CSE 703 – Wireless Network Security – Principles and Practices Dept. of Computer Science and Engineering Achieving security in Wireless Mesh Networks Abhishek Gautam Armaan Goyal Wireless Mesh Networks (WMNs) • Rising technology - offering low-cost high bandwidth community services that support several vital applications. • Examples of such applications: • Provisioning of internet access in rural areas • City-wide security surveillance • Information services in public transportation systems. • Emergency and disaster recovery operations. • High-speed wireless metropolitan area networks (WMANs) Where are WMNs being used currently? • U.S. Military are now using wireless mesh networking to connect their computers, mainly ruggedized laptops, in field operations. • The laptops in the One Laptop per Child program use wireless mesh networking to enable students to exchange files and get on the Internet even though they lack wired or cell phone or other physical connections in their area. Components of a WMN: The three tier Architecture • • • • MR – Mesh Router MC/WC – Mesh Client or Wireless Client IGW/GWN – Internet Gateway Router or Gateway Node BS – Base Station Routing techniques in Mesh Networks • Routing happens on similar grounds as in an IP network. • A number of nodes take part in the data forwarding process and hence the routing is based on multicasts. • A mesh-based multicast protocol is used that connects multiple sources and receivers. • Path selection is done on the basis of routing metrics just like in simple IP networks. Metrics will be discussed later. Design Concerns in Multicast Routing • Transmissions in multicast are less consistent than in unicast for several reasons. • Unicast communications are sent reliably, using link-layer unicast transmissions (involves link layer acknowledgements) • They also involve the RTS/CTS mechanism • Multicasts are sent unreliably using link-layer broadcast (no acknowledgements) and there is no RTS/CTS mechanism involved. Move to High-Throughput. How? • Conventional routing metrics – Hop Count? • Hop Count was used in Internal Gateway Routing Protocols like RIP. • Best path? – One containing the least number of hops to destination. • Scalable? – No. EIGRP and OSPF took over RIP. • Talking about WMNs, when the metric is hop count, the paths are likely to include lossy wireless links. • Focus has shifted to High-Throughput – choose paths depending on the quality of wireless links. The Metric! So how to achieve HT Multicast Routing in WMNs? • ODMRP takes care of this. • ODMRP stands for On-Demand Multicast Routing Protocol. • It is a multicast routing protocol for WMNs which uses a mesh of nodes for each multicast group. • Nodes are added to the mesh through a route selection and activation protocol. ODMRP Terminologies • Node – A device that participates in multicast group • Neighbor – Nodes that are within the radio transmission range. • Forwarding group - A group of nodes participating in multicast packet forwarding. • Multicast mesh - The topology defined by the link connection between forwarding group members. • Join query – The special data packet sent by multicast sources to establish and update group memberships and routes. • Join reply - The table broadcasted by each multicast receiver and forwarding node to establish and update group membership and routes ODMRP - Working • The source periodically recreates the mesh by flooding a Join Query message in the network in order to refresh the membership information and update the routes. • Join Query messages are flooded using a basic flood suppression mechanism, in which nodes only process the first received copy of a flooded message • When a receiver node gets a Join Query message, it activates the path from itself to the source • Receiver broadcasts a Join Reply message that contains entries for each multicast group it wants to join; each entry has a next hop field filled with the corresponding upstream node ODMRP - Working • When an intermediate node receives a Join Reply message, it checks for its own identifier. • This helps it to know whether it is on the path to the source or not, by checking if the next hop field of any of the entries in the message matches its own identifier. • If so, it makes itself a node part of the mesh (the Forwarding Group) and creates and broadcasts a new Join Reply built upon the matched entries • Once the Join Reply messages reach the source, the multicast receivers become connected to the source through a mesh of nodes (Forwarding Group) which ensures the delivery of multicast data ODMRP-HT • The High Throughput version of ODMRP. Enhance ODMRP with high-throughput metrics in the path selection process • Differs from ODMRP in two respects: 1. Instead of selecting routes based on minimum delay (which results in choosing the fastest routes), ODMRP-HT selects routes based on a link-quality metric 2. ODMRP-HT uses a weighted flood suppression mechanism to flood Join Query messages instead of using a basic flood suppression. Two reasons for vulnerabilities • Since many of the protocols assume a pre-existing cooperative relation among the nodes, for successful working of these protocols, the participating nodes need to be honest and wellbehaving with no malicious or dishonest intentions • Two prime reasons for vulnerabilities: 1. The assumption of pre-existing trust relationships among the nodes 2. The absence of a central administrator A discussion of various attacks • We discuss various attacks that can take place in WMNs categorizing them by different OSI layers. • We explore the attacks at L1-L4 and the defense mechanisms associated with them. • Then we discuss in detail one of the L3 attack i.e. Sybil Attack and one of the L4 attack i.e. the DoS attacks. Layer 1 - Jamming Attacks • The physical layer is responsible for frequency selection, carrier frequency generation, signal detection, modulation, and data encryption • Jammer is an entity who is purposefully trying to interfere with transmission (by transmitting RF signals in the wireless channel) and reception of message across the wireless channel • As with any radio-based medium, the possibility of a jamming attack in WMNs is high. • An adversary can potentially disrupt communication in the entire network by strategically distributing the jamming sources. • Jamming attacks can be more complex to detect if the attacking devices do not obey the MAC layer protocols Layer 2 – Passive Eavesdropping • Multi-hop wireless networks like WMNs are also prone to internal eavesdropping by the intermediate hops. • A malicious intermediate node keeps the copy of all the data that it forwards without the knowledge of any other nodes in the network. • Although passive eavesdropping does not affect the network, functionality directly, it leads to the compromise in data confidentiality and data integrity Layer 2 – Link Layer Jamming Attack • Attacker may transmit regular MAC frame headers (with no payload) on the transmission channel which conforms to the MAC protocol being used in the victim network. • Legitimate nodes always find the channel busy and back off for a random period of time before sensing the channel again. CSMA/CD (carrier sense multiple access collision detection) • This leads to the denial-of-service for the legitimate nodes and also enables the jamming node to conserve its energy. • These attacks can be effective even if encryption techniques such as wired equivalent privacy (WEP) and Wi-Fi protocol access (WPA) have been employed. This is because the sensor that assists the jammer can still monitor the packet size, timing, and sequence to guide the jammer. Layer 2 – Mac Spoofing • Modifying the MAC addresses in transmitted frames is referred to as MAC spoofing • MAC addresses are globally unique. So they have often been used as an authentication factor or as a unique identifier for granting varying levels of network privileges to a user. Example – IDS and ACLs • Since the identity is compromised, this enables the attacker to evade intrusion detection systems (IDSs) in the networks. • An attacker can eavesdrop on the network to determine the MAC addresses of legitimate devices. This enables the attacker to masquerade as a legitimate user and gain access to the network. Layer 2 – Replay Attack • An external malicious node (M) eavesdrops on the broadcast communication between two nodes A and B. • Later, it can replay the (eavesdropped) messages to gain access to the network resources. • The authentication information is replayed where the attacker M deceives a node B to believe that the attacker is a legitimate node A. • Malicious node M, which is an intermediate hop between two nodes A and B, can keep a copy of all relayed data. It can then retransmit this data later to gain an unauthorized access to the network resources Layer 2 – Replay Attack Layer 3 - Wormhole Layer 3 - Wormhole • Two or more malicious nodes collude together by establishing a tunnel using an efficient communication medium • During the route discovery phase of the on- demand routing protocols, the RREQ (Route Request) messages are forwarded between the malicious nodes using the established tunnel. • Consequently, the malicious nodes are added in the path from the source to the destination • Once the malicious nodes are included in the routing path, the malicious nodes either drop all the packets, resulting in complete denial of service, or drop the packets selectively to avoid detection. Layer 3 - Blackhole Layer 3 - Blackhole • Leads to denial of service in WMNs • Also exploits the route discovery mechanism of ondemand routing protocols • The malicious node always replies positively to a RREQ, although it may not have a valid route to the destination • Almost all the traffic within the neighborhood of the malicious node will be directed towards the malicious node, which may drop all the packets, causing a denial of service. Layer 3 – Sybil Attack • Malicious node creates multiple identities in the network, each appearing as a legitimate node • Redundancy in the system is exploited by creating multiple identities and controlling the considerable system resources • Legitimate nodes assume malicious nodes as regular distinct network nodes and add these identities to the list of distinct paths available to a particular destination. • When packets travel from source to destination, these intermediate malicious nodes process the packets and can perform any of the attacks discussed so far. • Phenomenon of Path Diversity is also diminished and this results in a poor performance. Sybil Attack Sybil Attack Overview: Dimension 1 • Direct Communication Sybil nodes communicate directly with legitimate nodes. When a genuine node sends a radio message to a Sybil node, a malicious devices listens it. Similarly messages sent from a Sybil node are in fact sent from a malicious device. • Indirect Communication In this version of the attack, no legitimate nodes are able to communicate directly with the Sybil nodes. Messages sent to a Sybil node are routed via one of these malicious nodes, which make up to pass on the message to a Sybil node. Sybil Attack Overview: Dimension 2 • Fabricated Identities The attacker can simply generate random new Sybil identities. For instance, if each node is identified by a 32-bit integer, the attacker can simply allocate each Sybil node a random 32-bit value. • Stolen Identities If a method is given to identify genuine node identities, an attacker cannot fabricate new identities. Then the attacker requires to assign other legitimate identities to Sybil nodes (Identity theft) Secure Multicast Routing – Defense against Sybil Attack • The discussion so far leads us to the question of achieving secure multicast routing while accommodating high-throughput metrics. • A secure version of ODMRP called S-ODMRP is used to achieve this. • S-ODMRP assures the delivery of data from the source to the multicast receivers even in the existence of Byzantine attackers, provided the receivers are reachable via non-adversarial paths. S-ODMRP • Uses Secure Message Authentication to eliminate processing non-authenticated messages. This avoids a variety of attacks • But what about the attacks on the mesh structure? • What about the packet dropping attacks? • These attacks are more challenging to defend when it comes to HT metrics. • To defend these attacks, RateGuard technique is used by S-ODMRP. RateGuard • Takes into account PDR(Packet Delivery Ratio) • Relies on the observation that attackers do not affect the multicast protocol unless they cause a drop in the packet delivery ratio (PDR) • Uses a detection and a reaction approach to work against these attacks. • Attacker nodes are detected through a measurement-based detection protocol component • Then such nodes are isolated through an accusationbased reaction protocol component. RateGuard - Measurement-based attack detection • Takes in account that whatever the type of attack may be, the effect of an attack is that data is not delivered at a rate consistent with the advertised path quality (the claimed one) • We rely on the ability of honest nodes to detect the discrepancy between the expected PDR (ePDR) and the perceived PDR (pPDR). RateGuard - Measurement-based attack detection • A node can estimate the ePDR of a route from the value of the metric for that route • the node can determine the pPDR for a route by measuring the rate at which it receives data packets from its upstream on that route • If (ePDR−pPDR) for a route becomes larger than a detection threshold δ, then nodes suspect that the route is under attack because the route failed to deliver data at a rate consistent with its claimed quality RateGuard - Accusation-based attack reaction • Uses a controlled-accusation mechanism • in which A node, on detecting malicious behavior, temporarily accuses the suspected node by flooding in the network an Accusation message • This message contains its own identity (the accuser node) and the identity of the accused node, as well as the duration of the accusation • As long as the accusation is valid, metrics advertised by an accused node will be ignored and the node will not be selected as part of the Forwarding Group RateGuard - Accusation-based attack reaction • To prevent the abuse of the accusation mechanism by attackers, a node is not allowed to issue a new accusation before its previously issued accusation expires. • Accused nodes can still act as receivers even though they are excluded from the Forwarding Group. Metric Poisoning effect • In order to deal with the metric poisoning effect due to metric manipulation attacks, the metric, in the whole network is refreshed shortly after attack detection. • The metric refreshment is achieved automatically through the periodic JOIN QUERY messages. S-ODMRP Mesh Creation • S-ODMRP Mesh is created in the same way as the ODMRP-HT mesh • Source node S periodically broadcasts to the entire network a Join Query message in order to refresh the membership information and to update the routes • The Join Query message is signed by S and is propagated using a weighted flood suppression mechanism • Nodes only process Join Query messages that have valid signatures – This is how secured nodes are achieved in a WMN. S-ODMRP Limitations • S-ODMRP limits a node to blame at most one other node at a time. This shows that attacker nodes should be a minority in the network. • This means that some attacker nodes will be left non-accused and will be susceptible to attacks and deny service to many receivers through metric manipulation. • This approach is not usable in Sybil attack scenario • To overcome this, Random Key Pre-distribution (RKP) technique is integrated with the S-ODMRP to defend against the Sybil attacks. The Defense Mechanism - RKP • What we really need? – validate that identity of each node is the only presented by the equivalent physical node. • RKP (Random Key Pre-distribution) facilitates the nodes to create secure links to other nodes. • RKP works by assigning a random group of keys to each node. In key set-up phase, each node can discover the common keys with its neighbors, which are used as a shared secret session key. • This ensures node-to-node secrecy The Defense Mechanism - RKP • At any time, network is capable of verifying the keys that an identity claims to possess. • So there is very slight probability that a fabricated identity will have a major intersection with key set. Hence it will not pass the key validation phase. • Validation can be of two types - Direct and Indirect • In direct, each node challenges an identity using the knowledge it possess and takes a decision independent of other nodes. Thus all nodes may not reach a global decision state • In indirect, there is a central trusted party that does the validation. Direct v/s Indirect • • Direct approach has less overhead. Messages exchanged only between two identities. Due to the memory constraint in WMNs, each node has limited knowledge about the other node that it could use to pose a challenge to its identity. Hence this offers lower defense performance • • Indirect approach has more overhead. Messages are exchanged between the nodes and the third party (validator) Indirect validation relies on the trusted party and hence this tradeoff doesn’t exist in this case. Hence offers higher defense performance. Another approach – The Key Pool Approach • This approach randomly assign k keys to each node from a pool of m keys. If any two neighboring nodes identify that they share q common keys, they can set up a secret link. • To use this scheme to defend against the Sybil attack, each node’s identity is the indices in sorted order of the keys that it holds • Ω(ID) = {Kβ1,Kβ2,...,Kβk} is the notation that explains that this is the set of keys assigned to ID where ID is the identity of the node and βi is the index of the ith key in the key pool. • The set of keys that ID possesses is given by: βi = PRFH(ID)(i), where PRF is a pseudo random function abd H is a hash function. The Key Pool Approach • For an attacker to construct new identities to use in the Sybil attack, it will need to capture genuine nodes and read off the keys and thus establish a compromised pool set S. • The attack will then try to fabricate usable Sybil identities. If an identity say ID’ passes the initialization phase and becomes a part of the WMN, it becomes a usable Sybil identity. • But the verifier can challenge the identity by requesting it to prove that it possesses the keys it claims to have. It goes by the rule that, If∃Ki,Ki ∈ Ω(ID′), Ki ∈/ S, and if some legitimate entity E in the sensor network knows Ki, then E can discover that ID’ is cheating by challenging ID’ using Ki. Performance Comparison Layer 4 Attacks • The Layer 4- transport layer is susceptible to Denial of Service attacks, Replay attacks • We will discuss two security protocols- EMSA and SAE and further a 4 way handshaking mechanism used for authentication • Suggested measures to secure the 4 way handshaking process using techniques like periodic updates and key hashing Mesh Security Systems: • Security relies on ability to protect message integrity • Requires guaranteeing confidentiality and authenticity of data packets • Can be achieved using a reliable authentication and association process • E.g. of attack- Black Hole attack • To secure network over long haul, use periodic updating of keys(in detail later) • Also updated when active attack has been detected Mesh Security Protocols: • Two major mesh security protocols- EMSA and SAE • EMSA-Efficient mesh security association • SAE-Simultaneous authentication of equals EMSA for Multigate Networks • Based on the use of a Mesh Key hierarchy • Tree based routing scheme is used(Extension of Hybrid wireless Mesh Protocol)-HWMP • Master Gateway acts as MA and MKD • MKD derives keys to create mesh hierarchy • Stores all MP’s authentication information • Consists of peer link establishment followed by EAP(Extensible Authentication Protocol) • 4 way handshaking for key derivation between every pair of MP • After Mesh Key Holder Security Handshake(MKHSH), the supplicant becomes MA • 4 way handshake results in deriving Pairwise transient key(PTK – for unicast messages) and (GTK- for multicast messages) EMSA for Multigate Networks-contd: • For MP to become an MA, PTK-KD(PTK for Key distribution) has to be derived from KDK • Then it starts the same process for one of its children • If child is already authenticated, only the process of Subsequent Authentication takes place • Process of link establishment continues untill each node has PTK, GTK and PTK-KD SAE for Multigate Networks: Single password used by all MPs for authentication No authentication server is involved unlike EMSA Any one of the pair can initiate protocol The parties are referred to as MP-A and MP-B and identified by MAC addresses • Before message exchange both will generate PWE based on shared password and MAC addresses • Two random numbers rand and mask produced and used to complete SAE • After success, both generate a PMK(Pairwise Master Key) used in the 4 way handshake to derive PTK and GTK • • • • Periodic key refreshment strategy • All key materials are updated at regular intervals • Achieved by initiating EAP or SAE and 4 way handshaking to get new keys before expiration of current keys • In EMSA, lifetime of PMK-MKD (Mesh Key Distributor PMK) and KDK (Key Distribution Key) shouldn’t be more than that of MSK (Master Session Key) • Lifetime of PTK and PMK-MA remains same as that of PMK-MKD • Lifetime of PTK-KD should be same as that of KDK • Post expiry of the lifetime, each key holder deletes their respective keys • Similar conditions on lifetime occur in SAE where PMK and PTK are bound to Master PMK • Upon expiration, MP’s operation ends and resumes after successful security process Periodic key refreshment strategy • Leads to disruption of network if lifecycle of keys is short • If keys are unchanged for longer time, network becomes vulnerable to attack • Hence, keys changed dynamically over regular intervals and in case of active attack • To update key materials before expiration, existing routes in the network must be maintained • If not, each update will cause fresh routing and association of involved MPs causing significant delays Key Updating for EMSA Key Updating for SAE Security Improved 4 way handshaking • Assumption- The PMK is known only to authenticator and supplicant • Attacks can occur only before generation of first PTK because of Link layer data encryption • Hence it all depends on protecting the integrity of the PTK • Security is developed and assessed on the assumption of a DoS attack by the adversary during the 4 way handshake • The intruder can forge other MP’s MAC addresses, eavesdrop and forge received messages • SPA, AA, Snonce and Anonce, represent Mac addresses and nonces of supplicant and authenticator • MICPTK{} represents Message Integrity Code Security Improved 4 way handshaking • First message is not encrypted, hence it can be tampered with easily • 4 way handshake can be disrupted by forging Message 1 • Intruder sends forged Message 1 after Message 2 is sent by supplicant forcing him to generate a new PTK-different from the earlier one, hence handshake terminated • A multiple message DOS attack can exhaust supplicant’s memory and cause significant delays Security Improved 4 way handshaking • DoS attack also possible on Message 3 • If intruder sends a forged Message 3 with a fake RSNE(Robust Security Network Element), it mismatches with the original RSNE and handshake terminates Security measures for 4 way handshake Two suggested techniques Merkle-Tree based hashing or a single hash function scheme Either of the techniques is used to protect both Message 1 and 3 MA can use one way hash function like SHA1 or SHA2 Merkle tree is a binary tree consisting of leaf tokens and internal nodes • Each node is the hash of the concatenation of its left and right children nodes • Merkle tree with height H has 2^H leaf tokens • Due to one wayness, computationally impossible to derive leaf tokens from root of tree • • • • • Security measures for 4 way handshake • To secure the 4 way handshaking, for Message 1, 2 Merkle trees are considered • In the first, MA derives the root and includes it in encrypted Message 1 • PMK not present in Message 1 • Supplicant received encrypted message and generates the root itself and then compares with the received value • Merkle Tree makes it impossible to derive the PMK • To prevent replay attacks, this tree is never reused • If 4 way handshaking has to be done again, a new tree is used, a new root is calculated and encrypted and sent to the supplicant along with PMK in Message 3 Security measures for 4 way handshake • For Message 3, AA RSNE can be protected using first Merkle tree in a similar fashion • A modification- replace the first Merkle tree by single Hash of Message 1/Message 3 • Single Hash is not computationally efficient for Authentication when compared to Merkle tree • The Merkle has the flexibility to construct the second Merkle tree which provides safety against replay attacks-not offered by single hash function Protocol Verification and Results • To analyse the flaw of the 4 way handshaking and the resistance of the proposed Merkle tree, ProVerif is used • It reconstructs attacks used during the handshake process • Due to the key refreshment strategy, PMK PTK and GTK are periodically updated • More frequent updates means less complex Merkle tree but with more overhead and delay • A simulation using three gateways GW 1,2 and 3 is run for the EMSA and SAE protocols • A set of MSK/MPMK lifetime values is used to study impact of overhead caused by periodical refreshment of keys • Multiple updates for PTK and GTK are performed to mitigate vulnerability Protocol Verification and Results • We can see that when no updates are done, nodes stop communication on PTK key expiry and EMSA or SAE authentications have to be reinitiated • The performance is only slightly worse than non secure systems when keys are periodically refreshed Protocol Verification and Results • Without periodical updating EMSA and SAE schemes give the worst performance because re-initiation of authentication pauses data transfer and causes overhead • SAE outperforms EMSA due to lower overhead • The system performance does not degrade significantly with more frequent key refreshment Protocol Verification and Results • It is evident that after applying Merkle tree based hashing, performances remain unaffected under DoS attack • Here PTK GTK and PMK are auto updated after every 200 secs • For blackhole attack, in SAE, an intruder cannot be excluded from the network without changing password in all MPs and restarting Protocol Verification and Results • For EMSA the system is capable of removing the intruder with the involvement of authentication server-an advantage over SAE Conclusion • Mesh networks are quite popular due to their numerous advantages but they are susceptible to a variety of attacks • We discussed layer by layer the attacks that are possible on the mesh networks • We discussed in detail- Sybil attacks and defense mechanisms and then DoS attacks and defense mechanisms for the same Acknowledgements The material used in the slides was sourced from the following papers: • Smart Grid Mesh Network Security Using Dynamic Key Distribution With Merkle Tree 4-Way Handshaking by Bin Hu, Senior Member, IEEE, and Hamid Gharavi, Life Fellow, IEEE - IEEE TRANSACTIONS ON SMART GRID, VOL. 5, NO. 2, MARCH 2014 • Security and Privacy Issues in Wireless Mesh Networks: A Survey by Jaydip Sen • Secure Multicast Routing in Wireless Mesh Networks against Sybil Attack by Dhivya.J - International Journal of Scientific & Engineering Research, Volume 5, Issue 4, April-2014 Thank You