MMIP: A New Dynamic IP Configuration Scheme with MAC Address Mapping for Mobile Ad hoc networks

Transcription

1 NCC 2009, January 16-18, IIT Guwahati 425 MMIP: A New Dynamic IP Configuration Scheme with MAC Address Mapping for Mobile Ad hoc networks Uttam Ghosh and Raja Datta, Member, IEEE Department of Electronics and Electrical Communication Engineering Indian Institute of Technology, Kharagpur, India, Kharagpur uttamg@iitkgp.ac.in, rajadatta@ece.iitkgp.ernet.in Abstract In this paper we present a novel dynamic IP configuration scheme that maps the MAC address for Mobile Ad-hoc Networks (MANET). Though not explicitly, the proposed scheme will also help in securing and keeping MANET organized as it grows from one node to many. It has been seen that within a lifetime of a MANET an individual node may exit and again re-enter the network as per its requirement. In most existing addressing schemes a node resurfaces with a new IP address to continue its work. As the IP address gets changed every time a node switches off and on within the network lifetime, the process of unique IP address assignment, partitioning and merging of the network becomes complex. In this work we propose a technique to map the MAC addresses of the nodes along with the IP addresses which are assigned at the time when a node enters the network. Performance analysis shows our proposed addressing scheme has less addressing latency and control overhead compared to the similar existing schemes. I. INTRODUCTION IP configuration in mobile ad hoc networks (MANET) is a prime concern as it requires that the allocated addresses are unique and robust in nature. Further, the scheme should have low addressing latency and communication overhead. Unlike in infrastructure based wireless networks, the unique characteristics of MANETs such as open network architecture, shared wireless medium, stringent resource constraints and highly dynamic network topology pose new challenges. Security is also a concern as till very recently research done on MANETs assumed a friendly and cooperative environment, which may not apply in real-life scenarios. Typical real-life scenario may be a MANET set up in a battlefield or in case of natural calamity or simply to facilitate outdoor conferences. Several schemes have been proposed in the literature for dynamic IP configuration. This work focuses on the problem of unique address allocation in an independent MANET where network security is also a consideration. Recently, a number of addressing schemes for ad hoc networks have been proposed [3] [7], all of which aim to provide efficient address assignment in a dynamic network environment so as to enable correct communication in the network. These approaches bear many similarities to each other, such as self-organizing, self-healing behavior in order to better adapt to the dynamic and resource-constrained environment of the ad hoc network. However, these approaches also differ in a wide range of aspects, such as address format, usage of centralized servers or full decentralization, hierarchical structure or flat network organization and explicit or implicit duplicate address detection. As a consequence, these approaches are likely to have different performance properties under varying network conditions. For example, due to the limited resources of the mobile devices as well as of the wireless network, the address assignment process should not incur significant traffic load on the network. Limiting the control traffic generated by the assignment process enables the protocol to more easily scale to larger networks. On the other hand, a node should obtain a valid address in a timely fashion, regardless of the network size, so that its communication is not delayed. To deploy largescale mobile ad hoc networks, it is important to understand and analyze these different approaches under a wide range of network conditions so that their performance and suitability can be predicted. Traditionally, a user can either configure the address of a host manually or have the host acquire its IP address dynamically through certain dynamic methods, such as Dynamic Host Configuration Protocol (DHCP) [1]. Manual address configuration is very troublesome and in most cases inapplicable to MANETs. However, DHCP requires the presence of a centralized DHCP server which maintains the configuration information of all hosts in the network. Since a MANET is devoid of any fixed infrastructure or centralized administration, this approach cannot be used. The existing IP address allocation schemes can be classified into three categories [2]. The best effort allocation schemes do not guarantee address uniqueness. Under this the Prophet scheme [5] propose a function f(n) to generate a series of random numbers for address allocation. But prophet scheme needs some mechanisms, such as passive duplicate address detection (DAD) [8] to resolve address conflicts. In Passive DAD, nodes use periodic link state routing information notify other nodes about their neighbors. But this causes broadcast storm problem in passive DAD. The second category is the Leader-based allocation scheme. In this scheme nodes obtain valid IP addresses from an elected leader or server of the network e.g., DHCP [1]. In DHCP, a new node needs to broadcast to discover the server discovery and then use DAD to verify the uniqueness of the IP address. In ODACP [2], instead of discovering the server, server periodically broadcasts the address request to reduce the overhead of broadcasting. But it requires longer latencies for hosts to obtain addresses. Decentralized allocation is the third category in which a host can acquire an IP address either itself or from a neighbor and then performs the DAD to ensure the uniqueness of the address. In MANETconf [4] scheme, every node maintains a

2 NCC 2009, January 16-18, IIT Guwahati 426 list of IP addresses which are in use by the network. Network partitions and mergers are detected throughout the DAD. In Prime DHCP [6], address can be allocated to the new host without broadcasting over the whole MANET. Prime DHCP makes each host a DHCP proxy of the MANET and run a prime numbering address allocation algorithm individually to compute unique addresses for address allocation. In this paper we propose a distributed dynamic IP configuration scheme that maps the offered IP address with corresponding MAC address of each host. Our distributed protocol ensures that no two hosts in the MANET acquire the same IP address. The scheme ensures that during a MANETs life time a node within that MANET will not be able to resurface (after switching off) with a different IP address to continue with its work. This also helps in securing the network though not explicitly. We describe enhancements to the solution that can handle problems that may arise due to host failures, message losses, mobility of the hosts and network partitioning and merger. The rest of the paper is organized as follows: We describe the system model in section II. In section III we present our dynamic address allocation algorithm (MMIP). Section IV gives the performance comparison and conclusions are given in section V. II. SYSTEM MODEL We consider an autonomous ad hoc network working on its own. It has no gateway or connection to the external world. The network is formed starting from one node and then the other nodes add up one by one. The nodes are free to move around and can join or leave the network at any point of time. Hence the size and the topology of the network is dynamic and unpredictable in nature. We define the lifetime of a MANET as the time from which the first node configures itself with an IP address till the time when all the nodes are switched off, i.e., the time when the MANET ceases to exist. A. Protocol Requirements: For assigning an IP address, the protocol should meet the following requirements: i. Each node should have an unique IP in the MANET, i.e., at any given instant of time there should not be two or more nodes with the same IP address. ii. The protocol should ensure that in the event of a node switching off and trying to join the network sometime later (within the lifetime of the network), the IP assigned to the node remains unchanged and cannot be changed under any circumstances. iii. A node should be denied an IP address if and only if there are no available IP addresses in the network. iv. The protocol should handle the network partitioning and merging. When two different partitions merge, there is a possibility that two or more nodes have the same IP address. Such duplicate address should be detected and resolved. III. THE MMIP ALGORITHM In this section we present our proposed algorithm for dynamic IP configuration where the MAC address is mapped with the allotted IP address. We call this algorithm as MAC Mapped IP (MMIP) algorithm. Mentioned in section I, most of the address allocation schemes for MANET rely on broadcasting a message to discover the server or uses DAD. Also, none of the existing schemes have considered the security of the MANET. Here we propose a technique where every node acts as a proxy to a new node (N n ). Therefore all the nodes are eligible to assign addresses and a new node N n can acquire an address simply from its neighbors. Each proxy computes a unique IP address for a new host N n from its own IP address and as a result DAD is not necessary here. The MMIP address allocation algorithm has two parts: one for the new node (N n ) and the other for the proxy that assigns its IP address. The MMIP algorithm for a new node and for the node acting as proxy using pseudo code is given in Algorithm 1 and Algorithm 2 respectively. Algorithm 1: Address Allocation for host N n Set threshold 2 Set begin true Set configured false Set counter 1 if begin = true and counter threshold then 1-hop broadcast DISCOVER message start offertimer begin false else exit if multiple OFFER(NEWIP) messages is received from proxies then stop offertimer select minimum IP from OFFER messages send SELECT message to selected proxy start miptimer else if OFFER(OLDIP) messages is received from proxies then send STATREQ message to parent if STATUS message is received then update allocation status if MIP is received from selected proxy then stop miptimer configured true MAPPING( ) refer to function. SWITCHOFF() refer to function. if timeout(offertimer) if timeout(miptimer) a) MMIP address allocation: When a new node N n wants to join a MANET, proposed MMIP algorithm periodically issues a DISCOVER broadcast message to its neighbors till it either receives an OFFER message or a DENY message. If no OFFER or DENY message is received, the new node N n configures itself to the IP address 169.X.1.1 and generates an unique network ID (NID, which is random number) as a network identifier. This NID is sent to the new node whenever the node joins that MANET. The DISCOVER message contains its MAC address as an identifier of the host N n and as the IP address (refer Algorithm 1). The new node N n then waits for multiple responses from its neighbors and chooses the smallest IP address (last byte of IP address i.e., the value of i) that has been offered to it. This smallest IP address is then unicast in a SELECT message back to the proxy offering that IP address. The other OFFERs are ignored. On receiving the SELECT message from the new node N n,the MAC address of N n is now mapped with the said IP address

3 NCC 2009, January 16-18, IIT Guwahati 427 Algorithm 2: Address Allocation for the proxy Set threshold 2 Set begin true Set counter 1 if DISCOVER message is received from N n then if MAC address in RESERVE( ) then send previously allocated OLDIP in OFFER message if STATREQ message is received from children then send allocation status in STATUS message else if MAC address in MAPPINGTABLE( ) then send previously allocated OLDIP in OFFER msg. else if free IP address is available then NEWIP = GENERATEUNIQUEIP( ) send OFFER message with NEWIP to N n else send REQUEST message to parent for NEWIP if NEWIP is received from parent then send NEWIP in OFFER message to N n else send DENY message to N n if SELECT message is received from N n then broadcast MIP message if begin = true and counter threshold then send UPDATE message to N n start updatedtimer begin false UPDATEMAPPINGTABLE( ) refer to function. if DEPOSIT message is received from children then RESERVE( ) refer to function. if timeout(updatedtimer) and is encapsulated in a MIP (Mapped IP) message. This MIP message is then broadcasted to all nodes within its radio range (including N n ). After receiving the MIP message from the selected proxy, N n performs a final check on the configuration parameters (e.g., ARP for allocated network address) specified in the MIP message and configures itself. The neighbors after receiving the MIP message broadcasts it to its neighbors and so on such that the whole of the MANET receives it within a short time. The nodes then updates their mapping table by inserting the MAC address and the IP address of the new host N n. In case N n does not receive any OFFER message from the proxy within a given time it retransmits the DISCOVER message. When the neighbor host receives DISCOVER message (Algorithm 2), they start serving as proxies of the host N n if the MAC address and a corresponding IP address of the host is not found in the mapping table (i.e., when the said node has not been allocated an IP address earlier anytime during the lifetime of the MANET). Here, if the proxy finds the MAC address and a corresponding IP address of the host N n in the ReserveList (refer Function RESERVE), then the said allocated IP address will be again assigned to N n. b) Unique IP address generation: Here we describe the algorithm given in Function GENERATEUNIQUEIP that generates unique IP address for a new node. As mentioned earlier each network node acts as a proxy and can assign IP address for a new node. We assume the class B IP address format 169.X.j.i for describing the algorithm and for the purpose of illustration. We also assume that the first node s IP address is 169.X.1.1 that initiates the address allocation process. Here 169.X be the net ID which is fixed for a network and the values i and j are Function GENERATEUNIQUEIP getmyip 169.X.j.i Set static count 1 Set I ((i 1) K + count) Set m Imod254 Set J (j + m) if I 1and I 254 and count K then count (count +1) return NEWIP 169.X.j.I else if I 255 and I 508 and J 254 and j =1and count K then I 1 count (count +1) return NEWIP 169.X.J.I else IP is not available variables. Thus, the address can be allocated from 169.X.1.1 to 169.X for the new node. Here each proxy can assign from ((i 1) K + count) to (i K), wherei is the last byte of its own IP address. K( 1) is the number of total nodes that a proxy can assign and count is initialized to 1 for each host in the MANET and is incremented by 1 after each IP assignment for the new node by that proxy j=2 j= Fig. 1. An example of address allocation tree, wherein K = 10 Fig 1 gives an example of how unique address can be allocated by a node acting as proxy. In the figure the last byte of an IP address (i) is shown within the circle and the corresponding third byte (j) is shown just outside the circle. In this illustration we have assumed K =10. The first node of the network which is shown as the root node in the figure has an address of i =1, j =1and can allocate from i =2, j = 1 to i = 10, j = 1 and the node having i = 2, j =1can assign from i =11, j =1to i =20, j =1. Similarly, a proxy having i =26, j =1can allocate addresses from ((26 1) ) mod 254 i.e. i = 251, j =1to (26 10) mod 254 i.e. i =1, j =7. Also a node having i =1and j =2can assign addresses having i =2, j =2 to i =10, j =2. Due to the limited IP address range, some proxies may not be able to allocate addresses to new nodes. In that case the proxies will have to request their parent proxy for an IP address that can be allocated to the new node. In this way the network grows up to the largest address bounded by the address space. Thus, the address can be allocated from 169.X.1.1 to 169.X in the network uniquely. In this algorithm each node has to maintain its allocation status that is the value of count to record the last assigned address. Here, no two nodes acting as proxies can generate the same IP address, and thus DAD is not required during the process of address resolution. Besides, each proxy can easily derive, from its own address and the value of K, the address

4 NCC 2009, January 16-18, IIT Guwahati 428 of its parent proxy. c) Graceful Switch-off of node: In case of our proposed MMIP algorithm every node has to Gracefully Switch-off whenever it wants to permanently or temporarily switch off itself from the network. This is required firstly to inform about its intention and also to transfer its allocation status to its parent node prior to switching off. This allocation status is reverted back to the same node when it wants to re-enter the network some time later. This is required because the allocation status contains the value of count which is used to generate unique IP address. After a node re-enters the network, if it still has available IP addresses, it may have to assign them to the new nodes entering the network for the first time. The value of count in the allocation status before a node has gracefully switched off will then be used so that unique IP is generated as per the Function GENERATEUNIQUEIP. The algorithm for a node that wants to do graceful switch-off is given in the Function SWITCHOFF and the corresponding parent proxy updates the reserve list is given in Function RESERVE. Function SWITCHOFF if configured = true and counter threshold then send DEPOSIT message to parent start oktimer if OK message is received from parent then stop oktimer configured false counter 1 if timeout(oktimer) then Function RESERVE if DEPOSIT message is received from children then send OK message to the host N n updatereservelist( ) The Function SWITCHOFF and the Function RESERVE works as follows: When a node (other than the root node) decides to switch off either permanently or temporarily within the lifetime of a MANET, it informs its parent by sending DEPOSIT message and waits for an OK message. If the root node decides to switch-off the network, it has to inform its child node (from among all children nodes) who has been allocated the minimum IP address. This child node becomes the new root node of the network. When a parent node receives the DEPOSIT message, it updates the reserve list and records the allocation status for that node. When a node that has switched off earlier wants to rejoin the MANET, it has to broadcast the DISCOVER message to the neighbors as is done in case of a new node. When the neighbors receive the DISCOVER message, it searches the reserve list to check whether the said node has been previously allocated with an IP address or not. In case the neighbors find that it is already allocated with an IP address it sends back the same IP address. The concerned node configures itself with the said IP address and sends information to the original parent node to obtain the allocation status which it has sent to it before switching off. Subsequently the original parent node sends the allocation status which enables the concerned node to act as proxy to new nodes. This is possible if there are IP addresses still available with him. The scheme ensures that once a host is assigned an IP address, it will never be able to change it within the lifetime of the MANET and also ensures the uniqueness of the IP addresses in the network even if the network is partitioned or merged. If the host leaves the network without sending the DEPOSIT message, it will be treated as an address leak. Thus a host can leave only after getting the OK message from its parent. d) Network Partitioning and Merger: In the initial state of address allocation, more than one node may configure themselves with the same IP address. This duplication of IP addresses can be detected when the nodes with different NIDs come in the radio range of each other. The difference in NID can be detected through the HELLO messages used by the routing protocols. The node with larger NID resets its configuration and reconfigures again using the MMIP algorithm. The network can partition and then again can merge at any instant of time due to the dynamic and unpredictable nature of the MANET. In our proposed MMIP protocol there will be no address conflicts in the network even if the network partitions. The split MANET needs a new root only. Though, the split MANET can continue with the same NID. Even if a node within the lifetime MANET switches off from the network, the IP address of the said node will not be allocated to any other host as described earlier. Therefore, there will be no address conflicts even if two split networks merge again as every node has the capability of generating unique IP addresses for a new host. In addition to this the proposed protocol also handles the following scenarios efficiently: i. A node which has not undergone a graceful switch-off (usually a malicious node) will be automatically caught if it tries to re-enter the network within its lifetime. ii. The MANET can partition and then the split networks can merge later without any address conflict. iii. Two separately configured MANETs i.e., if two different MANETs having different net IDs may merge without any conflicts of the IP addresses. e) Constructing and updating Mapping Table: In our proposed MMIP scheme, the mapping table is very important and has to be updated whenever a new node enters the network. Each node in the MANET, maintains a mapping table where the MAC addresses are mapped with the IP addresses of each node. Mapping table is created for a new host N n, after getting the IP from its parent proxy. Each node acting as Proxy in the MANET updates their mapping table by inserting MAC address and IP address for the new host N n. The algorithm for the mapping table for the new host N n and the proxy is given in Function MAPPING and Function UPDATEMAPPINGTABLE respectively. Function MAPPING create mapping table if UPDATE message is received from parent then send UPDATED message to parent updatemappingtable() Function UPDATEMAPPINGTABLE if UPDATED message is received from N n or MIP from other proxy then stop updatedtimer updatemappingtable()

5 NCC 2009, January 16-18, IIT Guwahati 429 IV. PERFORMANCE COMPARISON Table I presents the comparison of our proposed MMIP scheme with the existing dynamic addressing approaches. We focus on qualitative evaluation of all the approaches. Suppose the number of mobile hosts is n, the number of links is l, the average 1-hop latency is t and the network diameter is d. The existing DHCP [1] gives the guarantee of the uniqueness but can not be deployed in mobile ad hoc network. Further, DHCP needs to find out the server. Thus the latency is 4 t d and communication overhead is O(n 2 ). Another dynamic allocation scheme MANETconf [4] requires a positive acknowledgment from all known nodes indicating that the address is available for use. DAD is also necessary for MANETconf. Thus, the latency of MANETconf is 2 t d and the communication overhead is O(n 2 ). In ODACP [2], every host need to register with an address authority to reduce the communication overhead from O(n 2 ) to O(2l) and the latency from 4 t d to 2 t d. Prophet [5] is the only mechanism that can not guarantee about the uniqueness of addresses. Both Prophet and Prime DHCP [6] send their request to neighbors for an IP address and therefor the latency is 2t. The communication overhead of Prophet and Prime is the average degree (n/2) of each node in the network. MANETconf and Prophet are complex address allocation scheme. Most of the approaches use explicit mechanisms to detect network events such as partitions and merges except Prophet and our proposed MMIP acheme. The detection is normally accomplished by utilizing a unique network identifier which is either broadcast throughout the network by a leader node, or is contained in periodic hello messages (refer to P.Msg in table I) exchanged between neighbors. Finally, none of them have considered MANET security. TABLE I COMPARISON OF DYNAMIC ADDRESSING APPROACHES Metrics DHCP MANETconf ODACP Prophet Prime DHCP MMIP Uniqueness Yes Yes Yes No Yes Yes Latency O(4td) O(2td) O(2td) O(2t) O(2t) O(2t) Overhead O(n 2 ) O(n 2 ) O(2l) O(n/2) O(n/2) O(n/2) Complexity Low High Low High Low Low P.Msg Yes Yes Yes No Yes No Security No No No No No Yes In the proposed MMIP, every proxy in the network offers an unique IP address to a new host and that the IP cannot be changed during the lifetime of the MANET. So, unique IP is guaranteed in the network at any instant of time even after any partitions or mergers. After the partition, network can grow independently and if the partitions are merged at any later stage, there is no need to change the IP addresses of the nodes. This shows that the scheme is perfectly stable and robust. If a node initiates the address allocation procedure in the network then the proposed scheme eliminates the periodic message exchange between neighbors to detect network events such as partitions and mergers. The MMIP host N n sends request to neighbors only for an IP address, thus the latency is 2t and the communication overhead is the average degree (n/2) of each node of the network, assuming that the address space is sufficient. Thus, the latency and the overhead is less for the address allocation. To update the mapping table for a new host, we need to flood the MIP message in the network. Thus, the communication overhead is O(n 2 ). As there is no need to maintain any block of addresses nor there is a requirement to generate any complex function for an IP address the complexity is low for the address allocation in our scheme. Also the IP addresses for new nodes are generated from a node acting as proxy which reduces the complexity even further. The memory requirement for mapping table depends on the size of the network. If there are n nodes in the network, then the number of entries in the table are equal to the (n 1). Finally, it will be difficult, if not impossible, for a node to shed its IP address by switching itself off from the network, as its MAC address is mapped with the allocated IP address. Even though it is possible to change the MAC address of a node, our proposed MMIP scheme makes a network somewhat secure as it ensures that a node within the MANET will not be able to resurface (after switching itself off) with the original MAC address and a different IP to continue its work. V. CONCLUSION In this paper we proposed a dynamic IP address allocation algorithm for mobile ad hoc networks. The algorithm assumes the importance of an organized MANET along with low latency and communication overhead. The scheme maps the MAC address of a node with the allocated IP address. It ensures that a node will not be able to change its allocated IP address within the lifetime of a network if of course the MAC address of the node is not changed altogether. This eliminates the periodic message exchange between neighbors to detect network events such as partitions and mergers. In the algorithm, every node in the network also acts as a proxy and has the capability to assign IP addresses to new hosts. The signaling messages except MIP message need not be flooded all over the MANET saving considerable bandwidth. In addition, as each host can assign a unique IP address for a new host, DAD broadcasting is not required and is capable of handling host arrivals and departures. Performance analysis shows that our proposed scheme has low complexity, low overhead and is robust in comparison to the existing addressing schemes. REFERENCES [1] R. Droms, Dynamic host configuration protocol, RFC 2131, Mar [2] Y. Sun and E. M. Belding-Royer, A study of dynamic addressing techniques in mobile ad hoc networks, Wireless Communications and Mobile Computing, April [3] C. E. Perkins, J. T. Malinen, R. Wakikawa, E. M. Belding-Royer, and Y. Sun, IP address autoconfiguration for ad hoc networks, draftietfmanetautoconf- 01.txt, IETF MANET Working Group, July [4] S. Nesargi and R. Prakash, in MANETconf: Configuration of Hosts in a Mobile Ad Hoc Network, in Proc. INFOCOM, 2002, pp [5] H. Zhou, L. M. Ni, and M. W. Mutka, Prophet address allocation for largescalemanets, inproc. INFOCOM, 2003, pp [6] Y. Hsu and C. Tseng, Prime DHCP: A Prime Numbering Address Allocation Mechanism for MANETs, in IEEE Communicatons, August [7] M. Mohsin and R. Prakash, IP address assignment in mobile ad hoc networks, in Proc. IEEE MILCOM, pp , Sept [8] K. Weniger., Passive Duplicate Address Detection in Mobile Ad Hoc Networks, In WCNC, Florence, Italy, February 2003.