall which encrypted the original packet and added an additional
(http://www.3com.com). These early investigations defined header to the packet, but the protocol in the new header used
the “group” as a group of hosts where all communications its own protocol number, IP Protocol 174, rather than the
between group members would be fully protected by the standard IPsec ESP protocol number (IP Protocol 50). On
same cryptographic key. The notion that groups could be the receiving side, when a packet with the VPG protocol
defined along a criteria other than host identity, such as the number was received, it was decrypted and passed through
service to which they engaged, was proposed by Ioannidis the packet filter.
et al [4] in their Virtual Private Services (VPS). However, All key management was performed by the EFW Pol-
VPS relied on traditional certificate-based IPsec to encrypt icy Server. The EFW NICs stored all VPG keys so that
that service traffic. VPS were not intended to simplify key they were not accessible even if the host was compromised.
management or to address the sharing concerns targeted by EFW VPGs could be used to encrypt all traffic between a
VPGs. group of nodes or just traffic that matched certain packet
In summary, VPGs use shared keys among group partici- filtering rules.
pants and a centralized management component that results
Several fundamental advances were made during this
in simpler key management and provides flexibility over the
work that serve as a base for broader use of VPGs.
policy specification criteria.
Service-oriented Policy Management: The traditional
approach of specifying distributed firewall policy exactly as
5. Implementation examples it is to be enforced, that is, one firewall at a time, is time
consuming and tedious even for small deployments. Our
In this section we describe two implementations of VPGs work resulted in a new approach, called Conversations, that
that we have done at Adventium. The first was based on a elevates policy writing to simply specifying which hosts
commercial product, the 3Com Embedded Firewall (EFW). should be authorized to engage in specific network services.
It involved making modifications to the proprietary code It then provides a method for translating this higher-level
that was running on the EFW NIC under a licensing agree- specification into firewall-specific policies such that consis-
ment between Adventium Labs, 3Com and Secure Comput- tency between host-based policies is assured. The policy for
ing (the original developer of the EFW), and its focus was a single enforcement point may be composed from several
on developing a commercializable version of VPGs for pro- conversations, and these conversations may be specified by
tecting critical infrastructure networks. different authors.
The second implementation was focused on developing Transparent VPGs: VPGs are associated with conver-
a version of VPGs that was not dependent on a specific pro- sations. There is no overhead for specifying a VPG; it is
prietary hardware device. It was developed on the DARPA’s as simple as noting that a particular conversation should be
SRS Phase 2 program and was used to restrict and identify protected. The same VPG may protect many conversations,
malicious insiders. and the required cryptographic keys will be generated auto-
matically and bound to the resulting policy rules.
5.1. A Proprietary Solution: 3Com EFW Enforcement Mechanism Independence: Architec-
VPG tural changes and abstractions implemented during this
work enable a conversation to be agnostic with respect to
EFW VPG was developed as an enhancement to the its underlying enforcement mechanisms. Hence, by imple-
3Com EFW 2.5 product. It provides endpoint to endpoint menting specific translation components that produce low
encryption and high-level policy specification integrated level policies from the high-level conversation specifica-
with the EFW filtering and centralized management. EFW tions, the specifications can be used to define policy for a
VPG allowed all network traffic between group members variety of underlying enforcement mechanisms. To demon-
(i.e., protected hosts) to be encrypted and authenticated. strate this feature, we implemented a translation component

120
for a Linux host-based firewall as well as an EFW one. by naming the necessary Security Policy Identifiers (SPIs)
The VPG concept has been used to great advantage on and the cryptographic keys bound to those SPIs. Outgo-
a variety of exercises and deployments and has withstood ing packets are checked against the SPD, and if there is a
several Red Team trials. It played a prominent role on the match, the SAD is searched for the SPIs and cryptographic
DARPA OASIS Dem/Val program. Adventium’s work ad- keys that will encrypt the packet. Under manual keying, an
dressed the lessons learned from the OASIS Dem/Val expe- appropriate SAD entry must be found, or IPsec will return
rience [7] to develop a commercially viable and industrially an error. Under negotiated keying, the endpoints use IKE
robust VPG functionality. to quickly define the necessary SAD entries, after first au-
The 3Com EFW, without VPGs, was successfully used thenticating themselves to each other using either the shared
on the LOGIIC program to help control and identify threats secret or the certificate.
to a process control network. During that work it became Thus, IKE is responsible for populating the SAD, as re-
clear that VPGs could provide a significantly enhanced pro- quired by IPsec policy, whenever an appropriate security
tection capability for these networks and that conversations association cannot be found. Because the negotiation is
could be used to easily specify appropriate high-level poli- pairwise, the newly created security association is unique
cies. to that pair. As a result, the keys are known only to those
endpoints, which makes it very difficult to share them to the
5.2. An Open Solution: IPsec based degree required by a VPG, i.e., within a group of n hosts
and with a third party.
By contrast, VPGs are straightforward to implement us-
DARPA’s Self-Regenerative Systems (SRS) program
ing manually keyed IPsec, because while all SPIs in the
gave us the opportunity to build on our earlier work and
SAD must be unique per peer, there is no requirement that
develop a bump-in-the-wire VPG solution, whose policies
the cryptographic keys bound to each SPI must also be
were controlled by conversations but whose defenses were
unique. Thus, as long as we keep SPIs unique per peer,
based entirely on open standards. This solution, called the
we can distribute cryptographic keys in a manner that is in-
DRED for Detection, Response, Embedded Device, tar-
dependent of those endpoint identities.
geted a different type of critical infrastructure, a mission
Each DRED’s SPD and SAD was populated automati-
planning network [2]. While the DRED contained many
cally from the output of the conversations tool. The con-
network defenses, including packet filters, Ethernet frame
versations tool enabled us to specify that the same crypto-
filters and proxies, we focus below on the DRED’s imple-
graphic keys could be used across multiple hosts, and other
mentation of VPGs over IPsec.
tools ensured that the SPIs specified in each DRED’s SAD
A traditional VPN is implemented in IPsec using one of
would be unique. Thus, without altering IPsec, we were
three methods: manual keying, negotiated keying based on
able to populate the DREDs’ SPDs and SADs in a way that
a shared secret, and negotiated keying based on identity-
achieved the desired effect of VPGs.
based certificates. Manual keying is discouraged for any-
However, while VPGs reduced the keys to manage, they
thing beyond simple testing because it is difficult to protect
could not reduce the size of each endpoint’s SAD over what
keys that must be distributed over untrusted networks, and
would have been required using VPNs, because the SAD
keys must be refreshed often during the lifetime of the VPN.
entries still had to be pairwise and identity-based. This
Negotiated keying schemes provide trustworthy key distri-
highlights an important obstacle using IPsec: it was de-
bution through the Internet Key Exchange (IKE) protocol
signed under the assumption that encryption policy would
[3], which negotiates not only the initial keys but all future
always be based on the identity of the hosts, so any other
keys required by the VPN. If both endpoints are controlled
criteria for encryption must first be recast into this identity-
by the same authority, then negotiation based on a shared
based scheme in order to work.
secret is often preferred. For large enterprises, though, ne-
gotiation based on certificates scales more effectively. Of
these three methods, only manual keying supports all of the 6. Management approach
requirements for VPG implementation. To understand why,
it is useful to review how IPsec actually works and the role Traditional policy management tools focus more on the
that IKE plays. rules to be implemented by a particular policy enforcement
An IPsec connection between a pair of endpoints re- device than on the authorizations that will be enabled by
quires two rules, one for each direction, in each endpoint’s those rules. This can obscure these authorizations, espe-
Security Policy Database (SPD) and two security associa- cially for readers not familiar with the enforcement device.
tions, one for each direction, in each endpoint’s Security Separating authorization from enforcement, on the other
Association Database (SAD). The SPD governs when IPsec hand, clarifies the authorizations and enables them to be im-
must be used, while the SAD describes how it will be used plemented more easily by multiple, different devices.

121
Adventium developed a novel policy management ap- enforcing the particular authorizations resulting from a con-
proach called conversations that separates authorization versation. Bindings are relationships between a conversa-
from enforcement. Conversations is a group-based policy tion entity and the enforcement device that protects it. A
management approach where the groups are related service binding typically exists between a host and a device, but it
instances. Conversations enable policy to be specified at can also exist between a user and a host. In that case, transi-
a very high level (authorizations) and have those changes tive closure helps us determine which device is responsible
propagate automatically into enforcement rules, as shown for that user.
in Figure 2. Adventium has applied conversations success- Under traditional policy management, bindings are im-
fully in several problem domains and typically observes a plicit because the policy is already being written for a par-
two orders of magnitude improvement, in terms managed ticular device. Under Conversations, a change in the bind-
entities, over writing the enforcement rules directly. ings can result in completely new policies at the enforce-
To scale policy specification more effectively, conver- ment layer, even if the conversations themselves were not
sations separate the concerns of what must be authorized, changed. Bindings also clarify relationships that are re-
where must each authorization be enforced (i.e., what de- quired by some defenses. For example, IPsec requires both
vice must enforce it), and how must each authorization be the host and device identities in order to specify tunnel
enforced (e.g., what rules are necessary). mode encryption policy (where the device is a gateway).
A set of conversations and bindings together determine
6.1. What is authorized? the permissions that must be enforced by each device. Dur-
ing the policy translation process, every conversation gen-
A service instance is a network service, such as HTTP erates the following permissions, ∀c ∈ C, s ∈ S, p ∈ P
or DNS, from a specific provider(s). A conversation autho- named in that conversation,
rizes service instances for a set of consumers. The services
may have different providers, but all must share the named δc : c, sclient , p, k,
consumers. Formally a conversation is denoted
+
(C, ((S, P ) )) (1) δp : c, sserver , p, k
where C is the set of service consumers, S is a set of ser- where δc and δp name the enforcement device associated
vices, P is a set of providers that all provide S, and + means with c and p, respectively. These expressions mean “δc en-
one or more. Each (S, P ) defines a service instance. C and forces that c is authorized to request s (i.e., be a client of
P may represent users, roles, hosts or the EFW or DRED it- s) from p and this communication is protected according to
self. If C were a role, then the conversation would resemble k” and “δp enforces that p is authorized to accept requests
a role-based access control policy, except that conversations from c on s (i.e., be a server of s) and this communication
define policy simultaneously for both C and P , not just for is protected according to k”, respectively.
C (i.e., a role) alone. In a single conversation, there may be
multiple consumers, multiple services provided by multiple
(and not necessarily identical) providers. 6.3. How is the authorization enforced?
To enable VPGs, we extend Definition 1 to specify that
all communications authorized must be protected by cryp- Once we have identified where an authorization is en-
tography: forced, we can turn our attention to how it is enforced. The
+
(C, ((S, P ) ))k (2) enforcement semantics differ greatly depending on whether
where k is the cryptographic protection strategy. In our the target device is an EFW, a DRED, or something else.
case, k names a VPG key that is shared between C and P For this reason, the service s has, so far, remained abstract.
for all communications authorized by this conversation. The function of service semantics is to bind s to the nec-
The Conversation Manager (CM), a MySQL-based ap- essary enforcement rules that will enable s. Service seman-
plication, enables the construction and maintenance of con- tics are rule templates whose placeholders (entity identities
versations. It also supports the graphical display of conver- and services) are derived from the conversations. Service
sations in multiple ways, e.g., by service, by user, by en- semantics are independent of any high-level policy, so once
forcement device, etc., according to the needs of the viewer. defined they can be reused indefinitely. Service semantics
are created for each service in a conversation and for each
6.2. Where is the authorization enforced? defense (e.g., IPsec) that enforces its use. More specifically,
the service semantics for each defense are applied to each
Along with conversations, the CM is used to specify the permission assigned to the device to generate the necessary
bindings that determine which devices are responsible for rules for that defense (on that device).

122
Figure 2. Conversation specification with translation to device policies.

6.4. Benefits References

Separating concerns in this way offers many benefits. [1] M. Carney, R. Hanzlik, and T. R. Markham. Virtual private
First, conversations clarify sharply what is authorized be- groups. In Proceedings of the 3rd Annual IEEE Information
cause the reader is not confronted with the detailed syntax Assurance Workshop, 2002.
[2] J. T. Haigh, S. A. Harp, R. C. O’Brien, C. N. Payne, J. Gohde,
of the policy enforcement rule. Second, there is relatively
and J. Maraist. Trapping malicious insiders in the spdr web.
low cohesion and coupling between conversations, bindings In Proceedings of the Forty-Second Annual Hawaii Interna-
and service semantics so each can be managed nearly inde- tional Conference on System Sciences (HICSS-42), Waikoloa,
pendently of the others. This offers significant advantages Big Island, Hawaii, January 2009. To appear.
for environments where what is to be authorized may be de- [3] Internet Engineering Task Force, Network Working Group.
fined by a different authority than where it must be autho- Internet key exchange (ikev2) protocol, December 2005.
rized or how it must be authorized. Third, rather than rely- http://tools.ietf.org/html/rfc4306.
ing on a single individual with expertise in the chosen policy [4] S. Ioannidis, S. M. Bellovin, J. Ioannidis, A. D. Keromytis,
enforcement point, conversations may be defined by multi- and J. M. Smith. Design and implemention of virtual private
ple authors, each with a particular expertise in the named services. In Proceedings of the 12th IEEE International Work-
shop on Enabling Technologies: Infrastructure for Collabo-
service and how it should be deployed.
rative Enterprises (WETICE), pages 269–274, Linz, Austria,
June 2003. IEEE.
7. Conclusions [5] T. Markham, L. Meredith, and C. Payne. Distributed embed-
ded firewalls with virtual private groups. In DARPA Infor-
mation Survivability Conference and Exposition, 2003, vol-
Critical infrastructure networks operate under differ- ume 2, pages 81–83, April 2003.
ent trust assumptions than conventional business networks. [6] C. Payne and T. Markham. Architecture and applications for
Nodes in these networks are trusted for specific functions, a distributed embedded firewall. In 17th Annual Computer
and the communication paths between them are relatively Security Applications Conference, December 2001.
limited and static. As a result, using hardware-based so- [7] P. Rubel, M. Ihde, S. Harp, and C. Payne. Generating poli-
lutions that are independent of the protected node and cies for defense in depth. In Proceedings of the 21st Annual
conversation-based VPGs can provide the necessary pro- Computer Security Applications Conference. IEEE, 2005.
tections in a way that acknowledges the more limited op-
erational and trust models of these networks.

123

S-ar putea să vă placă și