cs0605088/bus.tex
1: 
2: \section{TARMAC: Basic Scheme}\label{bus}
3: 
4: \rednote{Good start.  Recommend:
5: -comparative analysis to the existing schemes
6: -identification of the tradeoffs present in the design (size of
7: packet, time-period, etc..).  Present the discussion of the drawbacks
8: in a way that leads into the next section (the improvements), such
9: that the next section can start with something like: ``The basic scheme
10: has two major drawbacks, inability to adapt to the level of traffic
11: and excessively high overhead.  In this section, we address these
12: drawbacks by...''}
13: 
14: \rednote{To make description easier and more accurate, maybe we should
15: use packet for data, and frame for the buses}
16: 
17: In this section, we present the basic TARMAC scheme.  We also discuss
18: tradeoffs and expected behavior of this base model.  The drawbacks of
19: the model in terms of energy efficiency and increased contention pave the way for the suggested improvements in the next Section.
20: 
21: \subsection{Basic TARMAC Scheme}
22: 
23: TARMAC is a MAC layer protocol that takes in to consideration the
24: makes traffic resistant to analysis by using uniform (but possibly
25: time-shifted) transmission schedules at all nodes. TARMAC emulates the
26: communication network alternative of a city bus system.  TARMAC uses
27: encrypted frames that are sent at pre-scheduled times (it can
28: inter operate with different link/MAC layers), irrespective of the
29: presence of data packets.  Different outgoing data packets, possibly
30: targeted to different destinations and next hops, are placed in the
31: same TARMAC frame; any empty slots are left empty.  Note that empty
32: slots are hidden by the encryption and the occupancy of each frame is
33: not visible to the attacker.  Each receiver of the frame can examine
34: it and decide which packets are its responsibility to forward:
35: virtually any routing protocol can be used to establish routing
36: responsibilities.
37: 
38:   TARMAC frames resemble public buses in that they have a fixed number
39: of slots and leave at pre-scheduled times, with passengers
40: (data-packets) taking up seats in the bus.  In typical public
41: transportation systems, provisioning is not uniform: for example,
42: stops at remote routes have much lower bus activity and perhaps
43: smaller buses than a central hub where many routes converge.  Thus,
44: extracting the structure of the public transportation network by
45: simply observing the level of activity at a stop becomes possible.
46: 
47: To protect against such analysis TARMAC requires that all nodes follow
48: the same schedule (later, we consider the possibility of a
49: time-shift).  {\em The basic TARMAC scheme we discuss in this section
50: is one where all the nodes transmit with a fixed period and fixed size
51: packets}; later we explore relaxing this model.  Having the
52: transmission times and sizes be uncorrelated to the data transport
53: being carried through TARMAC makes it impossible to detect the
54: presence of, for example, event based data.  Furthermore, the
55: broadcast nature of the wireless medium hides information about the
56: receiver of the packet; thus, a TARMAC frame is like a number of
57: concurrent buses that leave to each of the one-hop neighbors.
58: 
59: In contrast to these advantages of TARMAC, conventional MAC transmit
60: data only when there is data to transmit, providing valuable
61: information to attackers.  Further, a data packet can be followed as
62: it is retransmitted by intermediate hops to extract the full
63: connection and the eventual destination.
64: 
65: \subsection{Basic Parameters and Provisioning}
66: 
67: In the basic model with periodic equal size frame, the relevant
68: parameters are the transmission period $\tau$ and the size of the
69: frame in slots $s$.  The capacity, measured in slots per node per
70: second, for the basic TARMAC can theoretically be expressed as
71: $\frac{s}{\tau}$.  Within the physical limitations of the channel,
72: increasing either $s$ or reducing $\tau$ leads to increasing capacity
73: by either sending larger packets or sending packets more often
74: respectively.  However, sending smaller frames more frequently fosters
75: shorter delays, but increases frame overhead.  In addition, larger
76: size frames are more vulnerable to collisions and transmission losses.
77: 
78: A tension between the capacity of the network and the energy
79: efficiency of TARMAC arises.  At one extreme, all nodes may be made to
80: appear like a remote station, leading to energy efficiency but loss of
81: capacity since the bottleneck nodes now do not have sufficient
82: capacity to carry the offered load.  On the other extreme, the network
83: may be provisioned so that all nodes are transmitting at a sufficient
84: rate to enable the bottleneck nodes to continue to forward the
85: traffic.  This leads to excessive overhead in remote or idle areas.
86: These and other drawbacks are discussed in the next subsection.
87: 
88: \subsection{Drawbacks of Basic TARMAC}
89: 
90: To be able to carry the required traffic, the basic TARMAC must be
91: provisioned sufficiently such that $\frac{s}{\tau}$ is greater or
92: equal the required traffic at bottleneck nodes.  However, provisioning
93: for the worst case has the following drawbacks:
94: \begin{itemize}
95: \item {\em Over or under-provisioning:} it is difficult to predict
96: what the maximum bottleneck capacity is for some networks.  The
97: maximum reporting rate may be difficult to predict or the deployment
98: may be ad hoc making a-priori analysis of bottleneck nodes difficult.
99: As a result, the choice of $s$ and $\tau$ may lead to insufficient
100: capacity to carry the reported data and leading to increased delay and
101: loss of data as buffer sizes grow.  Alternatively, it may lead to more
102: aggressive sending and loss of efficiency as most slots remain idle.
103: 
104: \item {\em Possible High Energy Cost:} By provisioning to the rate of
105: the expected bottleneck, most of the nodes will be transmitting at a
106: rate higher than that needed to carry their traffic.  We note here
107: that as long as an average occupancy of more than one slot per frame
108: the total number of transmissions will be reduced.  However, the size
109: of each frame will be likely be bigger than the size of the individual
110: data packets due to over-provisioning; some savings in framing and
111: protocol overhead may result from combining multiple packets into a
112: single transmission.
113: 
114: 
115: \item {\em Low maximum throughput:} As the node capacity is increased
116: by reducing the period or increasing the size, all nodes start sending
117: more aggressively and the contention level increases.  This includes
118: nodes that do not have high occupancy, limiting the maximum throughput
119: that can be achieved by nodes that do have data to send.
120: 
121: \item {\em Increased delay:} Since each node is not forwarding the
122: packets as soon as it gets it, there may be a larger packet delivery
123: latency as packets wait for the next frame transmission at every
124: intermediate hop.
125: \end{itemize}
126: 
127: \noindent
128: In the next section, we identify a number of improvements to the base
129: TARMAC that address some of these drawbacks.
130: 
131: 
132: 
133: 
134: 
135: 
136: 
137: 
138: 
139: 
140: 
141: 
142: 
143: 
144: 
145: 
146: 
147: 
148: 
149: 
150: 
151: 
152: 
153: 
154: 
155: 
156: 
157: 
158: 
159: 
160: 
161: 
162: 
163: 
164: 
165: 
166: 
167: 
168: 
169: 
170: 
171: 
172: 
173: 
174: 
175: 
176: 
177: 
178: 
179: 
180: 
181: 
182: 
183: 
184: 
185: 
186: 
187: 
188: 
189: 
190: 
191: 
192: 
193: 
194: 
195: 
196: 
197: 
198: 
199: 
200: 
201: 
202: 
203: 
204: 
205: 
206: 
207: 
208: 
209: 
210: 
211: 
212: 
213: 
214: 
215: 
216: 
217: 
218: 
219: 
220: 
221: 
222: 
223: 
224: 
225: 
226: 
227: 
228: 
229: 
230: 
231: 
232: 
233: 
234: 
235: