1 Some Issues in Ad Hoc Networks Nitin Vaidya University of Illinois at Urbana-Champaign Keynote talk presented at the International Workshop on Theoretical Aspects of Wireless Ad Hoc, Sensor, and Peer-to-Peer Networks Illinois Institute of Technology, Chicago, June 11, 2004 © 2004 Nitin Vaidya

2 Outline  Preliminaries  Advertising  Preaching

3 Ad Hoc Networks  Formed by wireless hosts which may be mobile  Without necessarily using a pre-existing infrastructure Hybrid architectures using infrastructure likely in many applications

4 Why Ad Hoc Networks ?  Potential ease of deployment  Decreased dependence on infrastructure

5 Many Potential Applications  Personal area networking  cell phone, laptop, ear phone, wrist watch  Military environments  soldiers, tanks, planes  Civilian environments  taxi cab network  meeting rooms  sports stadiums  boats, small aircraft  Emergency operations  search-and-rescue  policing and fire fighting

6 Challenges (Opportunities)  Broadcast nature of the wireless medium  Limited wireless transmission range –Hidden terminal problem  Packet losses due to transmission errors  Mobility-induced route changes  Mobility-induced packet losses  Battery constraints  Potentially frequent network partitions  Ease of snooping on wireless transmissions (security hazard)

7 State of the Art  Lot of research activity on:  Routing  Medium access control  Quality of service

8 State of the Art  More recently …  Capacity of wireless networks << Information theory community –Pure wireless networks –Hybrid networks –Delay-throughput trade-off  Graph-theoretic problems<< Algorithms/theory community –Topology control –Dominating sets –Connectivity problems –Coverage problems in sensor networks

9 State of the Art  Many more (academic) problems … rich area  (Too) Many conferences  MobiHoc  SenSys  MASS  SECON  …

10 What’s Lacking?  Real applications still lacking (beyond military)  Hard to evaluate protocols in a vacuum  But there is hope … applications on the horizon  Community networks starting to use ad hoc routing  Vehicular networks  Sensor networks

11 What’s Lacking  Primitives to build distributed applications  Much work on distributed algorithms on fixed and dynamic networks wherein dynamism comes from “random” link failures  But little on ad hoc networks, where the dynamism comes from node mobility and channel variations  Need to revisit distributed computing problems in the new context

12 Outline  Preliminaries  Advertising  Preaching

13 Our Research Themes Exploiting physical layer capabilities  Protocols for directional antennas  Rate adaptation  Power control & Power save mechanisms  Multi-channel mechanisms

14 Our Research Themes Distributed algorithms for ad hoc networks  Address assignment  Mutual exclusion  Leader election  Token circulation

15 Our Research Themes Misbehavior in Wireless Networks  Protocol design for misbehavior detection

16 Some of our past research …  Weak duplicate address detection  Misbehavior detection  Mutual exclusion

17 Weak Duplicate Address Detection

18 Address Assignment  Dynamic auto-configuration important for autonomous operation of an ad hoc network  Goal: Assign each node a unique address OR Assign each address to at most one node  Can be viewed as distributed mutual exclusion with an address being a resource

19 Auto-Configuration in Ad Hoc Networks  Worst case network delays may be unknown, or highly variable, or unbounded  Partitions may occur, and merge

20 Duplicate Address Detection in Ad Hoc Networks  Several proposals  One example [Perkins]:  Host picks an address randomly  Host performs route discovery for the chosen address  If a route reply is received, address duplication is detected

21 Example: Initially Partitioned Network D’s packets for address a routed to A

22 Merged Network  Duplicate address detection (DAD) important to avoid misrouting

23 Strong DAD  Detect duplicate addresses within t seconds  Not possible to guarantee strong DAD in presence of unbounded delays  May occur due to partitions  Even when delays are bounded, bound may be difficult to calculate Unknown network size

24 DAD  Strong DAD impossible with unbounded delay  How to achieve DAD ?

25 Design Principle If you cannot solve a problem Change the problem

26 Weak DAD: Requirement Packets from a given host to a given address should be routed to the same destination, despite duplication of the address

27 Example: Initially Partitioned Network D’s packets for address a routed to A

28 Merged Network: Acceptable Behavior with Weak DAD Packets from D to address a still routed to host A

29 Merged Network: Unacceptable behavior Packets from D to address a routed to host K instead of A

30 Weak DAD: Implementation  Integrate duplicate address detection with route maintenance

31 Weak DAD with Link State Routing  Each host has a unique (with high probability) key  May include MAC address, serial number, …  May be large in size  In all routing-related packets (link state updates) IP addresses tagged by keys  (IP, key) pair

32 Weak DAD with Link State Routing  Address duplication not always detected  Duplication detected before misrouting can occur  Weak DAD  Reliable, but potentially delayed

33 Link State Routing (LSR): Example

34 Weak DAD with LSR

35 Weak DAD with LSR Host X with key K_x joins and choose IP_A (address duplication) X

36 Weak DAD with LSR If host D receives a link state update containing (IP_A, K_x), host D detects duplication of address IP_A Two pairs with identical IP address but distinct keys imply duplication

37 Just-in-Time DAD  Duplication detected before routing tables could be mis-configured

38 Moral of the Story  Traditionally, address assignment and routing are independent algorithms  Duplicate address detection integrated with route maintenance can provide stronger properties

39 Misbehavior Handling Joint work with Pradeep Kyasanur

40 Problem Definition Wireless channel Access Point AB Nodes are required to follow Medium Access Control (MAC) rules Nodes can benefit by misbehaving AB

41 IEEE overview  Distributed Coordination Function (DCF)  Widely used for channel access  DCF is a Carrier Sense Multiple Access/ Collision Avoidance (CSMA/CA) protocol

42 CSMA/CA  Don’t transmit when channel is busy  Defer transmission for a random duration on idle channel

43 Backoff Example  Choose backoff value B in range [0,CW]  CW is the Contention Window  Count down backoff by 1 every idle slot wait Transmit wait B2=10 B1=20 B2=10 B1=0 S1 S2 CW=31 B1=15 B2=25

44 Possible Misbehavior  Backoff from biased distribution  Example: Always select a small backoff value Transmit wait B1 = 1 B2 = 20 Transmit wait B2 = 19 B1 = 1 Misbehaving node Well-behaved node

45 Potential Solutions  Prevent misbehavior  Detect misbehavior  Penalize misbehavior

46 Game Theoretic Solutions [MacKenzie]  Assumes there is some cost for transmitting  Nodes independently adjust access probability  Under some assumptions, network reaches a fair equilibrium  Game theoretic solutions to the misbehavior problem so far assume complete knowledge of the channel (difficult to have in multi-hop networks)  Not yet clear whether partial information is adequate

47 Charging  Charge for transmitted packets  Transmitting more packets costs more  Disadvantages  Per-packet charging can still allow misbehavior that decreases the user’s delay  Need to implement charging mechanism

48 Goals of proposed scheme  Detect misbehavior  Penalize misbehavior

49 Detecting Misbehavior  Observe each node  If a node does not wait long enough before transmitting, then conclude that it is misbehaving  Penalize the misbehaving node

50 Issues  Idle duration is a function of backoff interval chosen by a node  Observer does not know exact backoff value chosen by a sender  Sender chooses random backoff  Hard to distinguish between maliciously chosen small values and a legitimate random sequence  Wireless channel introduces uncertainties  Channel status seen by sender and receiver may be different

51  Observe backoffs chosen by a sender over multiple packets  Backoff values not from expected distribution  Misbehavior  Longer delay in detection, since the distribution of non-deterministic backoff must be determined Potential Solution: Use long-term statistics

52 A Simpler Approach  Remove the non-determinism

53 A Simpler Approach  Receiver provides backoff values to sender  Modification does not significantly change behavior

54 Modifications to R provides backoff B to S DATA Sender S Receiver R CTS ACK(B) RTS S uses B for backoff for next packet RTS B

55 Detecting deviations  Receiver counts number of idle slots B obsr Condition for detecting deviations: B obsr <  B  ≤ 1 Sender S Receiver R ACK(B) RTS Backoff B obsr

56 Misbehavior Detection IF  The detection would always detect misbehavior IF all nodes observe identical channel status at all times  But all nodes do not see same channel status  Hidden terminals  Fading  In general, cannot diagnose misbehavior with 100% accuracy

57 Penalizing Misbehavior ACK(B+P) CTS DATA B obsr Sender S Receiver R ACK(B) RTS Actual backoff < B When misbehavior is suspected, larger backoff intervals are assigned  penalty mechanism

58 Penalty Scheme  Misbehaving sender has two options  Ignore assigned penalty  Easier to detect  Follow assigned penalty  No throughput gain  With penalty, sender has to misbehave more for same throughput gain

59 Diagnosing Misbehavior  If misbehavior suspected for “long enough” duration, conclude that the misbehavior is intentional  Higher layers / administrator can be informed of misbehavior

60 Multiple Observers  Currently, single observer is used (receiver)  Data from multiple observers can be combined to improve diagnosis S B AR S sends a packet to R A, B also monitor S Information from A, B, R may be combined

61 Moral of the Story  MAC layer misbehavior can severely affect throughput of well-behaved nodes  Improving predictability improves ability to detect misbehavior  Open issues:  Using multiple observers  Integrating diagnosis with higher layers

62 Distributed Mutual Exclusion Joint work with Jennifer Welch and Jennifer Walter

63 Approach 1: implement existing distributed primitives on top of existing ad hoc routing protocols. User Application Distributed Primitive Routing Protocol Ad-Hoc Network Approach 2: modify distributed primitives to be aware of information from lower layers User Application Distrib. Primitive Routing Protocol Ad-Hoc Network Why Design New Algorithms for MANETs?

64  Token-based: Only the node possessing the token may enter critical section  Nodes must have a way of sending requests to the token holder  One solution: Mutual exclusion for fixed topology + Routing on ad hoc networks Distributed Mutual Exclusion

65 Link Reversal Algorithm [Gafni81] (Routing Protocol) AFB CEG D

66 Link Reversal Algorithm [Gafni81] AFB CEG D Maintain a directed acyclic graph (DAG) for each destination, with the destination being the only sink This DAG is for destination node D Links are bi-directional But algorithm imposes logical directions on them

67 Link Reversal Algorithm Link (G,D) broke AFB CEG D Any node, other than the destination, that has no outgoing links reverses all its incoming links. Node G has no outgoing links

68 Link Reversal Algorithm AFB CEG D Now nodes E and F have no outgoing links Represents a link that was reversed recently

69 Link Reversal Algorithm AFB CEG D Now nodes B and G have no outgoing links Represents a link that was reversed recently

70 Link Reversal Algorithm AFB CEG D Now nodes A and F have no outgoing links Represents a link that was reversed recently

71 Link Reversal Algorithm AFB CEG D Now all nodes (other than destination D) have an outgoing link Represents a link that was reversed recently

72 Link Reversal Algorithm AFB CEG D DAG has been restored with only the destination as a sink

73 Link Reversal Algorithm  Goal: Maintain DAG pointing to the “destination” despite topology changes

74 E F D A B C  Static topology  Spanning tree with edges directed toward the token holder Mutual Exclusion in Static Networks [Raymond89]

75 A B C E D E D F A B C E D E F

76 Raymond’s Algorithm on Ad Hoc Networks  The algorithm can be implemented on top of routing protocol –Routing algorithms provides abstraction of a fully connected network  Maintain a spanning tree using logical links in the “fully connected” network  “Adjacent” nodes in the spanning tree may be far from each other  Potentially poor performance

77 Mutual Exclusion in Ad Hoc Networks  Gafni  Variable topology, fixed sink  Raymond  Fixed topology, moving sink  Proposed algorithm: Mutual exclusion in ad hoc networks  Variable topology, moving sink

78 Moral of the Story  Existing algorithms not always appropriate  Algorithms for dynamic networks can be applied to ad hoc networks, but performance may be poor  Taking into consideration lower layer information can help

79 On to the preaching …

80 Abstractions  Of necessity, algorithm designers work with abstractions  Physical layer is messy  Abstractions hide “unnecessary” physical layer details

81 Abstractions  But some details are important. Many common mistakes.  I am guilty too … but hopefully learning from the mistakes

82 Transmission “Range”  Transmission range R R

83 Transmission “Range”  Given the thermal noise, beyond a certain distance reliable communication infeasible at a desired rate  Converse often assumed true: Within transmission range, reliable communication is assumed always feasible  This assumption is not accurate Reliability depends on SINR  Assumption may perhaps be OK for order statistics, but the constants matter in practice

84 Interference “Range”  Interference “range” assumed to be the distance over which a transmission “collides” with another transmission  Assumed that if a host transmits, no other transmission within interference range will succeed  Not accurate: Reliability depends on SINR

85 Interference “Range” CFABED DATA Interference “range” Whether A’s interference results in unreliable reception at D depends on SINR at D

86 Graceful Degradation  Transmission “range” (or reliability) depends on SINR and bit rate  Even if transmission at a higher rate fails, low rate transmission may be feasible Distance Throughput Modulation schemes provide a trade-off between throughput and “range”

87 Energy Consumption  Common assumption: Energy required to transmit on a hop = k d  k and θ typically assumed to be constants  Proofs relying on constant k, θ may break when they are not constants θ

88 Energy Consumption  When k,θ = constant, links AC and BD cannot BOTH be on energy efficient routes (considering only transmit energy)  With constant k,θ, energy efficient routes do not need to intersect [Narayanaswamy02] A B C D

89 Energy Consumption  Consider routes A  C and B  D  With fixed k and fixed θ > 2, energy optimal routes are A-B-C and B-C-D (direct links A-C and B-D are not optimal)  Energy-efficient routes do not intersect A BC D

90 Energy Consumption  Let k be much smaller on diagonal links (alternatively, θ ≈ 2 on diagonal links, and 3 on other links)  Diagonal links cheaper than other routes  Energy efficient routes must intersect A BC D

91 Geographic Location  Many algorithms rely on knowledge of physical location  Location estimates in practice contain some error  The error can affect correctness of geographic routing [Saeda04]

92 Summary  Physical layer characteristics matter  Can affect algorithm performance and correctness

93 End of preaching …

94 Interesting Open Problems  Protocols that achieve “capacity”  Distributed algorithms for ad hoc networks  Shared memory  Message ordering  Group communication  …  Complexity as a function of mobility  Applications for ad hoc networks

95 Thanks!

96 Thanks!

97 Handling other misbehavior  Receiver may misbehave by assigning large or small backoff values  Sender can detect receiver assigning small backoff values  Backoff assigned by receiver has to follow well-known distribution  Sender uses larger of assigned backoff and expected backoff

98 Handling other misbehavior  Detecting receiver assigning large backoff values not handled  Equivalent to receiver not responding at all  Need higher layer mechanisms  Collusion between sender and receiver  Harder to detect  Requires an observer that can monitor both sender and receiver