Wireless Network Security Spring 2015

Transcription

1 Wireless Network Security Spring 2015 Patrick Tague Class #10 Network Layer Threats; Identity Mgmt Patrick Tague 1

2 Class #10 Summary of wireless network layer threats Specific threats related to identity 2015 Patrick Tague 2

3 Wireless Networking Message Source Relays / Routers Sink / Destination 2015 Patrick Tague 3

4 Network Layer Functionality The network layer is primarily responsible for establishing end-to-end paths and delivering packets over them Includes several fundamental services: Addressing: network ID management Routing: finding/establishing paths Forwarding: delivering packets Interactions with Transport layer and Link/MAC layer 2015 Patrick Tague 4

5 Addressing Before routing can be performed, nodes need some sort of ID or address Address/ID types range from local to global, just like in the postal system (streets up to ZIP codes) In very large-scale systems (e.g. Internet), addresses must have some sort of structure IP addresses follow a specific hierarchy and are reused within each domain Within a domain and in small-scale systems (e.g. MANET/WSN), addresses are typically unstructured or random Address management needed within a domain to prevent duplication and other failure scenarios 2015 Patrick Tague 5

6 Addressing Threats Addresses can be changed arbitrarily Allows for address spoofing Masquerading as other node(s) Potential for a large number of attacks Changing identity to prevent detection/punishment Attackers can infiltrate address management protocols (ARP, DHCP) to cause problems Inducing address duplication Forcing frequent address changes Manipulating forwarding schemes 2015 Patrick Tague 6

7 Routing Routing = path management Routing does not involve actual sending of packets from source to destination(s), only sets up the path Lives in the control plane Involves path setup/discovery, maintenance, and teardown Challenges in MANET/WSN environments Route using multiple untrusted relay nodes Resource and capability limitations No centralized authority or monitor Secure routing often relies on existing key mgmt Patrick Tague 7

8 Routing Threats Just as with other types of misbehavior, routers can be greedy, non-cooperative, or malicious Greedy routers can refuse route discovery requests in order to save their own resources Non-cooperative routers can choose to selectively accept route requests to specific sources/dests Malicious routers can persuade route discovery protocols so paths pass through them, avoid them, or take unnecessary detours 2015 Patrick Tague 8

9 Black-hole attack: Path Attraction A malicious router broadcasts false claims of being close to the destination in order to attract all traffic and drop it Gray-hole attack: Similar to black-hole attack, except it only drops some packets selectively Ex: forward all routing control packets but drop all data Worm-hole attack: Colluding routers create a low-latency long-distance outof-band channel to attract routing paths and control data flow 2015 Patrick Tague 9

10 Detours: Path Manipulation A malicious router can modify/inject control packets to force selection of sub-optimal routes Gratuitous detours : Greedy routers can avoid being on a selected route by advertising long delays or creating virtual nodes Could be considered a form of Sybil attack, where all personalities are on the routing path 2015 Patrick Tague 10

11 Targeted blacklisting: Route Subversion In any routing protocols using blacklisting, attackers can accuse/slander/blame others to force them onto the blacklist DoS Rushing attacks: Attackers can quickly disseminate forged requests, causing later valid requests to be dropped 2015 Patrick Tague 11

12 Forwarding Forwarding = point-to-point data management Forwarding involves actual sending of packets from source to destination(s) on given routing paths Lives in the data plane Correct forwarding involves Sending the correct packets Maintaining packet order Respecting headers and rules Relaying in a timely manner Respecting rate control mechanisms 2015 Patrick Tague 12

13 Forwarding Threats Misbehavior in the forwarding mechanism (often called Byzantine forwarding) includes various ways of going against forwarding rules Dropping packets Modifying packet contents or header information Injecting bogus packets on source's behalf Forwarding to the wrong next hop Disrespecting rate control (flooding or throttling) 2015 Patrick Tague 13

14 Network Privacy Threats Routing protocols inherently reveal information to curious/malicious eavesdroppers An attacker can listen to route discovery interactions and learn (1) locations of source and destination nodes, (2) type of interactions between nodes, (3) commonly used paths, (4) network events, or (5) data These are all issues of location privacy, network privacy, and data privacy due solely to the routing process 2015 Patrick Tague 14

15 Let's go through these different threats in some detail, starting with addressing 2015 Patrick Tague 15

16 Addressing In traditional networking, each device (radio) has two identities, in the form of addresses MAC address: hardware address of the radio needed for link layer communication (e.g., 802.3, ) Hard-coded into the NIC In theory, unique and static IP address: network layer address used for routing and some other higher layer services Virtual software address 2015 Patrick Tague 16

17 MAC Addresses MAC addresses in the Internet Ethernet and WiFi use MAC addresses for link layer communication Independent of any higher-layer functionality Link layer frames carry source and destination MAC addresses (6B each) MAC addresses in other systems Not typically used in sensor networks due to overhead Not needed if other addressing is available 2015 Patrick Tague 17

18 IP Addresses IP addresses in the Internet Network layer and above use IP addresses for some identity purposes Independent of whatever is below the network layer IP addresses must be unique IP addresses in other systems To support common applications, most designers are aiming to support IP addressing (to some extent) 2015 Patrick Tague 18

19 IP Address Resolution In most Internet domains, IP addresses are assigned centrally using DHCP and bound to MAC addresses using ARP DHCP = Dynamic Host Configuration Protocol: host asks server for IP address, which it keeps until expiry ARP = Address Resolution Protocol: host asks other hosts for MAC address corresponding to an IP address 2015 Patrick Tague 19

20 ARP image from [Whalen et al., 2001] 2015 Patrick Tague 20

21 Limitations MAC addresses are no longer hardware-bound Most Linux-like systems allow software to change MAC address used, despite hard-coded MAC address Many devices don't have (unique) MAC addresses DHCP is impractical for distributed systems Requires centralization High overhead in dynamic systems ARP has high overhead in distributed systems Requires request flooding 2015 Patrick Tague 21

22 Distributed Addressing Problem: How should IP addresses (or other suitable identities) be determined in a distributed system such that: Addresses are compact(-able) for low-overhead communication in sensors or embedded devices Network overhead is (relatively) low Addresses are (sufficiently) unique Systems can split and join Duplicate addresses can be detected and fixed Address space is large enough and dynamic 2015 Patrick Tague 22

23 A Few Approaches Random selection with duplicate address detection (DAD) Send a query to the selected address; if no response, the address probably isn't in conflict Requires flooding a query through the entire network Merging existing networks is difficult MANETconf Configured initiator nodes act like a server that can assign addresses to requesters who arrive later Configured node floods notification and assigns address if no nodes respond negatively Merging existing networks is difficult 2015 Patrick Tague 23

24 PACMAN [Weniger, JSAC 2005] PACMAN = Passive Auto-Configuration for Mobile Ad hoc Networks Architecture for efficient distributed MANET address auto-configuration 2015 Patrick Tague 24

25 Address Assignment To avoid overhead of flooding network to check address uniqueness, PACMAN assigns addresses passively and relies on network to expose conflicts Each node chooses its address using a probabilistic algorithm and a list of known used addresses See the paper for details 2015 Patrick Tague 25

26 Address Encoding To minimize overhead, PACMAN encodes MANET IP addresses MANET uses a fixed IP prefix (2B for IPv4, 8B for IPv6) Node ID only needs log 2 N bits to support N nodes Pad with lots of zeros, but only need to know #0s 2015 Patrick Tague 26

27 Passive DAD PDAD relies on observation of events that: Never occur in case of unique addresses but always occur in case of address duplication e.g., receiving a route reply when no request was sent Usually don't occur in case of unique addresses but sometimes occur in case of address duplication e.g., link states in a route reply change completely Upon detection of duplication, at least one node can reinitialize the address assignment This also allows relatively easy management of network split and merge events 2015 Patrick Tague 27

28 Security Issues PACMAN and many similar approaches were not designed with malicious behaviors in mind Threats [Wang et al., 2005]: Address spoofing attacker spoofs the IP address of a victim and hijacks its traffic False address conflict attacker injects conflict messages (or events) to a target victim Address exhaustion attacker claims many addresses to deny service or prevent nodes from joining Negative reply in cases where approval is needed to join, attacker can prevent nodes from joining 2015 Patrick Tague 28

29 Secure MANET Auto-Conf [Wang et al., 2005] Bind the IP address to a public key to authenticate auto-configuration processes New node A chooses an IP address as the hash of its public key A sends a query to the network for the IP address using a signed, time-stamped Duplicate Address Probe If a receiving node B has an IP conflict, it checks signatures (authenticity, replay prevention, etc.) and conditionally replies with a signed, time-stamped Address Conflict Notice If A receives ACN from B, it checks signatures and conditionally starts over with a new key pair If no reply within a fixed time period, A joins the network using the generated IP address 2015 Patrick Tague 29

30 Benefits of the Approach Forces the attacker to find a public key that hashes to a victim's IP address before launching the attack Even with relatively small address space, computation/storage overhead is prohibitive Detailed analysis in the paper 2015 Patrick Tague 30

31 Trust-Based Auto-Conf [Hu & Mitchell, 2005] Misbehavior in the requester-initiator model (MANETconf) Initiator can intentionally assign conflicting address Requester can flood requests repeatedly, causing resource depletion and/or DoS Malicious node can falsely claim that candidate addresses are already in use, causing excess request floods, resource depletion, and DoS Instead of relying on arbitrary nodes, keep track of which nodes are good and which are bad 2015 Patrick Tague 31

32 Trust Model For nodes A and B, A's trust in B is given by T A (B) Each node A has a threshold trust value T A * such that it will interact with B only if T A (B) T A * Trust values can be computed based on past behaviors 2015 Patrick Tague 32

33 Choosing a Trustable Initiator New requester N broadcasts a Neighbor_Query with its threshold T N * Each receiver sends N a InitREP reply with neighbor IDs who have trust values T N * N can chooses its initiator as the neighbor appearing in the most InitREP messages Malicious node is unlikely to be chosen unless majority of neighbors are malicious 2015 Patrick Tague 33

34 Duplicate Address Check If initiator A gets an Add_Collision message from a node B in response to an Initiator_Request: If B has been previously blacklisted, ignore it If T A (B) T A *, then believe B and start over Otherwise, declare B malicious, add B to the blacklist, and send a Malicious_Suspect message about B to other nodes Other nodes only believe A's Malicious_Suspect message if their trust value in A is high enough 2015 Patrick Tague 34

35 3rd-Party Duplication Check If a node B detects an address collision between two other nodes, it notifies both of them If a receiving node A gets such a notification from node B: If B has been previously blacklisted, ignore it If T A (B) T A *, believe B and choose a new address Otherwise, add B to the blacklist 2015 Patrick Tague 35

36 Calculating Trust Values A calculates its trust in B by: Monitoring and measuring a number of performance characteristics, including past behaviors in address assignment Each characteristic maps to a value in [0,1] Trust T A (B) is a weighted combination of the various percharacteristic values For distant nodes B, A must rely on establishing a trusted path to B More detail on evaluating trust in a later class 2015 Patrick Tague 36

37 Summary Discussed distributed addressing, threats, and a few approaches to secure auto-configuration PACMAN: Passive auto-configuration for MANETs [Weniger; JSAC 2005] Secure address auto-configuration for MANETs [Wang, Reeves, and Ning; MobiQuitous 2005] Secure auto-configuration in MANETs using trust [Hu and Mitchell; MSN 2005] 2015 Patrick Tague 37

38 Feb 19: Routing and Forwarding Security 2015 Patrick Tague 38