ACM / Crossroads / Xrds9-4 / The ABC..I of 802.11 WLANs

The ABC...I of 802.11 WLANs

by Moustafa A. Youssef and Arunchandar Vasan

Introduction

Wireless Local Area Networks (WLANs) are rapidly being deployed to provide last-hop user mobility. They use a collection of electromagnetic signals in space, called a channel, as the communication medium to transmit data.

The design choices of WLAN technology are quite different from those of traditional wired network technology such as Ethernet. WLANs are inherently error-prone, subject to interference from other electromagnetic sources, and are affected by numerous other communication phenomena, which cause performance degradation. They have constraints of power usage imposed on them due to the mobile nature of hosts using the WLAN (e.g., PDAs and laptops). It is easy for an intruder to eavesdrop on wireless traffic compared to tapping a wired network physically. WLANs are suitable candidates for ad-hoc routing, where nodes discover the presence of other nodes and route data among themselves without any explicit configuration.

We present a survey of 802.11 WLANs [10, 13]. In particular, we focus on the physical and MAC (Medium Access Control) layers, which have been defined by the IEEE. Finally, we explain the various standards currently debated by the IEEE such as 802.11a, 802.11b, 802.11c, 802.11d, etc., and outline recent trends in WLAN research.

History of 802.11 Wireless LANs

In 1985, the Federal Communications Commission (FCC) [7] made the commercial development of radio-based LAN components possible by authorizing the public use of the Industrial, Scientific, and Medical (ISM) bands. This band of frequencies resides between 902 MHz and 5.85 GHz, just above the cellular phone operating frequencies. The choice of the free ISM band was a major force behind the widespread of the 802.11 Wireless LANs, as the end users do not have to obtain FCC licenses to operate the products. Without a standard, however, vendors began developing proprietary radios and access points.

In the late 1980s, the Institute of Electrical and Electronics Engineers, Inc. (IEEE) 802 working Group, responsible for the development of LAN standards began development of standards for wireless LANs. The IEEE 802.11 Working Group developed the Wireless LAN Medium Access Control and Physical Layer specifications. The IEEE Standards Board approved the standard on June 26, 1997, and the IEEE published the standard on November 18, 1997.

Architecture of 802.11 WLANs

A typical 802.11 WLAN consists of a number of wireless stations associated with Access Points (AP). Such a setup is called the infrastructure mode of operation (Figure 1). An AP is a special station with two interfaces: one on the WLAN and another wired interface. It takes 802.11 frames from the WLAN and bridges them (i.e., converts the link layer headers) to packets suitable for the wired format (typically Ethernet) and vice versa. This allows the stations to send packets to and receive packets from the wired network, and, from the Internet.

Figure 1: An example of the infrastructure mode.

When a station moves within the coverage area of the WLAN, it could potentially leave the coverage area of an AP. If this happens, the station sends out probe messages. APs that cover the new area respond to this, and the station associates with one of these APs, which it chooses from factors such as signal strength and noise level. This procedure of associating with a new AP is called a handoff. Note that 802.11 provides only for link layer mobility, i.e., the IP address cannot change. Any disruption in communication due to the handoff is handled by recovery at higher network layers (such as a reliable transport like TCP).

802.11 also provides for an ad-hoc mode (Figure 2), where there is no need for any infrastructure like the AP. In this case, stations talk to each other directly. As there is no central station, like the AP, to relay data frames, the coverage area of ad-hoc mode of operation is typically limited. However, it is possible to implement ad-hoc routing protocols [13] (e.g., TORA and DSDV) on ad-hoc 802.11 networks for data transfer beyond one network hop.

Figure 2: An example of the ad-hoc mode.

The Physical Layer

A channel is a range of electromagnetic frequencies over which data is transmitted. This is much the same as the channel in radio, television, or cable TV. (The same word is used to refer to the medium itself as the wireless channel.)

A basic signal called carrier is used to carry the data. The carrier is modulated, i.e., modified using some algorithms at the source to encode data. This modulated carrier is demodulated, i.e., interpreted at the destination.

The carrier signal is spread across a channel or channels using specified schemes. The schemes used in 802.11 are Direct Sequence Spread Spectrum (DSSS) and Frequency Hopping Spread Spectrum (FHSS). The former uses a single channel and spreads the carrier in one wide channel, while the latter uses several channels for spreading the carrier signal. The reasons for spreading the signal are to prevent another source from jamming the signal and to decrease the probability of error. The 802.11b standard, which specifies the higher rates of 2Mbps and above, uses the DSSS exclusively.

IEEE 802.11 specifies the encoding schemes to be used to represent bits in terms of the carrier signal. This standard provides for 1Mbps, 2Mbps, 5.5Mbps, and 11Mbps. Different modulation schemes offer different bit rates over the same channel.

The hardware on the wireless card implements the PHY layer in an 802.11 station. This layer implements all the modulation, transmission, and carrier sensing. Carrier sensing reports to the MAC layer if the channel is busy or free. The MAC uses this to decide when to send data.

The 802.11 standard provides half duplex communication. A station cannot send as well as receive data simultaneously. This allows the same hardware (antenna, signal processors, circuitry, etc.) to switch between reception and transmission.

The Medium Access Control Layer

As the name suggests, the MAC layer decides when a station can access the medium, i.e., send data. The MAC can run either a distributed algorithm or a centralized algorithm to decide who can send data. The centralized algorithm is implemented on an Access Point.

The 802.11 standard defines both centralized (Point Coordination Function) and distributed algorithms (Distributed Coordination Function) for medium access.

The Distributed Co-ordination Function (DCF)

We can summarize DCF as "Listen before talk." Any station senses the media first to see if it is free for a specified duration. If so, it transmits; if not, it defers for a random duration. As the random duration could be potentially negligible, DCF also has mandatory spacing between any two successive frames called Inter Frame Spacing (IFS). The exact type of IFS used depends on the type of frame sent.

DCF provides a positive acknowledgement. A station acknowledges any unicast data frame (i.e., a frame meant for this station alone, not for the entire WLAN) it receives. DCF is an illustration of a family of protocols called the "Carrier Sense Medium Access with Collision Avoidance CSMA/CA". This differs from Ethernet, which is based on "Carrier Sense Multiple Access with Collision Detection." The key difference is the extra conservative approach of CSMA/CA in which it tries to avoid collisions by spending extra time sensing the medium. This provides greater robustness, as unlike Ethernet, 802.11 stations cannot sense collisions unlike. This is because 802.11 allows only for half-duplex communication. The details of DCF algorithm follow.

Sending a data frame:

Sense the media.
If the media is free for one DIFS (Data Inter Frame Spacing), go ahead and transmit.
If the media is busy any time during sensing, enable the back-off procedure.
If backing-off is enabled, choose a random number of slots for the back-off duration.
After waiting for the DIFS duration, wait an extra back-off duration.
If the medium becomes busy, freeze the back-off timer.
Once the medium becomes free, resume the back-off timer.
Once the timer expires, go ahead and transmit.

Receiving a data frame:

On receiving a data frame, wait for a SIFS (Short Inter Frame Spacing), and immediately send an ACK frame back to the sender.

Note that there is no carrier sensing or backing-off for an ACK.

By generating an ACK for every data frame, 802.11 trades off efficiency for throughput. By this we mean, instead of sending just a single data frame, 802.11 spends more time in sending an ACK back to the source. This is an important example of a design decision that is glaringly different from typical wired LANs like Ethernet. In Ethernet, there are no ACKs for any sort of reliability. However, the random errors in a wired LAN are negligible, and thus, there is no need for this trade-off.

DCF with RTS/CTS

The carrier sense mechanism not only implements physical carrier sense but also virtual carrier sense in this mode. A station sends a special frame called (RTS) Request to Send to its destination node. The destination sends a (CTS) Clear to Send to indicate permission to send. The RTS and CTS include fields that specify the time to be taken for sending the data frame and the ACK, which follows this RTS-CTS pair.

All stations near the sender hear the RTS. They update their network allocation vector (NAV) and defer from sending. Likewise, all stations in the neighborhood of the receiver hear the CTS and update their NAV to defer from sending. The NAV stores the time for which the medium is busy, and the PHY reports the medium as busy during this time. This is the virtual carrier sense mechanism.

Intuitively, the RTS/CTS reserves the medium for a specified duration of time. This reduces the probability of collisions and increases throughput under heavy loads (under light loads, RTS/CTS is more of an overhead as there aren’t many collisions anyway). Note that a station sends the RTS frame after deferring for the medium like the normal DCF.

The Hidden and Exposed Terminal Problems

Suppose A, B, and C are three stations that want to communicate (one of which could be an AP). If the station B is within the hearing range of both A and C, and A and C cannot hear each other, then A is a hidden terminal with respect to C and vice versa. This is because C could transmit to B when A is already transmitting to B (as it doesn’t hear A) causing a collision at B (Figure 3).

Figure 3: The hidden terminal problem.

Suppose A, B, C, and D are four stations, B wants to send to A, and C wants to send to D. If A and D are silent, both B and C can transmit at the same time causing a collision. Therefore, C is an exposed terminal with respect to B, and vice versa (Figure 4).

RTS/CTS solves these problems. This is because a station trying to send data will first send the RTS. Whoever succeeds in sending the RTS is guaranteed to have nearby sources silent when it is trying to send. Likewise, the CTS guarantees that stations near the receiver do not try to send data.

Figure 4: The exposed terminal problem/

The Point Coordination Function

PCF is a centralized function, in which the AP controls the medium access explicitly. DCF and PCF can co-exist. The AP signals the start of PCF mode by broadcasting a special beacon frame using DCF mode of access. Once this is done, the AP takes over control of the medium explicitly for duration called the Contention Free Period (CFP).

During the CFP, the AP polls each station in its list in a round-robin fashion. On getting a poll from the AP, a station responds with an ACK. If it has any data to send, the station responds to the poll with a data frame. In the case of an unsuccessful transmission, the station resends it during the next poll either in the same CFP or the next CFP. However, as there are likely to be no collisions (as stations talk only when the AP lets them), this is acceptable.

In order to preclude stations implementing DCF from disrupting PCF traffic, PCF frames use PIFS (Priority Inter Frame Spacing), which is smaller than the DIFS used by DCF. This means that the PCF station will get to transmit, and the DCF station, seeing the medium busy, will defer to PCF stations.

At the end of the CFP, the AP relinquishes control of the medium by broadcasting a special beacon frame again. This signals the stop of PCF and the normal mode of operation reverts to DCF.

Performance of 802.11 in Practice

In practice, most wireless cards and APs implement DCF, although a few cards support PCF. The firmware of the cards and the AP implement most of the MAC layer; users can control some settings using software.

DCF offers a throughput of about 6.5Mbps out of a bit rate of 11Mbps, when a single station transmits data [15]. This difference is because of physical layer headers, use of an ACK for each packet, and the mandatory delay between frames (Inter Frame Spacing). In other words, the protocol trades off throughput for robustness at the link layer. The performance degrades with increasing numbers of stations competing for the channel. This is due to collisions and back-offs in the medium access.

Fairness issues abound in 802.11 [12], as contention for the channel is dependent on the physical location of the stations. For instance, a station closer to the AP could potentially get throughput; this is not the case for stations on wired LANs like Ethernet.

Rapid deployment of 802.11b has led to the formation of wireless hotspots in public places like airport terminals, malls, etc., where many users try to access the WLAN simultaneously. This is quite similar to popular web servers attracting a lot of traffic.

Security

Since the medium is shared, everything that is transmitted or received over an 802.11 network can be intercepted. This is easier than a wired network as one does not need any physical tapping into wires. Therefore, encryption and authentication are important for a wireless network. The IEEE 802.11 standard provides the Wired Equivalent Privacy (WEP) protocol as a mechanism to make wireless traffic as secure as wired traffic, and hence the name [10, 16]. WEP provides a mechanism to encrypt the traffic and authenticate 802.11 nodes.

Wired Equivalent Privacy protocol

Encryption

WEP uses a key shared between the access point and the stations to encrypt and decrypt the data. The WEP encryption algorithm works as follows (Figure 5):

Figure 5: WEP encryption.

Two processes are applied to the plain text data. One encrypts the plain text; the other protects against unauthorized data modification. The secret key (40-bits) is concatenated with an Initialization Vector ("IV", 24-bits) resulting in a 64-bit total key size. The resulting key is input into the Pseudo-random Number Generator (PRNG). The PRNG outputs a pseudorandom key sequence based on the input key. The resulting sequence is used to encrypt the data by doing a bitwise XOR. The result is encrypted bytes equal in length to the number of data bytes that are to be transmitted in the expanded data plus four bytes. This is because the key sequence is used to protect the Integrity Check Value (ICV, 32-bits) as well as the data.

The above discussion assumes a 64-bit key. For a 128-bit key, the only difference is that the secret key size becomes 104 bits and the IV remains 24 bits.

To protect against unauthorized data modification, an integrity algorithm (CRC-32) operates on the plain text to produce the ICV.

The IV, plain text, and ICV triplet forms the actual data sent in the data frame (Figure 6). In decryption, the IV of the incoming message is used to generate the key sequence necessary to decrypt the incoming message (Figure 7). Combining the ciphertext with the proper key sequence yields the original plain text and ICV. The decryption is verified by performing the integrity check algorithm on the recovered plain text and comparing the output ICV' to the ICV transmitted with the message. If ICV' is not equal to ICV, the received message is in error.

Figure 6: WEP data frame.

Figure 7: WEP decryption.

There are two types of 802.11 authentication:

Open system authentication: This is the default authentication service.
Shared-key authentication: This involves a shared secret key to authenticate the station to the AP.

The open system authentication, usually implemented for simplicity, is null authentication; any station can associate with the access point.

The shared-key authentication approach provides a better degree of authentication. For a station to utilize shared-key authentication, it must implement WEP. Figure 8 illustrates the operation of shared-key authentication. The 802.11 standard does not specify how to distribute the keys to each station.

The process is as follows:

A requesting station sends an Authentication frame to the AP.
The AP will reply with an Authentication frame containing 128 bytes of random challenge text generated by the WEP engine in standard form.
The requesting station will then copy the challenge text into an Authentication frame, encrypt it with a shared key, and then send the frame to the AP.
The AP decrypts the value of the challenge text using the shared key and compares it to the challenge text sent earlier. If a match occurs, the AP replies indicating a successful authentication. If not, the responding AP will send a negative authentication.

The WEP PRNG (RC4) is the critical component of the WEP process, since it is the actual encryption engine. The IV extends the useful lifetime of the secret key. The secret key remains constant while the IV changes periodically. Each new IV results in a new key sequence, thus there is a one-to-one correspondence between IV and the output. The IV may change as frequently as every message, and since it travels with the message, the receiver will always be able to decrypt any message. As the data of higher layer protocols (e.g., IP) are usually highly predictable, an eavesdropper can readily determine portions of the key sequence generated by the (Key, IV) pair. If the same pair is used for successive messages, this effect may reduce the degree of privacy. Changing the IV after each message is a simple method of preserving the effectiveness of WEP as it makes it more difficult to determine the key.

Figure 8: Shared key authentication.

Is WEP Effective?

Several researchers worked on analyzing the WEP protocol. Researchers at Berkeley [6] demonstrated weaknesses due to key reuse and weak message authentication. A paper from University of Maryland [4] showed the weaknesses of 802.11 access control mechanisms, even those based on WEP's cryptographic authentication. A later paper showed that the weak message authentication made it possible to inject traffic into the network [5].

Fluhrer, Mantin, and Shamir found a flaw in the "key scheduling algorithm" of RC4 that made certain RC4 keys fundamentally weak, and they designed an attack that would allow a passive listener to recover the secret WEP key by collecting a sufficient number of encrypted frames [8].

802.1x

WEP was designed to provide both authentication and privacy, but had been shown to provide neither. To solve the user-authentication problem, the 802.11 working group adopted the 802.1x standard [3], which provides "per-port user authentication." It was designed to require user authentication before granting network access.

Another major underlying problem with the existing 802.11 standard is that the keys are cumbersome to change. In order to use different keys, each access point and NIC must be manually configured with the new keys.

The use of IEEE 802.1x offers an effective framework for authenticating and controlling user traffic to a protected network, as well as dynamically varying encryption keys. 802.1x uses a protocol called EAP (Extensible Authentication Protocol) to both the wired and wireless LAN media and supports multiple authentication methods, such as token cards, Kerberos, one-time passwords, certificates, and public key authentication. For details on EAP, the reader is refered to the IETF's RFC 2284.

Initial 802.1x communications begin with an unauthenticated supplicant (i.e., client device) attempting to connect with an authenticator (i.e., 802.11 access point). The access point responds by enabling a port for passing only EAP packets from the client to an authentication server located on the wired side of the access point. The access point blocks all other traffic until the access point can verify the client's identity using an authentication server (e.g., RADIUS). Once authenticated, the access point opens the client's port for other types of traffic.

The basic 802.1x protocol provides effective authentication regardless of whether the network uses WEP keys or no encryption at all. It's important to note that 802.1x does not provide the actual authentication mechanisms. When utilizing 802.1x, the user needs to choose an EAP type.

Even 802.1x is not perfect. 802.1x was designed for a wired network, which leads to several problems [11]. In the 802.1x, the AP is always trusted, which means that a man-in-the-middle attack can be carried by taking the role of the AP.

The Road Ahead

Since the ratification of the initial 802.11 standard, the IEEE 802.11 Working Group (WG) has made several revisions through various task groups. Task groups within the 802.11 WG enhance portions of the 802.11 standard. A particular letter corresponds to each standard/revision. Following is a list of WGs [1, 9]:

802.11a - OFDM in the 5GHz Band

802.11a is a Physical Layer standard (IEEE Std. 802.11a-1999) that specifies operating in the 5GHz band using orthogonal frequency division multiplexing (OFDM). 802.11a supports data rates up to 54Mbps. This is also called Wi-Fi5.

802.11b - High Rate DSSS in the 2.4GHz band

The task group for 802.11b was responsible for enhancing the initial 802.11 DSSS PHY to include 5.5Mbps and 11Mbps data rates in addition to the 1Mbps and 2Mbps data rates of the initial standard. 802.11 finalized this standard (IEEE Std. 802.11b-1999) in late 1999. To provide the higher data rates, 802.11b uses CCK (Complementary Code Keying), a modulation technique that makes efficient use of the radio spectrum.

Most wireless LAN installations today comply with 802.11b, which is also the basis for Wi-Fi certification from the Wireless Ethernet Compatibility Alliance (WECA).

802.11c - Bridge Operation Procedures

802.11c provides required information to ensure proper bridge operations. This project is completed, and related procedures are part of the IEEE 802.11c standard.

802.11d - Global Harmonization

In order to support a widespread adoption of 802.11, the 802.11d task group has an ongoing charter to define PHY requirements that satisfy regulatory within countries other than U.S., Europe, and Japan. This is especially important for operation in the 5GHz bands because the use of these frequencies differ widely from one country to another.

802.11e - MAC Enhancements for QoS

The 802.11e task group is refining the 802.11 MAC (Medium Access Layer) to improve QoS for better support of audio and video (such as MPEG-2) applications.

802.11f - Inter Access Point Protocol

802.11f is working on specifying an inter access point protocol that provides the necessary information that access points need to exchange to support the 802.11 distribution system functions (e.g., roaming).

802.11g - Higher Rate Extensions in the 2.4GHz Band

The aim of the 802.11g task group is to develop a higher speed extension (up to 54Mbps) to the 802.11b PHY, while operating in the 2.4GHz band.

802.11h - Spectrum Managed 802.11a

802.11h addresses the requirements of the European regulatory bodies. It provides dynamic channel selection (DCS) and transmit power control (TPC) for devices operating in the 5GHz band (802.11a). In Europe, there is a strong potential for 802.11a interfering with satellite communications, which have "primary use" designations. Most countries authorize WLANs for "secondary use" only. Through the use of DCS and TPC, 802.11h will avoid interference in a way similar to HiperLAN/2, the European-based competitor to 802.11a.

802.11i - MAC Enhancements for Enhanced Security

802.11i is actively defining enhancements to the MAC Layer to counter the issues related to WEP. 802.11i will incorporate 802.1x and stronger encryption techniques, such as AES (Advanced Encryption Standard).

Recent Trends in WLAN Research

Research in WLAN protocols at the MAC layer is primarily focused on providing QoS. Although the 802.11e standard is addressing this issue, many solutions have been proposed in recent literature [17].

Another area of active research is co-existence of 802.11 with Bluetooth and other standards which operate in the same 2.4GHz band [2]. Protocols are being studied for Bluetooth to detect interference from 802.11 and adapt its scheduling appropriately.

An area which holds great promise is location determination of mobile users [18], in which signal strengths from different APs at that user location is used to determine the location of the user.

Perhaps the killer application for WLANs is providing the communication technology for pervasive, ubiquitous, nomadic, and sensor network-based computing.

Concluding Remarks

802.11 is fast growing technology that is being rapidly deployed, due to the industry adopting it wholeheartedly. We have presented an overview on the basic schema of 802.11 protocols. In addition, we also surveyed the state of the art with the latest working groups constituted by IEEE to address pertinent issues and recent trends in WLAN research.

References

1: The 802.11 working group home page. <http://grouper.ieee.org/groups/802/11/>.
2: The 802.15 Task Group 2 (TG2) home page. <http://www.ieee802.org/15/pub/TG2.html>.
3: The 802.1x specification. <http://standards.ieee.org/getieee802>.
4: William A. Arbaugh, Narendar Shankar, and Y.C. Justin Wan. Your 802.11 Wireless Network has No Clothes. IEEE International Conference on Wireless LANs and Home Networks, December 2001.
5: William A. Arbaugh. An inductive chosen plain text attack against WEP/WEP2. IEEE Document 802.11-01/230, May 2001. <http://grouper.ieee.org/groups/802/11/Documents/index.html>.
6: Nikita Borisov, Ian Goldberg, and David Wagner. Intercepting Mobile Communications: The Insecurity of 802.11. Proceedings of the Seventh Annual International Conference on Mobile Computing And Networking, July 16-21, 2001.
7: Federal Communications Commission web site. <http://www.fcc.gov> (11/10/2002).
8: Scott Fluhrer, Itsik Mantin, and Adi Shamir. Weaknesses in the Key Scheduling Algorithm of RC4. Eighth Annual Workshop on Selected Areas in Cryptography, August 2001.
9: Jim Geier. 802.11 Alphabet Soup. <http://www.80211-planet.com/tutorials/article.php/1439551>.
10: The Institute of Electrical and Inc. Electronics Engineers. IEEE Std 802.11 - Wireless LAN Medium Access Control (MAC) and Physical Layer (PHY) specifications. 1999.
11: Arunesh Mishra, and William Arbaugh. "An Initial Security Analysis of the IEEE 802.1x Security Standard." February 6, 2001. <http://www.cs.umd.edu/~waa/1x.pdf>.
12: Thyagarajan Nandagopal , Tae-Eun Kim , Xia Gao , Vaduvur Bharghavan. Achieving MAC layer fairness in wireless packet networks. Proceedings of the sixth annual international conference on Mobile computing and networking, Boston.
13: Royer, E.M., Toh, C.K. A Review of Current Routing Protocols for Ad Hoc Mobile Wireless Networks. IEEE Personal Communications, Vol. 6., No. 2, April 1999 pp 46-55
14: William Stallings, "Wireless Communications and Networks ", Prentice Hall.
15: Arunchandar Vasan, A. Udaya Shankar. An Empirical Characterization of Instantaneous Throughput in 802.11b WLANs. <http://www.cs.umd.edu/Library/TRs>.
16: Sultan Weatherspoon. Overview of IEEE 802.11b Security. Intel Technology Journal, 2nd Quarter 2000. <http://www.intel.com/technology/itj/q22000/articles/art_5.htm>.
17: Drabu, Yasir. A Survey of QoS Techniques in 802.11. Department of Computer Science, Kent State University. <http://trident.mcs.kent.edu/~ydrabu/research/wmac/mac.pdf>.
18: M. Youssef, A. Agrawala, A. U. Shankar, and S. Noh. A Probabilistic Clustering-Based Indoor Location Determination System. <http://www.cs.umd.edu/~moustafa/papers/locdet_tr.pdf>.

Biographies

Moustafa Youssef (moustafa@cs.umd.edu) received his BSc and MSc in Computer Science from Alexandria University, Egypt in 1997 and 1999, respectively, and an MS from University of Maryland in 2002. He is currently a Ph.D. candidate at University of Maryland at College Park. His research interests include location determination techniques in WLANs, context-aware computing, energy-aware computing, and protocol modeling. Moustafa is a graduate fellow of the Computer Science department, University of Maryland, a life fellow for the Egyptian Society for Talented, an elected member of the honor society Phi Kappa Phi, among others. He is a member of various professional societies such as IEEE, IEEE Computer Society, and IEEE Communication Society.

Arunchandar Vasan (arun@cs.umd.edu) received his Bachelor of Technology in Computer Science and Engineering from the Indian Institute of Technology Madras in 2000 and expects to receive his MS in Computer Science in 2002 from the University of Maryland. He is currently a Ph.D. student at the University of Maryland, College Park. His research interests are in performance evaluation of large scale networks, QoS over WLANs, peer-to-peer protocols, and formal modelling. He is a recipient of scholarships, awards, or grants from the Central Board of Secondary Education of India, the Government of Delhi, the IIT Madras, DARPA, and ACM SIGMETRICS among others. He was elected member of Phi Kappa Phi in 2002.

Crossroads

The ACM Student Magazine