Fast and Seamless Handover for Secure Mobile ... - IEEE Xplore

6 downloads 177029 Views 857KB Size Report
The Atheros ath9k family ... Atheros chipsets AR5212 for the MN and AR9280 for the APs have been .... from website; http://wireless.kernel.org/mac80211book/.
The 5th IEEE International Workshop on Performance and Management of Wireless and Mobile Networks (P2MNET 2009) Zürich, Switzerland; 20-23 October 2009

Fast and Seamless Handover for Secure Mobile Industrial Applications with 802.11r Ahmad Ali Tabassam, Henning Trsek, Stefan Heiss, Jürgen Jasperneite inIT – Institut Industrial IT, Ostwestfalen-Lippe University of Applied Sciences 32657-Lemgo, Germany {ahmad-ali.tabassam, henning.trsek, stefan.heiss, juergen.jasperneite}@hs-owl.de Abstract – Concerning its temporal behavior the latency caused by a handover from one access point (AP) to another is of particular interest for mobile industrial applications, e.g. automated guided vehicles, due to their real-time requirements in the range of a few milliseconds. With the advancements of wireless local area networks, fast roaming support has become one of the most important issues in IEEE 802.11. The IEEE 802.11r standard amendment has been specified to address roaming capabilities of real-time applications with an attempt to minimize BSS transition time while still providing the 802.11i security and 802.11e QoS. 802.11r permits seamless connectivity with a fast and secure handover from one AP to another within the same mobility domain. It provides an opportunity to derive encryption keys prior to a reassociation to reduce the connectivity loss during the handover. It also provides features for resource reservation and central mobility management. This paper presents an implementation of 802.11r based on a Linux operating system. It uses a softMAC approach that allows moving the processing of MAC layer management entity (MLME) from kernelspace to userspace. Measurement results show the performance gain of 802.11r as compared to a legacy implementation in terms of a handover time reduction, which is a core requirement of wireless industrial networks. Index Terms – Wireless Industrial Networks, 802.11r, Basic Service Set Transition, Fast Handover, Extensible Authentication Protocol.

I. INTRODUCTION Standard wireless technologies are envisioned for deployment in future industrial real-time networks [1]. However, available wireless technologies are mainly designed for office environments and unable to meet the hard industrial requirements in terms of real-time performance, network management and IT security. Whenever wireless links are included, reliability and timing requirements are significantly more difficult to meet, as compared to wired networks, like Real-time Ethernets (RTE), [2]. Wireless technologies have many advantages compared to their wired counterparts. For example they are essential for realizing industrial applications with moving objects, such as automated guided vehicles (AGV’s). However, the IEEE 802.11 standard [3] was developed with the major goal to provide high throughput and almost continuous network connectivity within office and residential environments, which usually rises several challenges when used for industrial automation.

978-1-4244-4487-8/09/$25.00 ©2009 IEEE

Most mobile nodes (MN) use wireless connections to ensure easy mobility, which is provided by a cellular structure of the wireless network to cover huge spatial areas. This structure is made up of several different access points (APs). If the link quality of the current AP is getting weak, a handover to a new AP will be performed. Industrial applications usually require continuous support for various features, such as quality of service (QoS), data encryption, authentication services, etc. The terminologies for different handover types are defined in [4] as follows. •

Fast Handover: A handover that aims at minimizing handover latency, without any explicit interest in packet loss.



Smooth Handover: A handover that aims primarily at minimizing packet loss, with no explicit concern for additional delays in packet forwarding.



Seamless Handover: A handover without any considerable degradation of the quality of service required by the application.

Hence, a methodology for reducing the delay associated with connection re-establishment, to provide a seamless mobility for secured mobile industrial applications, is needed. The remainder of the paper is organized as follows: In section II some background information is provided. This is followed by an explanation of 802.11i in section III and the fast handover mechanism specified in 802.11r in section IV. Section V introduces a real implementation used for the actual measurement setup which is described in chapter VI. The results are presented in section VII followed by the conclusion and an outlook about future activities in section VIII. II. BACKGROUND AND RELATED WORK The IEEE 802.11 standard [3] has two operational modes. In ad-hoc mode, stations communicate directly and must be within each others communication range. In infrastructure mode, access points (APs) are involved in all communications, from and to mobile nodes in the same basic service set (BSS). Therefore, all mobile stations must be within the range of the access point, but no restriction is placed on the distance between mobile stations themselves. In an infrastructure network, mobile stations must associate with an access point to get network services. During association, the mobile station joins an 802.11 network. A mobile station can be associated

750

with only one access point at a time [3]. This process is always initiated by the mobile station, and access points may grant or deny access based on the contents of an association request. Since it is rather easy to attack a WLAN, the use of data encryption and authentication is mandatory for a secure

communication. The wired equivalent privacy (WEP) algorithm initially provided the necessary security services in 802.11 WLANs. However, many potential and critical flaws were rapidly discovered, followed by a complete break in 2001 [5].

Fig. 1. 802.11i RSNA establishment

The IEEE 802.11i amendment adds stronger encryption, authentication and key management methods to guarantee data and system security [3]. 802.11i enables a robust security network authentication using the extensible authentication protocol (EAP) framework, which is independent of authentication methods (cf. Fig. 2). The authentication methods can be based either on passwords or on public key infrastructures [6]. Basic primitives in 802.11i include •

TKIP/Michael for data encryption and message integrity check (MIC) based on the RC4 algorithm.



AES-CCMP (CTR with CBC-MAC Protocol) for data encryption and MIC computation based on the advanced encryption standard (AES).



802.1X Authentication (EAP) to authenticate stations and to distribute the necessary keys.

EAP supports multiple authentication methods, e.g. one-time passwords, certificates, public key authentication, and smart cards.

Fig. 2. IEEE 802.11 stack and EAP associated layers

The EAP framework [7] allows a peer (mobile station/STA) and an authentication server (AS) to negotiate the authentication method and to perform the authentication exchange through an authenticator (network access server). Fig. 3 shows the extensible authentication layered model.

751

Fig. 3. EAP layered model

According to our best knowledge, there is currently no 802.11r capable hardware offered in the open market. The work presented in the current research literature is mostly based on simulation case studies. In [8] a simulation model is presented to study the performance of the 802.11r protocol with the support of IEEE 802.11e QoS mechanisms. A set of simulation experiments have been conducted in this model, but it does not specify, which simulation environment is used. The details and the relative security properties of three different handover protocols, which are CAPWAP, HOKEY and 802.11r are presented in [9]. The performance is measured as the number of round-trips required for the execution of particular connection establishment and handover, without any information about the implementation. The implementation presented in [10] is based on an old chipset (Intersil Prism 2/2.5/3 chipset), which might be not compatible with the new wireless stack (mac80211) in the latest Linux kernel. The several recently developed authentication protocol that can reduce the re-authentication delay in intra-domain and interdomain handover scenarios are discussed in [11].

authentication server sends a master session key (MSK) to the AP. During the key generation and distribution phase TKey, the STA and AP perform a series of operations that establish the cryptographic keys in both entities. This phase includes two types of handshakes: a 4-way handshake and a group handshake. The 4-way handshake must be executed for a successful RSNA establishment, regardless of whether the pairwise master key (PMK) is derived from a MSK or from a PSK. The PMK is used to derive the pairwise transient key (PTK). The traffic between the supplicant (STA) and authenticator (AP) is protected using the confidentiality and integrity algorithm chosen during the discovery phase. Data frames exchanged in phase four are protected using the pairwise transient key (PTK) or group temporal key (GTK) and the negotiated cipher suite. GTK is a random value assigned by the AP. The PTK is used for unicast data transfer, whereas the GTK is used for broadcast and multicast data traffic. An existing association is terminated by tearing down the connection, i.e. the AP de-authenticates the STA. Security associations and temporal keys are deleted and the IEEE 802.1X ports are blocked (as shown in phase five of Fig. 1). B. Handover Procedure If a certain MN moves away from its current AP, resulting in a degraded signal-to-noise ratio (SNR), it is necessary to perform a handover. The moving MN needs to find potential new APs within its range.

III. 802.11I SECURITY A. RSNA Establishment During the robust security network association (RSNA) establishment, the STA discovers an AP to communicate with, either by sending Probe Request frames (active scan) to potential APs or by passively monitoring channels (passive scan) for Beacon frames. The complete procedure is comprised of five phases [6], as shown in Fig. 1. In the discovery phase (TProbe+TOpen) the STA finds an AP, which advertises its security policy in the RSN Information Element (RSNIE) field of the Probe Response frame or in Beacon frames [3]. After the open authentication, the STA associates with the AP, specifying one set of matching security capabilities (authentication, key management suite and cipher suite) from those advertised by the AP. The authentication phase T802.1X is only present, if EAP authentication is enabled [12]. In case of using pre-shared keys (PSK), it is not needed. The authentication frames pass through the AP using EAP. The authentication flow between the STA and the AS typically use the EAP over LAN (EAPOL) protocol [7]. The authentication server can be RADIUS or Diameter. After successfully completing the authentication phase the

Fig. 4. 802.11i handover procedure with PMK caching

Two types of delays might occur for scanning. The probe delay Tprobe for active scanning and the beacon delay Tbeacon for passive scanning. Since the active scanning is faster, only the probe delay will be considered in the rest of the paper. The probe delay Tprobe belongs to the first phase of the RSNA establishment procedure. After selecting the target AP, the

752

delay incurred by exchanging frames for authentication, reassociation and key derivation, can be considered as authentication delay. Thus, the overall handover latency is caused by two distinct logical steps, i.e. discovery and authentication. In fact, fast handover can be realized by reducing the probe delay TProbe and the authentication delay (T802.1X and TKey). The BSS transition time, referred to as Ttotal, is defined as the time between the last acknowledged data frame in the originating BSS, and the first successful transmission of a data frame in the new BSS as shown in Fig. 4. Tprobe is a probe request/response frame exchange with potential new APs, which starts after the last successful data frame exchange with the old AP. Topen is the open system authentication including the (Re)association time. The time T802.1X comprises the pairwise master key (PMK) derivation using EAP. It can be avoided, if the APs and STAs are using the PMK caching mechanism. Tkey is the time needed for the 4-way handshake to establish the PTK and the GTK. The four-way handshake time is required in 802.11i for both cases; connection establishment and handovers. Inspite of other features offered by 802.11r that will be discussed in the coming section; 802.11r incorporates four-way-handshake prior to reassociation process, to shorten this time during handovers. IV. FAST BSS TRANSITION WITH 802.11R A. RSNA Establishment The 802.11r standard amendment [13] specifies mechanisms for a fast BSS transition (FT) and was finally ratified in 2008. This section describes the RSNA establishment, followed by the handover procedure of 802.11r.

element (MDIE) or the fast BSS transition information element (FTIE). The MDIE contains the mobility domain identifier (MDID). All BSSs within an extended service set (ESS) that support fast BSS transition share a common MDID. The FTIE includes information needed to perform the FT authentication sequence during a fast BSS transition in a RSN [13]. 802.11r defines a key management framework, which is based on a three-tier key hierarchy. It consists of different key holders, to be more precise the R0 key holder (R0KH) and R1 key holders (R1KH), as shown in Fig. 5. The R0KH is colocated with the network access server (NAS) client functionality of the IEEE 802.1X authenticator. The R0KH identifier (R0KH-ID) is set to the identity of the co-resident NAS client, while the R1KH identifier (R1KH-ID) is set to the MAC address of the physical entity that stores the corresponding PMK-R1. The IEEE 802.11 station management entity (SME) provides key management [3]. In 802.11r key hierarchy of an authenticator (AP), R0KH and R1KH are part of AP SME RSNA key management, the computation of PMK-R0 and PMK-R1, and all the intermediate results in the computations, are restricted to the R0KH [13]. The computation of PTK, and all intermediate results in its computation, are restricted to the R1KH as shown in Fig. 5. In non-AP STA key hierarchy, S0KH and S1KH are part of the non-AP STA SME RSNA key management. The computation of PMK-R0 and PMK-R1 and all the intermediate results in the computations are restricted to the S0KH [13]. The computation of PTK, and all intermediate results in its computation, are restricted to the S1KH as shown in Fig. 5. In 802.11i, the PTK is derived from the pairwise master key (PMK), as shown in Fig. 6 on the left side. While in 802.11r the PTK is derived from PMK-R1, the second level key of the key hierarchy, as shown in Fig. 6 on the right hand side.

Fig. 6. Key derivations of 802.11i and 802.11r Fig. 5. IEEE 802.11r key hierarchy

The initial 802.11r RSNA establishment is very similar to the establishment of 802.11i as shown in Fig. 1. However, in 802.11r RSNA establishment some frames also contain new information elements, e.g. the mobility domain information

The R0KH is responsible to distribute the authentication keys (PMK-R1A, PMK-R1B, …) to the R1KHs. Hence, when a station roams across APs, the PMK-R1s is assumed present; in case of a handover to a new AP in the same mobility domain, the MN does not need to do the authentication again.

753

B. Handover Procedure The 802.11r handover aims at permitting a continuous connection when wireless devices are in motion, with a fast and secure handover from one AP to another managed in a seamless manner. 802.11r defines four roaming procedures, over-the-air, over-the-distribution system (DS), and both with and without resource reservation [13]. The required resources can be reserved prior to reassociation, if a network is 802.11k capable [14]. To shorten the handover gap, a fast reassociation is introduced in 802.11r that incorporates the four-way-handshake into 802.11 authentication (also known as FT authentication) exchanges prior to the reassociation [13]. In 802.11r, the mobile node and the access point use the FT authentication to specify the PMK-R1 security association and derive a new PTK based on the values of authenticator nonce (ANon) and supplicant nonce (SNon). This exchange enables fresh encryption keys to be computed in advance of a reassociation as shown in Fig. 7. As already mentioned 802.11 does not allow multiple associations at a time, i.e. a STA has to disassociate from the old AP to start the (re)association with the new one. After the successful (re)association 802.11i derives new encryption keys using the 4-way handshake, as shown in Fig. 4. So, there is no connectivity for data exchange during the exchange of encryption key derivation. 802.11r embedment overcomes this loss of data connectivity, by just deriving a fresh encryption key prior to reassociation without 4-way handshake.

implementation available from [15] has been realized. The new information elements (IE) added by IEEE 802.11r to the management frames are managed and processed by the MAC Sublayer Management Entity (MLME). This leads to the basic requirement of a fully controllable MLME as part of an extensible platform for research and development [16].

Fig. 8. IEEE 802.11 protocol model and MLME [3]

There are two types of wireless cards in the open market, either with a fullMAC chipset or with a softMAC chipset. In fullMAC chipsets, the MLME is managed in hardware, e.g. Marvell Libertas 8388 [17]. The softMAC chipsets manage the MLME in software [16], as shown in Fig. 8. The Atheros ath9k family is an example for this type of chipsets. Initially the MLME was implemented in the kernel, whereas the softMAC allows moving it into userspace. The complexity of developing and maintaining new drivers for the Linux kernel is reduced by splitting the implementation of certain features between kernelspace and userspace. As a result, only the most essential code is kept in kernel, while any other functionality, such as management and control code etc, are developed as userspace applications. The new wireless protocol stack mac80211 allows the MLME to be implemented in userspace. Having the code in userspace makes the system more replaceable and flexible, and allows for certain experiments to be carried out [16]. However, the communication performance suffers a little compared to kernel space protocols [18]. For a fast roaming, as specified in 802.11r, implementing the MLME entirely in userspace allows a faster deployment. It is easier to replace a single application, as compared to a replacement of the entire OS kernel. Finally, keeping things in userspace makes it possible to have a testbed for new research experiments. The overall layout of the communication path being used in this implementation is shown in Fig. 9. It is based on three communication components from the recent Linux wireless development [19].

Fig. 7. 802.11r handover procedure

V. FAST HANDOVER IMPLEMENTATION The mechanisms of 802.11r minimize the amount of time that data connectivity is lost between the STA and the distribution system (DS), when a STA moves from one BSS to another within the same ESS. Consequently, a modified 802.11r

• • •

754

mac80211: Wireless driver API for softMAC devices. cfg80211: Driver configuration API. nl80211: Userspace ÅÆ Kernelspace wireless driver. communication transport.

cfg80211 is the new Linux wireless configuration API, while the nl80211 is used to configure a device. The nl80211 is a new 802.11 Netlink interface public header, which is used to transfer information between kernel modules and userspace processes, provides an intra-kernel messaging system as well as a kernelspace/userspace bidirectional communication link, which is an inter-process communication scheme [20]. It consists of a standard sockets-based interface for userprocesses and an internal kernel API for kernel modules (mac80211). The mac80211 is the Linux API used to write softMAC wireless drivers [19]. mac80211 is a wireless subsystem or stack, which resides in kernelspace and contains different driver modules that are used depending on the wireless interface requirement. The nl80211 features are enabled in cfg80211 to provide a communication transport between both domains.

the MLME. The MLME implementation of the MN and the AP is moved to userspace for a successful deployment of a fast BSS transition compliant to 802.11r. As WLAN hardware the Atheros chipsets AR5212 for the MN and AR9280 for the APs have been used. VI. TESTBED SYSTEM ARCHITECTURE The testbed consists mainly of Azimuth Systems [22] components and was especially designed for reproducible handover measurements by means of a conducted test environment as specified in [23]. As shown in Fig. 10, two APs (AP1/AP2) and an authentication server (RADIUS [24]) are connected to the wired Ethernet backbone (distribution system). One mobile station (STA) is connected to one of the APs by a matrix of controllable attenuators, which emulate the real wireless channel. The APs and the STA reside in anechoic chambers to minimize environmental influences further. A traffic generator generates the necessary data frames and two traffic capture engines monitor and record the traffic on both channels in order to observe the whole handover procedure. Afterwards, the capture files are automatically analyzed and a detailed test report is generated.

Fig. 9. Component interaction of mac80211

The main communication flows consist of TX/RX paths and the management MLME. The configurations, for both operational modes the ad-hoc mode and the infrastructure mode, are initiated from the userspace using nl80211. The packet received by the driver, is passed to mac80211’s rx function (ieee80211_rx) [19] along with a rx_status info and the receive handlers are invoked. The data packets are converted to 802.3 format, and delivered to the networking stack, while the management packets are delivered to the MLME. In the transmit path, the packets are handed to the tx function (ieee80211_subif_start_xmit) [19] of the virtual interface, converted into the 802.11 format, and passed to ieee80211_xmit. After this the transmit handlers are invoked, and the packets are given to driver. Both paths include the en/decryption in software. The management MLME requests from userspace are converted into internal variables. For normal procedures, such as probe request/response, auth request/response and the assoc request/response, the state machine runs on userspace requests and then sends the notification to userspace [21]. The userspace notifications are handler of the cfg80211, which is part of the wireless subsystem (mac80211). Thus, the 802.11r management entities can be easily managed and processed by

Fig. 10. Testbed system architecture

As already described, the implementation of the APs and the mobile station are customized to the specific needs for this evaluation. They are based on a modified Linux wireless testing Kernel-2.6.28 and are using a new wireless stack of the Linux kernel [17]; mac80211. It implements only parts of its functionality in hard- or firmware. The remaining functionality is realized as process in the userspace. The APs are connected to a radius-server via an ordinary Ethernet backbone. The APs are equipped with an Atheros AR9280 chipset and are capable of working in master mode. The mobile node (Laptop) is equipped with a wireless CardBus Adapter, AR5212, capable of working in STA mode. These chipsets also support 802.11i security and 802.11e QoS. The security is achieved by using CCMP encryption based on the advanced encryption standard (AES). QoS is provided by using four traffic categories with different priorities for accessing the wireless medium.

755

VII. EXPERIMENTAL RESULTS First, real industrial WLAN devices have been evaluated with respect to their handover performance as a reference. The setup consisted of two industrial APs and a client using traffic characteristics identical to the other measurements. Moreover, key caching was enabled. The results are shown in Fig. 11, it can be seen that the initial handover took approx. 370ms, due to the full 802.1X authentication. All consecutive handover procedures are reduced to values ranging from 27ms to 30ms.

an anonymous identity is used as a security feature to establish a secure tunnel for phase 2. In phase 2, different EAP methods are used protected by a TLS tunnel that has been established in phase 1. MSCHAPv2 or TLS are commonly used in phase 2 of PEAP. A certificate authority (CA) certificate is required for PEAP. EAP-TLS is the most secure method among them. Besides a CA certificate, it also requires a client certificate and private key. Due to this a longer time is needed compared to other EAP methods. Furthermore, the different mechanisms of 802.11i and 802.11r with respect to handover and connection establishment are compared. Since the previously described probe delay TProbe depends only on the client implementation, we will assume TProbe=0, and exclude this phase from the presented measurement results. Only those phases, which depend on the infrastructure side, are shown in Fig. 13. They consist of the open authentication phase Topen (dark grey), the EAP authentication phase T802.1X (grey) and the key generation and distribution phase TKey (light grey).

Fig. 11. Handover performance of industrial APs and an industrial Client

Second, the performance of our implementation was measured. IEEE 802.11 makes use of IEEE 802.1X authentication that relies on EAP. The EAP methods used for WLAN deployments have their own requirements specified in [25]. As a first step, different EAP authentication methods for WLANs have been taken into consideration. These methods [7] comprised in the measurements are EAP-Transport Layer Security (EAP-TLS), Protected-EAP-MSCHAPv2, ProtectedEAP-TLS and EAP-MD5. Fig. 13. Connection establishment and Handover of 802.11i and 802.11r

Fig. 12. EAP authentication methods

Among these EAP methods, EAP-MD5 is one of the fastest, because it is based on passwords, however it has a poor protection level. PEAP is a certificate-based method, which is called either two-phase or tunneled EAP method. In phase 1,

For a deployment of the 802.11i mechanisms, the connection establishment as well as the handover time is 172ms. It can be seen that the handover time in 802.11i networks is significantly reduced to 48ms when using a caching mechanism. However, this improvement only works, if the moving node roams back to an already visited AP. In case of associating to a completely new AP the handover time is the same as for a full 802.11i connection establishment. The connection establishment for 802.11r requires some more time than the 802.11i establishment, because the establishment of the key hierarchy (PMK-R0/PMK-R1’s) needs additional computation. Furthermore, the key derivation in the userspace probably causes additional delays and offers potential for further optimizations. The result for an 802.11r handover is reduced to 19ms and only slightly increased compared to just having the mandatory open authentication. This is due to the unnecessary 4-way handshake for key generation and distribution. Consequently, 802.11r has a much better performance and a decreased connectivity loss during a handover.

756

VIII.

[6]

CONCLUSION

In this paper, different methods for a secure authentication have been described and their impact on the handover performance has been shown. Therefore, the 802.11r mechanisms were implemented on an evaluation platform, using the softMAC approach, i.e. the MAC layer management entity (MLME) is moved from kernelspace to userspace. This provides an extensible research platform for working with new standards. Developing userspace code is much simpler than kernel code. Having the new code in userspace makes it more replaceable and allows various experiments. On the other hand, the measurement results show the performance gain of 802.11r by reducing the handover duration from an infrastructure point of view. They also show that only a very short connectivity loss remains during the handover, which is a core requirement for mobile wireless industrial applications. For the application, 802.11r provides seamless connectivity with a fast and secure handover from one AP to another within the same mobility domain. This is achieved by deriving all necessary encryption keys prior to a reassociation. However, the initial probe phase still depends on the client implementation, which has to be considered and modified as well to obtain an optimal performance. Further steps in this work include the optimization of the current 802.11r implementation and an extensive simulation study to investigate scenarios with more than one client for generating different load conditions. The protection of management frames, such as the open authentication request/response, will be investigated, as it is also a crucial point for security. ACKNOWLEDGMENT The work presented in this paper was part of the project "RAvE - Realtime Automation Networks in moving Industrial Environments" with the signature 1787A07. It is funded by the German Federal Ministry of Education and Research. REFERENCES [1] A.Willig, K. Matheus, and A. Wolisz, “Wireless Technologies in Industrial Network,” in Proc. of IEEE, vol. 93, no. 6, June 2005, pp. 1130-1151. [2] IEC 61784-2, Industrial communication networks – Profiles – Part 2: Additional fieldbus profiles for real-time networks based on ISO/IEC 8802-3, ed. 1.0, Dec. 2007. [3] 802.11-2007, IEEE Standard for Information Technology Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Spec., June 2007. [4] J. Manner and M. Kojo, Mobility Related Terminology, IETF RFC 3753, June 2004; http://www.ietf.org/rfc/rfc3753.txt. [5]

S. R. Fluhrer, I. Mantin, and A. Shamir, “Weaknesses in the key scheduling algorithm of RC4,” in Selected Areas in Cryptography 2001, vol. 2259, Lecture Notes in Computer Science, S. Vaudenay and A. M. Youssef, Springer, 2001, pp. 1-24.

S. Frankel et al., Establishing Wireless Robust Security Networks: A Guide to IEEE 802.11i, NIST Special Publication 800-97, Feb. 2007.

[7] M. Nakhjiri and M. Nakhjiri, AAA and Network Security for Mobile Access Radius, Diameter, EAP, PKI and IP Mobility, John Wiley & Sons, 2006. [8] H. Ahmed and H. Hassanein, “A performance study of roaming in wireless local area networks based on IEEE 802.11r,” in Proc of IEEE 24th Biennial Symposium on Communications, Kingston, ON, June 2008. [9] T.C. Clancy, “Secure handover in enterprise WLANs: capwap, hokey, and IEEE 802.11r,” IEEE Transactions on Wireless Communication, vol. 15, pp. 80-85, Oct. 2008. [10] S. Bangolae, C. Bell, and E. Qi, “Performance Study of Fast BSS Transition using IEEE 802.11r,” in Proc. of ACM Int’l Conf. on Communication and Mobile Computing (IWCMC ´06), British Columbia, Canada, July 3-6 2006. [11] X. Zheng, “Handover Keying and Its Uses,” IEEE Transactions on Network, vol. 23, pp. 27-34, March 2009. [12] C. Jyh-Cheng et al., “Wireless LAN Security and IEEE 802.11i,” IEEE Transactions on Wireless Communication, vol. 12, pp. 27 - 36, Feb. 2005. [13] 802.11r-2008, IEEE Standard for Information Technology Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Spec., Amendment 2: Fast Basic Service Set (BSS) Transition, July 2008. [14] 802.11k-2008, IEEE Standard for Information Technology Telecommunications and information exchange between systems Local and metropolitan area networks Specific requirements - Part 11: Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) Spec., Amendment 1: Radio Resource Measurement of Wireless LANs, June 2008. [15] Host AP driver for Intersil Prism2/2.5/3, hostapd, and WPA Supplicant, available from website; http://hostap.epitest.fi/ [16] A. A. Tabassam, Implementation of Kernelspace to Userspace MLME: An extensible wireless research platform, tech. report TR-inIT010209, inIT – Institut Industrial IT, Ostwestfalen-Lippe University of Applied Science, Lemgo, Germany, 2009. [17] J. W. Linville, “Tux on the Air: The State of Linux Wireless Networking,” in Proc. of Linux Symposium, vol. 2, Ontario, Canada, July 2008. [18] M. Oey, K. Langendoen, and H.E. Bal, “Comparing kernel-space and user-space communication protocols on Amoeba,” in Proc. of 15th Int’l Conf. on Distributed Computing Systems, 30 May-2 June 1995. [19] J. M. Berg, The mac80211 subsystem for kernel developers, available from website; http://wireless.kernel.org/mac80211book/ [20] J. Salim, et al., Linux Netlink as an IP Services Protocol, IETF RFC 3549, July 2003; http://www.ietf.org/rfc/rfc3549.txt. [21] J. M. Berg, “WiFi Overview,” presented at Linux wireless mini-summit Berlin, Germany, June 2009. [22] Azimuth Systems Inc., http://www.azimuthsystems.com/, 2009. [23] 802.11.2-D1.01, IEEE Draft Recommended Practice for the Evaluation 802.11 Wireless Performance, Feb. 2008. [24] C. Rigney, S. Willens, A. Rubens, and W. Simpson, Remote Authentication Dial In User Service (RADIUS), IETF RFC 2865, June 2000; http://www.ietf.org/rfc/rfc2865.txt. [25] D. Stanley, J. Walker, and B. Aboba, Extensible Authentication Protocol (EAP) Methods Requirements for Wireless LANs, IETF RFC 4017, March 2005; http://www.ietf.org/rfc/rfc4017.txt.

757