Interception of 802.11 Communications Analysis

1. Intercepting Mobile Communications:The Insecurity of 802.11 Nikita Borisov Ian Goldberg David Wagner UC Berkeley Zero-Knowledge Sys UC Berkeley Presented by Kunjan Naik

2. Agenda • Introduction • WEP protocol Brief description Security goals • Keystream reuse attacks • Attacks involving message authentication • Countermeasures • Conclusion

3. Typical Scenario Ad-Hoc Network Infrastructure Network

4. WEP Protocol • Wired Equivalent Privacy • Link Layer Security Protocol • Goals : Confidentiality : Protection against eavesdropping Access Control : Restrict accessibility Data Integrity : Correctness of data

5. RC4 and Stream Ciphers • RC4 encryption is Vernam Cipher • RC4 is a stream cipher • Generates pseudo random keystream fromthe key( IV || Key) Pseudo-random number generator Encryptionkey Cipher text byte Plain text data byte

6. WEP Protocol • Mobile station shares key with Access Point • Transmitting a message M Compute checksum of M and append it to M Generate keystream using RC4(IV,Key) Xor <M,C(M)> with keystream Transmit IV and cipher text • Upon receiving Reverse steps

7. Packet Format and Encapsulation • Encryption Algorithm = RC4 • Key length = 40. IV length = 24 • C = RC4(IV,K) xor <M, C(M> Message CRC XOR Keystream = RC(IV,k) IV Cipher Text

8. Shared secret distributed out of band Challenge (Nonce) Response (Nonce RC4 encrypted under shared key) WEP Authentication • Authentication key distributed out-of-band • Access Point generates a “randomly generated” challenge • Station encrypts challenge using pre-shared secret • Denial of service attack Decrypted nonce OK?

9. So What are the Problems? • Shared key mechanism Same shared key in the network • Attacks based on Keystream Reuse IV collision • Decryption Dictionaries • Message modification • Message injection.

10. Shared key mechanism • Single key or array of shared keys betweenall mobile stations in the network • key length is just 40 bits. • Key management is a misnomer • Shared keys changes rarely. • Chances of IV collision proportional to number of users.

11. IV Collision • P1 and P2 packets with same IV • C1 = P1 xor RC4(IV,Shared Key) • C2 = P2 xor RC4(IV,Shared Key) • C1 xor C2 = P1 xor P2 • Attacker knows the Xor of two plaintexts • Given P1 or P2 easy to find other • More packets with same IV : More easier • Dragging cribs, frequency analysis methods

12. Key Reuse • Shared key same in both directions • Keystream depends on IV as Key is fixed • IV included in unencrypted portion of message • IV reset to 0 when initialized • Easy to find collisions • After 16 million packets ( worst case ) IV repeats

13. How to find keystream reuse? • IV space - 2^24 possibilities • Collision after few minutes on a busy AP • WEP standard recommends IV to be changed (but does not require) per packet • More so, IV set to 0 when re-initialized • Finding keystream reuse is therefore easy

14. How to get plaintext? • IP traffic predictable - well defined structures and message content • Login sequences and Welcome messages • Sniffing Authentication challenge - plain and cipher text both • Sending packets from outside - ping • Broadcast packets in both encrypted and unencrypted form - for some implementations

15. Attack from both ends Internet Attacker sends data Attacker AP AP encrypts plaintext data Attacker MS

16. Attack from both sides cont’d • Attacker will send packets from internet to mobile station and AP will encrypt them for attacker • Flip bits to change destination address to host we control - IP checksum needs to be modified • Sufficient number of packets with different IV’swill enable the attacker to build a decryption dictionary

17. Decryption Dictionaries • Xoring cipher text and plain text gives keystream • Store one to one mapping of IV to RC4(IV,Key) • Xor any packet with corresponding IV and read data • Number of entries in table 2^24 • 1500 bytes per packet - 24 GB • Independent of key size - depends on IV only. • Building table ensures immediate decryption

18. Message Authentication • CRC checksum for data integrity • CRC resilient against random errors and not malicious attacks • CRC is independent of IV and key • CRC and RC4 are linear • CRC(X xor Y) = CRC(X) xor CRC(Y) • So, changing bits in packet is easy

19. Message modification • C = RC4(IV,K) * {M,C(M)} • Let M’ = M * D • D is arbitrarily chosen and * => xor • C’ = C * { D, C(D) } RC4(IV,K) * {M,C(M)} * {D,C(D)} RC4(IV,K) * {M * D,C(M) * C(D)} RC4(IV,K) * {M * D,C(M * D)} RC4(IV,K) * {M’, C(M’)} • Effectively Attacker does C’ = C * {D, C(D)}

20. Message Injection • Attacker needs plain text and cipher text • Attacker has fake message F and computes C(F) • Computes C’ = {F,C(F)} xor RC4(VI,key) • Transmits (VI, C’) • Reuse old IV’s and circumvent access control • Attacker can authenticate himself using message injection

21. Message Decryption • IP redirection - Send encrypted packet to host on the internet ; IP checksum and firewall issues • Reaction attacks - TCP packets will be dropped for incorrect checksum and TCP ack for the correct packets. Modify packet and check recipients reaction

22. Attack Practicality • Use off the shelf wireless card and software radio • Sit outside competitor’s office and sniff packets • Reverse engineer firmware to inject packets • Dictionaries - Has to be done once

23. Countermeasures • Data encryption is not enough - access control through data authentication is must • Use block ciphers • Increase key length • Make checksum keyed function of message • Put wireless network outside firewall -treat it as public network

24. Conclusion • Public review is essential • All three goals Confidentiality - Attacker can read traffic Access Control - Attacker can inject traffic Data Integrity - Attacker can modify traffic • Use VPN, IPSec, SSH along with WEP • ESN is supposed to solve all problems