Trust Transfer Mechanism for Mobile Internet Access

Multimedia applications bring demands on broadband real-time data transfer, which needs resource reservation protocol (RSVP) helps to satisfy. However, handover processes bring delay, and existing authentication mechanism makes it further worse. Based on Pareto distribution patterns of mobile station communication traffic, proposed a trust transfer mechanism to execute pre-anticipated trust transfer and shorten the delay. It means that the dominant minority correspondent nodes designed in the VIP address books are informed about trust context messages, such as the information of handover, AAA, QoS, before anticipated handover. With the trust transfer mechanism, the simplification can be made to the authentication mechanism, which divides trust handover sequencing into several paralleled parts and budges some steps between stations. Thereby cut processes on each node by processes simultaneously carrying. Simulation with a modified NS2 platform showed that when time setforward equals 0.2, the delay is reduced by about 19.12% and the resource possession of RSVP is mitigated by 30.13%.


Introduction
High-speed mobile multi-media in portable Internet becomes multimedia applications rapidly advance in mobile Internet, with demands of broadband and real-time transfer. Only mobile IP scheme such as Hierarchical Mobile IP [1] or fast handover Mobile IPv6 [2] cannot support it very well, as a consequence of challenges, such as seamless handover (HO) and the QoS (Quality-of-Service) during the HO process [3]. It must work with the resource reservation protocol (RSVP) to ensure the QoS, broadband transfer and real-time transfer.
However, though it is designed for QoS, RSVP is so difficult to execute in handover processes because MIP buffer packets and then buffer enlarges delay so QoS demands are hard to be satisfied. Moreover, packets in tunnels cannot get the same QoS. MRSVP, MIPRSVP, and Dynamic RSVP, series of RSVP protocols [4] all requests reservation resources be reallocated before handover (HO). Since next-cell prediction is uncertain, it can but reserve resources at all possible place, which makes high frequency of IP-in-IP tunnel, afterwards, introduces extra overhead.
Further, the current authentication mechanism makes things worse. It uses the AAA server to carry authentication security, with plenty of complexity. Diameter AAA access authentication defined by IEEE needs 18 steps to build the authentication security process [5], while Radius is up to 32 steps, is exacerbated by the delay [6], which can make resource reservation timeout in great probability, not to mention services.
We designed a mechanism for the dominant minority based on Pareto pattern on communication traffic of mobile station, then, showed more details in trust transfer mechanism. Analysis and the simulation validated the availability.  Table 1 shows common acronyms appear in this paper.

Trust Transfer Processes
Noted that majority of communications traffic is for a limited size m of callers or callees in corresponding nodes (CN) group [7], set up the VIP address book contains m members (m=20 in our experiment, but can vary according to user level requirements for QoS). m denotes mobile nodes and mobile routers (MR) on which traffic concentrate, so that 95% of call traffic can be covered. So it means ∃ m, ∀ ( ) P m ξ > ≥ 95%. The classified aggregations of the data distributions on the communication traffic comply with Pareto distribution described as following: Trust transfer was optimized for the dominant minority group of callers or callees, which defined in a special VIP address book. Aiming to improve the 95% real-time services, mobile station will keep all these CNs informed of trust context messages in advance. So the above method was called address book informing, or ABI. Trust context in ABI messages contains QoS, AAA and HO messages, such as expiration data, the current Care of Addresses (CoA) of mobile nodes (MN) and the list of possible next Access Router (AR). Trust context transfer mechanism with ABI is just a frame containing information. QoS, AAA, and HO information will be forwards to relational nodes.
When in handover, trust transfer with ABI can also make the function. The original idea of tunneling buffer packets is smooth HO, but introduces more delay and difficulty of guaranteeing QoS in tunnel. With the mobile speed of mobile node (MN) increases, two and more tunnels may be needed. Things turn worse. Now if binding update cannot be achieved in time, previous access router (PAR) originally forwards packets from its corresponding CNs to the home agent, but now it is PAR can send the packets to next AR (NAR), which does not take a detour and then makes high-frequent IP-in-IP tunnel.
To avoid RSVP in tunnel, some literatures adopted the similar method, for example, Chaudet introduced mobility notification and location register [8], Yet cannot solve the call initialization question. Of course, MN have informed each of its CN of its CoA, but most of them achieve the optimized routing only when MN initializing a call or after CN initializing a call. However, if it is when CNs initialize a call, since CNs cannot know the CoA of MN and still have to go through HA of MN, there still forms a similar triangle routing. It prolongs waiting time of initializing. The phenomenon exists in almost all mobility schemes, no matter in HMIPv6, FastMIPv6, or the HADL scheme based on dual-link. The higher mobility speed of MN is, the worse the problem becomes, because the frequency of third party HO is increasing, the more cooperation steps between MN and ARs must be finished.
In the call initializing process with ABI, once a MN receives all these CNs context messages informed, it can prepare for the HO and be ready to rebuild an RSVP QoS path. Since MN and the current AR have known the next AR, most IP-in-IP tunnels can be avoided lest RSVP expires on preparing period. Therefore routing, HO and QoS got further optimized, which speed up the initialing process and partly solves the delay in the HO process. It is by multicast or by unicast that AAA, QoS and HO context can be sent to MN's probable next cell. By multicast is a better method. If only the contexts can be transferred before HO, the HO delay will be shortened and MSPEC can be minimized.

814
Information Technology for Manufacturing Systems III

Figure 1. Time Sequence of Trust Transfer Mechanism
Trust transfer context information can help to simplify the AAA process. There are 10 steps processes before HO, and six steps processes in HO step, which means greatly reducing in the AAA steps.
When a MN is approaching the rim of the edging of a cell, the trigger mechanism of trust transfer mechanism will be triggered preparing for HO. The edging cells denote the access router (AR) that the cell belongs to have a boundary with another router. Only the edging cells of a router need considering IP layer HO. Trigger distance D trigger from rim of the edging cell can be calculated by Advanced Engineering Forum Vols. 6-7 Where α is the time setforward, α=1 means double of the HO time, while α=-1 means no time setforward. α is an empiric value decided by experiments in order to insure the trigger mechanism, so that α greatly affects the quality of HO; T HO is HO time; V MN , the velocity of MN, can be acquired by wireless positioning with GPS info or sensor network and E-map. The time sequence of simplifying authentication mechanism is shown as following. So the HO authentication processes can be shortened to six steps for the dividing handover sequencing into several paralleled parts and budges some steps between stations. Thereby cut processes on each node.

Simulation
Simulation is conducted to verify availability of the above framework with modified NS2 platform. NS2, replacing network simulator version 2, can be found at http://www.isi.edu/nsnam/ns/.
The platform was composed of 256 50km-diameter cells C i (i=0-255). It is assumed that there are j MS j (j=0-9999) and their initial location (C j0 =C i ) and their constant speed (v j ) is randomly generated, with speed v j from 1km/h to 120km/h. The next location (C j t ), which decides movement direction, is randomly generated per seconds. Initiation of a call between two MS is also randomly generated on when (t p ), who (MS p1 ) to who (MS p2 ), how long (conversation time t p , obeys negative exponent distribution, average value 1/µ=200s) and how large (bandwidth requirement B p ). With time setforward α increases from -1 (means no any time setforward, it is a common Fast HO scheme) to 1, the HO delay is reduced. When α valued -1 to 0 the delay decreases evenly and the resource engross increases evenly. From α valued from 0 to 1, the delay not obviously increases, but the resource bandwidth engross augment greatly therefore useable bandwidth is greatly reduced. One maybe explanation is that with the setforward increase, system got enough time to prepare, but the next-AR-predict become vaguer, so an aggregate of locations augment, then MSPEC of resource reservation is more. With α increases (from -1 to 1, step 0.2), average MSPEC number increases too. There exists the best compromise result in α=0.2, with acceptable resource engross and a little HO delay.

816
Information Technology for Manufacturing Systems III  Figure 2. The effect of time setforward α on delay Fig.2 show that the effect of time setforward α on the average of the delays in Fast HO+MRSVP [9](called FHO+RSVP) and our scheme (ABI+MRSVP) when α=-0.2,0.2,0.6 and call volume of business (erl) shifts from 6 to 18.
Obviously, delays in ABI+MRSVP are smaller than FHO+RSVP no matter α is -0.2,0.2 or 0.6. Since best result exists α=0.2 from Table 2, the ABI+MRSVP average delays when α=0.2 is adopted to compare with FHO+RSVP, we have the reduced quantity of 19.12%.   Fig.3 show that the effect of α on average successful probability of resource reservation. When α=-1, no time setforward thereby no overhead. When α=-0.8, ABI+MRSVP performs worse than FHO+RSVP due to time setforward get too rare to prepare the ABI, much less the simplifying authentication mechanism. Howbeit ABI signaling overhead exists, so resource reservation possibilities are a little small than the FHO+RSVP. Only when α>-.6, the scheme is meaningful. When α=0.2, we can see, a large resource reservation probability with little overhead can be gotten. In the situation, and the resource possession of RSVP is mitigated by 30.13% comparing with FHO+RSVP.

Summary
This paper proposed the trust transfer mechanism for mobile Internet access with ABI scheme. With pre-anticipated trust transfer for dominant minority correspondent nodes and the time sequence of simplifying authentication mechanism, ABI can reduce delay and the probability of packets loss during both call initialization and HO processes, it can eliminate HO delay and minimizes resource possession during HO, enhance QoS in MRVSP and smooth HO procedure. ABI supports heterogeneous environments as well as a homogeneous network. Nevertheless, it will perform better in a network with QoS service. By modifying the details of this ABI scheme, it also can be used in other sceneries.