September 4 – 9 , 2016, Dagstuhl Seminar 16361
Network Attack Detection and Defense - Security Challenges and Opportunities of Software-Defined Networking
1 / 4 >
For support, please contact
(Use seminar number and access code to log in)
From September 4 through 9, 2016, more than 40 researchers from the domains of computer networks and cyber security met at Schloss Dagstuhl to discuss security challenges and opportunities of software-defined networking (SDN).
Software-defined networking has attracted a great attention both in industry and academia since the beginning of the decade. This attention keeps undiminished. In 2014, IDC predicted that the market for SDN network applications would reach $1.1bn. Especially in industry, the vision of "programming computer networks" has electrified many IT managers and decision makers. There are great expectations regarding the promises of SDN. Leading IT companies, such as Alcatel-Lucent, Cisco systems, Dell, Juniper Networks, IBM, and VMware, have developed their own SDN strategies. Major switch vendors already offer SDN-enabled switches.
Software-defined networking provides a way to virtualize the network infrastructure to make it simpler to configure and manage. It separates the control plane in routers and switches, which decides where packets are sent, from the data plane, which forwards traffic to its destination, with the aim to control network flows from a centralized control application, running on a physical or virtual machine. From this controller, admins can write and rewrite rules for how network traffic, data packets, and frames are handled and routed by the network infrastructure. Routers and switches in a sense become "slaves" of this application-driven central server. SDN-enabled networks are capable of supporting user requirements from various business applications (SLAs, QoS, Policy Management, etc.). This is not limited to the network devices of a certain vendor. It can be applied to devices from various vendors if the same protocol is used. Most SDN infrastructure utilizes the widely-used OpenFlow protocol and architecture to provide communication between controllers and networking equipment.
Security-related aspects of software-defined networking have only been considered more recently. Opinions differ widely. Some believe that the security problems introduced by SDN are manageable - that SDN can even bring security benefits; others think that Pandora's Box has been opened where SDN and SDN-enabled networks can never be secured properly.
No doubt, there are a number of serious security problems as the following examples show. SDN controllers represent single points of failures. The controllers as well as the connections between controllers and network devices might be subject to distributed denial of service attacks. Compromising the central control could give an attacker command of the entire network. The SDN controllers are configured by network operators. Configuration errors can have more complex consequences than in traditional settings because they may unpredictably influence the physical network infrastructure. Furthermore, the idea of introducing ‘network applications’ that interact with the controller to modify network behavior seems like a complexity nightmare in terms of required authentication and authorization schemes. Finally, the SDN paradigm is a major turn around with respect to the basic design rules that have made the Internet successful so far, namely a well-defined layered approach. Whereas in today’s world, applications have no say in routing decisions, SDN’s promise for highly flexible and application-tailored networking requires a way for applications to optimize networking decisions for their own benefits. However, it is unclear to what extent fairness can be ensured, how conflicting decisions can be resolved, etc. Along the same line, members of the security community worry about the possibility to intentionally design SDN applications that could eventually be turned into attack weapons or simply be misused by malicious attackers. Whether these fears are substantiated or not is something which has not received any scrutiny so far.
On the other hand, SDN is also considered by many researchers as an effective means to improve the security of networks. SDN controllers can be used, for instance, to store rules about the permission of certain requests which cannot be decided at the level of a single switch or router because this requires full overview over network status or additional information and interactions which are not contained in the current protocol versions. Attacks that can be detected this way are ARP spoofing, MAC flooding, rogue DHCP server, and spanning tree attacks. Also, by enabling the creation of virtual networks per application, people speculate that intrusion detection techniques relying on the modeling of the normal behavior of network traffic will become much easier to implement and more reliable in terms of false positive and negatives. Similarly, SDN apps could offer a very simple and effective way to implement quarantine zones for infected machines without cutting them off completely from the network since the quarantine could be customized at the application level (letting DNS and HTTP traffic for a given machine go through but not SMTP, for instance).
These two contrary facets of SDN security were the key ingredients for an extremely lively and very fruitful seminar. The seminar brought together junior and senior experts from both industry and academia, covering different areas of computer networking and IT security. The seminar started with two invited talks by Boris Koldehofe (TU Darmstadt, DE) and Paulo Jorge Esteves-Veríssimo (University of Luxembourg, LU) on the basics and security aspects of software-defined networking. After that we organized six working groups to discuss in two rounds the Good and the Bad of using SDN from the security point of view. Based on the outcome of the working groups and a plenary discussion, we formed another four working groups to discuss required research directions. The first six working groups focus on the following issues: (1) centralization in SDN, (2) standardization and transparency, (3) flexibility and adaptability for attackers and defenders, (4) complexity of SDN, (5) attack surface and defense, and (6) novelty and practicability. The research direction working groups dealt with (1) improving SDN network security, (2) a secure architecture for SDN, (3) secure operation in SDN-based environments, and (4) SDN-based security. The discussion in the working groups was supplemented by short talks of participants to express their positions on the topic or to report about ongoing research activities. Based on the talks, discussions, and working groups, the Dagstuhl seminar was closed with a final plenary discussion which summarized again the results from the working groups and led to a compilation of a list of statements regarding the security challenges and opportunities of software-defined networking. The participants agreed that SDN provides new possibilities to better secure networks, but also offers a number of serious security problems which have to be solved for being SDN a successful technology. The outcome of these discussions and the proposed research directions are presented in the following.
Creative Commons BY 3.0 Unported license
Marc C. Dacier and Sven Dietrich and Frank Kargl and Hartmut König and Radoslaw Cwalinski
Dagstuhl Seminar Series
- 14292: "Network Attack Detection and Defense: Securing Industrial Control Systems for Critical Infrastructures" (2014)
- 12502: "Securing Critical Infrastructures from Targeted Attacks" (2012)
- Security / Cryptology
- Software defined networking
- OpenFlow protocol
- Programmable networks
- Attack detection
- Targeted attacks
- Network monitoring
- Intrusion detection
- Vulnerability analysis
- Malware assessment
- Denial-of-service attack detection and response