Can a network really defend itself?
The short answer is, "Yes, it can." Network security has evolved from independently deployed products such as firewalls into the realm of system-wide solutions. And Cisco Systems® is at the forefront of the technology development that is making self-defending networks a reality.
The reason is simple: For today's companies, especially in this era of regulatory activity, preserving the integrity, confidentiality and longevity of corporate information is critical to success. As we move further into an information-driven global economy, the value of information, and controlled access to that information, has never been greater. The goal of IT infrastructure therefore is to create systems that can detect and protect against unauthorized access while providing timely access to legitimate users. Simply denying access in the face of an attack is no longer acceptable. Today's networks must be able to respond to attacks in ways that maintain network availability and reliability and allow a business to continue to function. In many respects, the goal of security is to make networks more resilient by making them more flexible. Rather than succumb, networks must be able to absorb attacks and remain operational, much in the same way the human immune system allows us to keep functioning in the presence of viruses and related bacterial infections.
The Necessity of the Self-Defending Network
Corporate networks, and the attacks used to exploit them, are now so complex that no single mechanism can be relied upon to keep them secure. This has led to the concept of "Defense in Depth." Until recently, this concept had been built on the notion of proactive defenses. But given the type of vulnerabilities and attacks that have accompanied our ever-changing networks, Cisco Systems believes in building better adaptive solutions. As a result, Cisco has begun looking to other, seemingly unrelated real-world examples such as our immune system as a model for the Self-Defending Network. Other real-world systems that have also proven to be instructive can be found in the field of epidemiology and in the way communities police themselves. A common theme with all of these systems is that they employ adaptive as well as proactive defenses.
Taking this observation a step further, the defenses to guard a system of this nature tend to be to be built in to every functional block. The key abilities of these adaptive defenses are that they:
- Remain active at all times
- Perform unobtrusively
- Minimize propagation of attacks
- Quickly respond to as-yet unknown attacks
These systems are built on the premise that resources are finite and must be marshaled carefully to avoid resource exhaustion. They also are designed to take full advantage of existing infrastructure with a minimum of disruption to a customer's IT operations.
The Self-Defending Network from Cisco Systems provides systems-based solutions that allow customers to use their infrastructure in new ways to reduce windows of vulnerability, minimize the impact of attacks, and improve overall infrastructure availability and reliability. It is also helps create autonomous systems that can quickly react to an outbreak with little to no human intervention. This type of rapid response is required to thwart the latest forms of misuse that are much more virulent than their predecessors.
The Cisco Self-Defending Network continues to improve its ability to respond to new threats. The first phase, Integrated Security, incorporates security elements in network elements such as switches and routers. The second phase, Collaborative Security, involves building linkages between network security elements and extending the network presence out onto endpoints that connect into a network. The latest phase of the Cisco Self-Defending Network introduces Adaptive Threat Defense (ATD) capabilities, which enhance the ability of a network to respond to threats based on a new set of Anti-X technologies.
The Building Blocks of the Cisco Self-Defending Network
Most customers will not adopt all of the components of the Cisco Self-Defending Network at one time, as it may be difficult to overhaul all of the required subsystems at once without disrupting the integrity of the IT services. Some customers may hesitate to turn over security controls to an automated system until they are confident the system will operate dependably. The Cisco Self-Defending Network initiative deals with these concerns by first providing products that can be usefully deployed independently of one another and then by offering solutions that can link these products together as confidence builds in each product and subsystem-a successful approach based on a combination of product development, product acquisitions, systems development, and partnering. With this in mind, it is worth reviewing the key milestones of the Cisco Self-Defending Network initiative to date.
One of the realities of viruses and worms is that they frequently create network congestion as a byproduct of rapid propagation as well as infection of endpoints. Cisco realized it could begin addressing both problems by offering its customers an endpoint intrusion prevention called the Cisco Security Agent. Cisco Security Agent uses novel forms of behavioral security to detect and prevent viruses and worms from gaining a foothold on an endpoint system, and prevents these viruses and worms from propagating across a network. In effect, Cisco Security Agent becomes a first order dampener to the virus and worm propagation effect.
A second and equally compelling reason for deploying Cisco Security Agent is that it establishes a presence on endpoints that can be used to establish a feedback loop between the endpoint and the network resulting in a network that rapidly adapts to emerging threats.
One of the most high-profile Cisco Self-Defending Network initiatives to date is the Cisco Network Admission Control (NAC) program. NAC allows customers to determine what level of network access to grant to an endpoint based on its security posture, which is based on the security state of the operating system and associated applications. In addition to controlling access, NAC gives IT administrators a way to automatically quarantine and remediate noncompliant endpoints. Making sure endpoints are compliance with OS patches and antivirus software updates is an effective second order dampener to the virus and worm propagation effect. Another way of looking at NAC is as an on-demand vulnerability assessment and patch management tool.
A distinguishing feature of NAC is that it provides both client and backend AAA interfaces that allow customers to plug in products from their preferred endpoint security and policy vendors to drive the NAC plane. Today, more than 30 industry-leading vendors are actively integrating NAC into their technologies.
It is important to extend NAC capability to small and medium-sized businesses (SMBs). To that end, Cisco recently acquired Perfigo, Inc., a developer of packaged network access control solutions that provide endpoint policy analysis, compliance, and access enforcement capabilities and now offers the bundled compliance solution under the name Cisco Clean Access to this market segment.
Strong network admission policies are not a cure all, and do not eliminate the need to continue monitoring devices once they enter a network. Determined attackers can evade just about any admission check, and the network cannot always rely on, or trust, an infected element to turn itself in. Compliant devices also can become infected through a variety of vectors once they are members of a network, e.g., a USB key with infected content. To further help protect the network, Cisco Self-Defending Network is designed to extend the security checks performed at the time of admission for the duration of the network connection. In addition, the Self-Defending Network can rely on other network elements, including other endpoints, to detect when another endpoint(s) is no longer trustworthy, much as a police force monitors crime in a community via a 911 call center. Cisco regards Infection Containment as a third-order dampener to the virus and worm propagation effect.
Existing authentication protocols are not designed to work beyond the initial exchange, however. So the Self-Defending Network must provide new ways of communicating the state of a device (context) and new ways of gauging the veracity of that information based on inferred as well as direct forms of trust. For example, an administrator can create a rule stating that notification received from an endpoint running Cisco Security Agent is more trustworthy than a notification from an unprotected endpoint. As a result Cisco has begun development of new types of correlation and feedback based on inferred attributes. Intelligent Correlation and Incident Response-In order for steady-state feedback mechanisms such as Infection Containment to work effectively, the Self-Defending Network needs to provide services such as real-time correlation of events, quick assessment of the security impact of an event, the ability to decide what action to take, the ability to identify the closest control point to implement a response, and more. To achieve this, Cisco recently announced the acquisition of Protego Networks, whose MARS family of products provides methods for overlaying feedback from a variety of points of presence (POPs) in the network (firewalls, network intrusion detection systems [NIDS], routers, switches, and hosts) with context it obtains from learning the L2 and L3 network topology. This ability helps the Security Incident Response team to rapidly identify where attacks are occurring in the network.
Cisco is also working with netForensics and other partners to enhance its correlation capabilities to better audit the Self-Defending Network.
An important area of ongoing security development at Cisco has been in the area of network intrusion detection systems (NIDS). One of the first Cisco innovations in this area was to integrate NIDS into its router and switching platforms. But in order for NIDS to fully deliver on its capabilities it needs to transform into an intrusion prevention system (IPS) with inline filtering capabilities. This provides a mechanism to remove unwanted traffic with fine-grained programmable classification engines.
Unfortunately most NIDS generate too many false positives to reliably operate as inline security services. Part of the problem is that a large amount of information (context) must be assembled and processed within a fairly short time window-particularly with the amount of context that must be amassed for application-based protocols. This is especially true for applications such as IP Telephony, which is very sensitive to packet transmission delays. To address this, Cisco is developing several methods that provide high-fidelity signaling to these inline classification engines.
Many legitimate activities can be mistakenly regarded by the network as anomalous, particularly in networks where the number of variables is significant. As a result Cisco intentionally adopted a conservative, incremental approach to anomaly detection, beginning with Cisco Security Agent, because it was acknowledged that operating systems are easier to model than network environments. Cisco then acquired Riverhead, which is effectively an inline prevention system with low false positive rates because denial of service (DoS) activity clearly stands out from other network activity. DoS attacks can be identified with greater accuracy by deploying the technology Cisco gained through its acquisition of Riverhead network. This again results in low false positives when identifying a DoS attack.
Based on lessons learned from Cisco Security Agent and the DoS capabilities acquired with the Riverhead family of DoS Guard products, Cisco is introducing an IPS that reduces the false positive rate through the application of innovative anomaly detection techniques and the sharing of state (context) between endpoints and network elements (linkages). It also provides multivector threat identification and meta-event correlation to assess both the vulnerabilities and exploits quickly. By incrementally introducing these technologies into the market, Cisco establishes greater customer confidence in its capabilities and by extension the Self-Defending Network.
Over the past several years, a number of new application-layer network products have emerged to help address new classes of threats that were not adequately addressed by classic firewall and NIDS products, including viruses and worms, e-mail based SPAM and phishing, spy-ware, Web services abuse, IP Telephony abuse, and unauthorized peer-to-peer activity. Cisco has developed the next generation of packet- and content-inspection security services to deal with these types of threats and misuse. This convergence brings granular traffic inspection services to critical network security enforcement points, thereby containing malicious traffic before it can be propagated across the network.
The consolidation of these services onto multiservice platforms represents an opportunity for vendor innovation as well as a reduction in customer cost of ownership. The Self-Defending Network will also become more application aware once these services are integrated into its framework.
Note that when applications employ end-to-end encryption, the Self-Defending Network can collect information from endpoints to compensate for the loss of visibility at the network edge.
Next Steps
Cisco will continue to invest heavily in building Self-Defending Networks that create linkages between POPs across networks up to and including endpoint systems. By doing so, Cisco helps organizations to obtain greater visibility and control of devices, users, and applications that communicate across their infrastructures. This is a necessary and important evolution of what is classically expected of the network, much like the introduction of intelligent routing protocols were to packet forwarding through a communication fabric.
Bob Gleichauf
Vice President/Chief Technology Officer
Security Technology Group
Cisco Systems, Inc.