Download NEA One of the first questions when designing a neural net architecture is how many layer and more Exams Computer Networks in PDF only on Docsity!
What is a denial-of-service attack : The goal of a denial-of-service attack is to stop a network from functioning or to disrupt the services a network provides. Jamming attack : A jamming attack is a DoS attack at the physical layer, where an adversary interferes with the radio communications of sensor nodes by continuously transmitting a strong signal that lowers the signal-to-noise ratios on the sensor nodes. Exhaustion attack: The goal of an exhaustion attack is to prematurely deplete a sensor node of its energy resources, e.g. by increasing the overheads and workloads of a sensor node. For example, an adversary could exploit the error recovery mechanisms of a sensor node by continuously interfering with transmissions, thereby triggering retransmissions of entire sensor messages. Tampering attack: A tampering attack occurs when an adversary obtains physical access to a sensor node, allowing the attacker to destroy or modify the device. Consider routing attacks such as selective forwarding, sinkhole, blackhole, Sybil, rushing, and wormhole attacks. Describe briefly each type of attack and discuss how these attacks could take place in the following types of networks: An adversary positioned on the route between a sender and receiver could drop packets that meet certain characteristics (selective forwarding attack) or even all packets (blackhole attack). A sinkhole attack occurs when an adversary attempts to position itself on the route of as many sensors streams as possible, thereby drawing traffic towards the adversary, who can then disrupt the traffic or modify packet content. A Sybil attack occurs when an attacker claims to have several identities (or locations) thereby convincing more sensor nodes to choose the attacking node as forwarding node. In a rushing attack, an adversary quickly forwards route request messages towards the destination to increase the attacker’s probability to be on the sensor stream’s route. Finally, in a wormhole attack, two collaborating attackers attempt to convince other nodes that they have a good link between them such that they will be chosen as forwarder nodes for as many routes as possible. (a) A network using a table-based routing protocol such as OLSR. In table-based protocols, the goal of an attacker is to appear as next-hop neighbor in as many tables as possible. Therefore, the attacker can pretend to have low forwarding costs to all destinations. Once the attacker receives packets to forward, it can modify or drop them as desired. (b) A network using an on-demand routing protocol such as DSR. In an on-demand routing protocol, the goal of the attacker is to be on the path chosen by either the receiver or sender device. Therefore, the attacker could either rush RREQ packets towards the destination or it could modify the content in the RREQ packet headers such that the receiver believes that the route with the attacker is the one with the lowest cost. (c) A network using a location-based routing protocol such as GEAR. In location-based routing protocols, the goal of the attacker is to make its neighbors believe that it is closer to the destination than all other neighbors or that its cost for forwarding towards the destination is lower that those of all the other neighbors. For example, this can be achieved by pretending to have multiple identities and locations or by changing the location information sent to its neighbors. For example, if an attacker node sees that one of its neighbors frequently sends packets towards a destination, but not via the attacker node, the attacker can “change” its location such that it becomes the better forwarding choice towards the destination. What are the security models provided by IEEE 802.15.4. What is the purpose of the trust center in ZigBee? IEEE 802.15.4 has four basic security models: access control, message integrity,
confidentiality, and replay protection. The trust center in ZigBee (a responsibility typically assumed by the ZigBee coordinator) is responsible for the authentication of devices wishing to join a network, maintaining and distributing keys, and enabling end-to-end security between devices. What is “data freshness” and why is it important in sensor networks? Data freshness ensures that sensor data is recent and no old recording of such data are being replayed. For example, this is important for key distribution schemes to prevent later replays of key distributions. While “typical” computers are in homes, offices, labs, etc., wireless sensor nodes are often placed in places that are publicly open and accessible. What kind of attacks could an adversary initiate by accessing a single sensor node in a large-scale WSN? A variety of attacks are possible once an attacker gains access to a device. It can be used to learn about encryption algorithms and keys used in the network. It could be used as a starting point for a variety of denial-of-service attacks and man-in-the- middle attacks. Attacks on the routing layer (e.g. selective forwarding attacks, etc.) or on data aggregation are also possible. Explain some of the characteristics of a WSN that make routing security difficult to implement. # The resource constraints of wireless sensor networks make it easier to use attacks with the goal to exhaust a device’s or network’s resources. It is also more difficult to use resource-intensive security measures (e.g. CPU-intensive algorithms). The lack of centralized control in a WSN means that many security measures must also be decentralized. The remote location of sensor nodes makes it difficult to protect them from physical access by an attacker. Communication in a WSN is error-prone, which may interfere with security-related communications.# Why do you think authentication can be a particularly significant problem in a WSN? Many sensor nodes are placed in publicly accessible areas, i.e. it is often easy to gain physical access to sensor devices. An attacker may then be able to modify or replace the device or analyze the device for its content (e.g. to learn about security keys and algorithms). It is often necessary to share cryptographic techniques and keys with all nodes in the sensor network, making it easier for an intruder to obtain such information. Authentication is particularly important since it is difficult to distinguish between a new node that joins a network or a a node that was temporarily disconnected from the network from a sensor node that has been inserted into the network or compromised by an intruder. A sensor node in a WSN using the lighthouse approach for localization detects the first beam of light at time 0 s and the second beam of light at time 0.25 s. The next time the first beam of light is detected is 7 s. The distance of the two light sources (beam width) is 10 cm. What is the distance of the sensor to the light emitter? # The distance is computed as: d = b 2sin(α/2) where α is computed as: α = 2π tbeam tturn Since tbeam = t1 − t2 = 0.25 − 0 = 0.25 and tturn = 7 s, α = 0.224 and d = 25.5 m. # For the APIT test, can you show a concrete scenario where a node M would come to the wrong conclusion that it must be inside a triangle? Use a scenario where node M has at least 3 neighbors. Can you also show an example where node M would come to the wrong conclusion that it must be outside a triangle? Explain the difference between range-based and range-free localization? Range-based localization techniques are based on distance measurements using ranging techniques (RSS, ToA, etc.) between nodes, e.g. between sensor nodes with unknown locations and anchor nodes with known locations. Range-free localization techniques do not use distance measurements based on ranging techniques. The advantage of range free techniques is that they typically do not require extra hardware and are therefore more cost effective.
refraction, scattering, etc. RF profiling allows localization techniques to take such effects into consideration. Objects include trees, buildings, traffic signs, mountains, etc. What is the main disadvantage for both TDoA and AoA ranging techniques? Both techniques requires extra hardware (e.g. two radios for TDoA and an array of antennas for AoA), which may be infeasible for many low-cost and low-power sensor networks. Time of Arrival (ToA) is one example of a ranging technique. Answer the following questions (assume a propagation time of 300 m/s): (a) What is the advantage of two-way ToA over one-way ToA? The one-way ToA requires that sender and receiver are accurately synchronized. In the two-way ToA, the round-trip time is measured using the clock on only on device, thereby removing the need for clock synchronization. (b) In a synchronized network with unknown synchronization error, an anchor node periodically broadcasts an acoustic signal to sensor nodes in its range. At time 1000 ms on the anchor node’s clock, the anchor node issues a beacon, which is received by node A at time 2000 ms (on node A’s clock). What is the distance that A can now compute? The propagation delay of the acoustic signal is 1000 ms and the signal travels at 300 m/s, i.e. the distance is 300 m. (c) Instead of computing the distance itself, node A also responds with an acoustic signal issued at time 2500 ms, which is received by the anchor node at time 3300 ms. What is the distance computed by the anchor node? What can you say about the synchronization of anchor node and node A? The round-trip time is (2000 − 1000) + (3300 − 2500) = 1000 + 800 = 1800. The distance is then the round- trip time divided by 2 and multiplied by the velocity, i.e. 270 m. Since the propagation time of the signal from anchor to sensor is 1000 ms and the propagation time from sensor to anchor is 800 ms, the offset between the sensor node’s clock and the anchor’s clock appears to be 100 ms. Define the terms anchor-based localization and range-based localization. Anchor-based localization refers to the use of reference nodes (anchors) with well known locations. Communication between sensor nodes and these anchors can be use to estimate the locations of the sensors. Range-based localization estimate locations based on range measurements, i.e. measurements of the distances between nodes. A node’s position in two-dimensional space is (x, y) = (10, 20) with a maximum error of 2 in the x direction for 95% of all measurements and a maximum error of 3 in the y direction for 90% of all measurements. What is the accuracy and the precision of this location information? The accuracy of the location information is 2 in the x direction and 3 in the y direction. The precision is 95% in the x direction and 90% in the y direction. Why is localization needed in wireless sensor networks? Name at least two concrete scenarios or applications where localization is required. Sensor nodes are often deployed in an ad-hoc fashion, without knowledge of their exact deployment position. Localization is then necessary to provide a physical context to sensor readings, i.e. most sensor readings are meaningless without knowledge of where the readings were obtained. For example, localization is required in sensor network that detects wild fires to ensure that firefighters can quickly locate the affected areas and to predict the spread of the wild fire. Similarly, in surveillance applications, location information is required to be able to guide law enforcement officers quickly towards an intruder. Why is time synchronization needed in a WSN? Name at least three concrete examples. Time synchronization is needed to achieve accurate temporal correlation of observed events, e.g. to detect the speed of a moving object. Time synchronization is also needed for many MAC-layer protocols, e.g.
to accurately determine when a node can transmit in a TDMA-based system. Finally, many networks utilize duty cycling, i.e. devices are awake for brief periods of time and sleep otherwise. Accurate timing information is needed to determine when a node has to turn its radio back on. Explain the difference between external and internal time synchronization and name at least one concrete example for each type of synchronization. External synchronization means that the clocks in a network are synchronized with an external source of time (reference clock). For example, GPS can be used to achieve external synchronization. Internal synchronization refers to the process of synchronizing the clocks of sensor nodes in a network with each other, without the support of an external reference clock. To achieve internal synchronization, sensor nodes could elect a master node which periodically broadcasts its time to all other nodes in the network, allowing them to reset their clocks to correct for any drifts. Consider two nodes, where the current time at node A is 1100 and the current time at node B is 1000. Node A’s clock progresses by 1.01 time units once every 1 s and node B’s clock progresses by 0.99 time units once every 1 s. Explain the terms clock offset, clock rate, and clock skew using this concrete example. Are these clocks fast or slow and why? Comparing the two clocks, the clock offset is the difference in time between the two clocks. In this example, the current clock offset is 100. The clock rate indicates the frequency at which a clock progresses, i.e. node A’s clock has a clock rate of 1. and node B’s clock has a clock rate of 0.99. The clock skew indicates the difference in the frequencies of the two clocks, which is 0.02. Clock A is fast since its clock readings progress faster than real time. Similarly, clock B is slow since its clock readings progress slower than real time. Assume that two nodes have a maximum drift rate from the real time of 100 ppm each. Your goal is to synchronize their clocks such that their relative offset does not exceed 1 s. What is the necessary resynchronization interval? Each clock can deviate from real time by 100 μs per second in the worst case, i.e. it takes up to 10 000 s to reach an offset of 1 s. However, since both clocks have a drift rate of 100 ppm, the relative offset between them can be twice as large as the offset between a single clock and the external reference clock. Therefore, the necessary resynchronization interval is 5 000 s. You need to design a wireless sensor node and you have three choices for clocks (1..3) with maximum drift rates of ρ1 = 1 ppm, ρ2 = 10 ppm, and ρ3 = 100 ppm. Clock 1 costs significantly more than clock 2, which in turn costs significantly more than clock 3. Explain when one would choose clock 1 instead of clock 2 or clock 3 and vice versa. # The smaller the maximum drift rate, the less frequent re- synchronizations do have to occur. Re-synchronizations can incur large communication overheads, therefore more accurate clocks are preferred. However, cost is also an important factor for many wireless sensor networks. Therefore, in networks where cost is the primary factor, a cheaper and less accurate clock may be chosen. In networks where communication overhead due to synchronizations is a bigger concern than cost, more accurate clocks should be chosen. A network of five nodes is synchronized to an external reference time with maximum errors of 1, 3, 4, 1, and 2 time units, respectively. What is the maximum precision that can be obtained in this network? Since the error can go either way, i.e. a clock can be faster or slower than the external reference time by the amount of the error, the maximum precision is then the sum of the two largest errors, i.e. 3+4=7. Node A sends a synchronization request to node B at 3150 (on node A’s clock). At 3250, node A receives the reply from node B with a timestamp of 3120.
synchronization source. Which node detected the event sooner and by how much? Assume a signal speed of 300 m/s. When the acoustic signal arrived at node A at time 10 (on A’s clock), the signal had to travel 300 m more to reach node B, which takes 1 s. That is, When node A’s clock reading was 10 s, node B’s clock reading was 15-1 = 14 s. That is, node B’s clock is 4 s faster than node A’s clock. Node A observed the event at 15 s, while node B observed the time at 19.5 s, which corresponds to 15.5 s on A’s clock. That is, node A observed the event first by 0.5 s.
(a) synchronization messages experience send and access delays with high variance and all other
delays are negligible RBS is a better choice since the only the time off arrival is important and everything before is irrelevant, i.e. send and access delays do not affect it (unlike TPSN). (b) synchronization messages are sent using acoustic signals and the distances between nodes are unknown When propagation delays are large (such as in acoustic communications), they have an important effect on the quality of synchronization. When two nodes have different distances to the sender, nodes using RBS will experience different propagation delays, which introduce an error into the synchronization. Protocols based on round-trip time measurements (such as TPSN) are more suitable in this case since they consider propagation delays in their synchronization process. (c) synchronization messages experience send and access delays without variance and all other delays are negligible RBS does not depend on send and access delays and TPSN works well when these delays do not vary, therefore both protocols will work well. (d) synchronization messages experience significant receive delays that may differ from node to node Both TPSN and RBS will be affected by the variation in receive delays and therefore both will be good choices. In AODV, is it possible that route discovery packets travel in the network forever? Why or why not? Each node is identified by its source and a sequence number (broadcast ID), allowing forwarding nodes to discard duplicate packets. Further, most routing protocols use time to-live values (maximum number of hops a packet travels) to prevent a packet from traveling in the network for too long. Explain the three challenges of flooding # A node receiving a packet forwards this packet to all neighbors regardless of whether these neighbors have already received a copy of this packet. This is known as the implosion problem. The redundancy in sensor data from sensors that monitor similar physical environments is known as the overlap problem. Finally, flooding does not consider the resources available on all nodes, i.e. it is resource-blind. WSN application for each of the following categories: time-driven, event driven, and query-driven. An example of a time-driven WSN application is environmental monitoring, where temperature, pressure, or humidity may be reported periodically to obtain statistics for weather and climate studies. An example of an event-driven WSN application is a network that detects and reports wild fires, i.e. sensor data is only generated and disseminated when events of interest occur. A query-driven WSN application could be a surveillance system, where law enforcement officers access sensors to detect or track a suspect in public places. What is data-centric routing? Why is data-centric routing feasible (or even necessary) compared to routing based on identities (addresses)? If the focus is on the data generated by sensors and not the identity of the sensors generating these data, then we call a routing protocol data-centric. For example, if sensor data describing the environmental temperature must be routed to all sinks interested in such data, irrelevant of the data originator (i.e. which sensors produced the data), the focus is on the gathered data itself.
MAC protocol affects the design, performance, and efficiency of the routing protocol? The choice of MAC protocol can have several impacts on the design or performance of a routing protocol. For example, communication latencies are determined by the type of MAC protocol (contention-free versus contention-based) and on the use of duty cycling at the link layer. The MAC protocol is responsible for error recovery due to collisions and other interferences, i.e. the routing layer may decide to rely on the MAC layer for reliable communications or it may implement its own mechanisms to ensure reliability. Some approaches discussed in the chapter on MAC protocol rely on a tight integration of MAC and routing protocols, e.g. the LEACH protocol (which uses clusters that affect both medium access and routing decisions) or the DMAC protocol (which uses fixed routes from sensor nodes to a sink). Explain how the Z-MAC protocol allows nodes to determine their own local time frames instead of using a single global time frame. What are the disadvantages of Z-MAC? In Z-MAC, a node learns about its 1-hop and 2-hop neighbors during a setup phase and uses this information to select a time slot that only belongs to that node within its 2-hop neighborhood. Based on the neighborhood information, a node can then determine the periodicity of its assigned slot, i.e. the time frame to be used in its neighborhood. The consequence is that in denser neighborhoods (nodes with many neighbors), the time frame will be large, while in neighborhoods with low density the time frames may be very small. The main disadvantage of Z-MAC is that it requires an explicit setup phase, consuming both time and energy. Also, Z-MAC’s ECN (explicit contention notification) messages add to the traffic overheads. Describe the concept behind PMAC’s approach to adapting a node’s sleep durations to observed traffic. In PMAC, nodes use “patterns” to describe their tentative sleep and awake times, where a pattern is a bitmask, each bit representing a time slot (a ’0’ bit indicates a sleep interval and a ’1’ bit indicates an awake interval). Further, “schedules” are used to describe actual sleep/awake sequences, where a schedule is a series of ’0’s, followed by a ’1’. Comparable to TCP’s slow start behavior, PMAC doubles the sleep interval whenever it has no data to send until a certain thresholds of ’0’ has been reached, after which the ’0’s are increased linearly. However, once a node has data to transmit, the pattern is immediately reset to 1, allowing the node to wake up quickly to handle its traffic load. What is the “early sleeping problem” and how does T-MAC address this problem? The early sleeping problem is shown in Figure 6.11 in the book. The problem is that a node losing the medium, but wishing to transmit after the current transmission has completed will stay wake, while its intended receiver may be unaware of the node’s intent and will go to sleep. Therefore, the future-request-to- send technique adds another control message, the FTRS packet, which is sent by the node planning a future transmission after it has observed the CTS from the receiving node of the current transmission. This FTRS packet will be seen by its intended receiver. Which shortcoming of S-MAC does T-MAC address? Explain briefly T-MAC’s ability to adapt to traffic density. S-MAC uses a fixed-size listening period, whereas T-MAC uses an active period that adapts to traffic density. In T-MAC, nodes awake briefly to listen for activity, but return quickly to the sleep mode if no activity is detected. Nodes transmitting, receiving, or overhearing a message, on the other hand, stay awake for a brief timeout period after the transmission has finished to see if more traffic can be observed. That is, when traffic is heavy, nodes will stay awake longer, but if traffic is light, they only awake for very brief periods of time. How does the S-MAC protocol reduce the duty cycles of sensor nodes? How does the S-MAC protocol attempt to reduce collisions? How does it address the hidden-terminal problem? Name at least three disadvantages of the S-MAC protocol. S-MAC uses virtual clusters, where nodes using the same sleep- wake schedule belong to the same cluster. Schedules are exchanged via SYNC messages to allow each
radios on each sensor node, because of the increased device cost and the increased energy consumption of a multi-radio design. What are the advantages and disadvantages of the TRAMA protocol (compared to contention-based protocols)? TRAMA reduces the probability for collisions and dynamically determines when a node is allowed to transmit (based on traffic), thereby increasing the throughput. Since TRAMA uses a time- slotted channel, it allows nodes to determine when they must stay awake and when they can enter low-power sleep modes. Nodes must be awake during the random-access intervals. As with other contention-free protocols, communication in TRAMA may experience larger latencies than in contention based protocols since a node must wait for its scheduled slot before it can begin transmission. Five requirements of MAC protocols for wireless sensor networks: energy efficiency, scalability, adaptability, low latency, and reliability. Can you describe a concrete WSN application for each of these five requirements, where the requirement would be more important than the others? Energy efficiency is most important in WSN applications with battery-powered sensors that must operate without human intervention for an extended period of time. For example, sensor networks that detect wild fires or monitor civil infrastructure must be able to operate for months or years before their power sources can be replaced or recharged. Scalability is important in sensor networks that cover large geographic areas, e.g. a sensor network measuring the flow of lava on a volcano or in many military applications. Adaptability is crucial in networks where changes in network topology, density, and traffic are common, including vehicular sensing applications and other applications where sensors are mobile. Low latency is important when sensor data must be reacted upon quickly, e.g. in networks that detect explosions, the presence of toxic fumes, or the impending collapse of a bridge. Finally, reliability is essential in networks where the loss of important sensor data can be catastrophic. For example, in military sensor networks, sensor data reporting enemy movements must reliably reach the command and control center. Describe how the design of the MAC protocol affects the energy efficiency of a sensor node. Since the MAC protocol determines when a node may access a wireless channel for communication, its design determines when and for how long a node can power down its wireless radio to preserve energy. Some MAC protocols require nodes to be awake continuously to ensure that no messages are missed. Other protocols use schedules that determine exactly when a node must stay awake, allowing a node to sleep without the risk of losing messages. Other MAC protocol characteristics affecting energy efficiency include packet header overheads, reliability features (such as retransmission and error control mechanisms), and the number (and sizes) of control messages exchanged between nodes. Why is the IEEE 802.15.4 standard preferable over the IEEE 802.11 standard for most wireless sensor networks? The IEEE 802.15.4 standard was created specifically for low-power devices, with much lower data ranges compared to IEEE 802.11 and channels in the 868 MHz, 915 MHz, and 2.45 GHz ranges. Its focus on low power and low data rates much better meets the demands of wireless sensor networks than the IEEE 802.11 standard. Explain why the IEEE 802.11 standard uses three different “interframe spaces”. The “standard” interframe space DIFS is used to ensure that the medium has been sensed idle for a certain minimum amount of time before a transmission is attempted. IEEE 802.11 can use RTS and CTS frames to reserve the wireless channel, therefore, to ensure that no other node will begin a transmission when a reservation has occurred, a second (shorter) interframe space (SIFS) is used to separate RTS frames from CTS frames, CTS frames from DATA frames, and DATA frames from ACK frames. Finally, in the
PCF mode, an even shorter interframe space (PIFS) is used to ensure that PCF traffic has priority over traffic generated by devices in the DCF mode. The NAV field does not prevent the hidden-terminal problem. It is used to reduce the need for continuously sensing the medium, thereby allowing a node to preserve more power. What are the specific features of the IEEE 802.11 PSM (Power Saving Mode) and what are the main difficulties of using it in wireless sensor networks? IEEE 802.11 offers the power saving mode for nodes operating in the PCF mode. A device can inform the base station that it wishes to enter a low-power sleep mode using special control messages. To ensure that incoming messages can be received, the device periodically awakens to receive beacon messages from the base station. Assume that the RTS and CTS frames were as long as the DATA and ACK frames. Would there be any advantage to using the RTS/CTS approach? Explain why or why not. The reason RTS and CTS frames are used is to reserve the channel for a DATA frame. When a frame collides with another frame, it must be re-transmitted at a later time. If the DATA frame is large, it is preferable to exchange the much shorter RTS and CTS frames first, since their collision will result in less overhead. However, if the DATA frame is short (e.g. as short as the RTS frame), there is no benefit in first transmitting RTS and CTS frames. As a matter of fact, in such cases, using RTS and CTS frames introduces additional overheads because two additional frames must be transmitted. In a CSMA/CA network, nodes use a random delay before accessing the medium. Why is this being done? In CSMA, nodes can access the wireless medium immediately after it has been sensed idle. CSMA/CA is a variation of the CSMA protocol, where the number of collisions are reduced by randomly delaying medium access. That is, a node with a short random delay may access the medium sooner than a node choosing a larger random delay. The node with the larger delay will therefore overhear the transmission of its neighbor and it will not initiate a transmission, thereby avoiding a collision. Describe the problems in using CSMA as medium access control mechanism in a WSN. CSMA does not address the hidden-terminal problem and collisions can be costly, particularly if the transmitted frames are large. Neighboring nodes also do not know how long transmissions will take, making it difficult to decide whether to power down a wireless radio (and for how long). Contention-free medium access strategies avoid collisions by ensuring that when one device transmits, all other devices that could interfere with the reception of the transmitted data remain silent. Avoiding collisions can have positive impacts on communication latency and throughput. Most contention-free approaches are based on schedules, indicating exactly when a node can transmit or must listen. This also facilitate power management strategies, e.g. using such a schedule, a device knows exactly when it can power down its radio. On the other hand, contention-based medium access strategies do not avoid collisions, but instead provide mechanisms to recover from collisions. An advantage of such a strategy is that a device does not have to wait (and delay), but instead can transmit immediately. In networks with low traffic loads, this approach may result in lower latencies. A contention based strategy may also be easier to implement, since no schedules are required. Besides the potential collisions (which may lead to increased latency and reduced throughput), it may be more difficult to implement duty cycle techniques in a network using a contention based MAC protocol. Contention-based access strategies may be preferable in networks with light traffic loads and unpredictable (or sporadic) traffic patterns. Contention-free access schemes are useful when the traffic is predictable (e.g. sensor nodes periodically reporting their sensed data) and when it is important to maximize the sleep times of the wireless radios.
