1. SECURING BGP Matthew Nickasch nickaschm@uwplatt.edu University of Wisconsin-Platteville Dept. of Computer Science & Software Engineering

2. BGP – Quick Overview External Routing Protocol ▫Interior vs. Exterior Gateway Protocols ▫The Autonomous System (AS) ▫Routing between ISPs BGP - Only EGP in use

3. Functions of BGP Shortest path not priority Routing policy Removal of routing loops Broken link removal Determine which IPs “go where” ▫Responsibility of address blocks

4. Basic Operation of BGP Connections between border routers peer with neighboring ASes TCP port 179 Manual session creation ▫Complete copies of routing table sent to neighbors ▫Evaluate received routes ▫Better route through neighbor? UPDATE ▫Only update when routes change

5. Intra/Inter AS Routing AS 100 AS 200 AS 500 AS 400 AS 300 (PEERING RULES) AS 500 LAN_CORBOR_1 BOR_2

6. Routing Policy When to send routes? Where to send routes? Peering Responsibility ▫Accept routes from known peers ▫Don’t accept routes from non-peers ▫Route efficiency hampered by political boundaries ▫Route preference configuration ACL (Access Control Lists) Error Checking

7. An ISP’s Use of BGP Importance of filtering Address block size ▫Prefix “overload” (small networks/subnets) ▫Delegate BGP handling to ISP Utilization of peering paths

8. BGP Looking Glass Demo Looking Glass ▫route-views.ab.bb.telus.comroute-views.ab.bb.telus.com ARIN AS Whois Lookup Prefix / AS Query

9. Security Considerations BGP – Single Point of Failure? ▫Only EGP in use ▫Comparison with IGPs  OSPF, RIP, IS-IS, etc. ▫EGP standardization difficult  “Big” router vendors  Early Cisco stronghold  Now Juniper, Nortel, etc.  Different vendors want different implementations

10. Security Considerations A “trusting protocol” ▫Very little error checking  Route verification requires route lookups  30,000 + ASes! * 120,000 unique routes! = TIME  Garbage in, garbage out Physical Infrastructure ▫9/11 “Meet Me Facilities” ▫Peering Points ▫Physical Router Compromisation

11. Security Considerations Human error ▫Human error to human intent (exploit errors) Remote router compromisation ▫IOS vulnerabilities, etc. Social Engineering vulnerabilities BGP traffic sniffing ▫Message injection / modification ▫Man-in-the-middle

12. Security – Assembling the Risk SPOF Trusting protocol + lack of error checking Physical Infrastructure Human Error Router security flaws Social Engineering Unencrypted message transport DoS / MIM / TCP-style attacks Supporting entire Internet routing structure

13. YouTube Route Hijacking Prime example of human ‘error’ ▫Illustrates violation of route trust ▫Easily replicated by attacker ▫Proves that attack vectors are in-place  Compromised router could cause similar results  Relatively simple attack, “invalid route announcement”  Potential large worldwide attack against many ASes

14. YouTube Route Hijacking YT always announces 208.65.152.0/22 Pakistani Telecom announces 208.65.153.0/24 Routes propagate to bordering ASes ▫Traffic destined for network directed to PT YT announces 208.65.153.0/24 Duplicate announcement entry (shortest path) YT announces 208.65.153.128/25 ▫Longest-prefix-match-rule  Most specific route

15. YouTube (AS 36561) Pre-Hijack

16. Pakistani Tel (AS 17557) Hijack

17. Detecting Invalid/False Routes Response ▫“Firefighter” mentality to BGP problems ▫Symptom-based response too late ▫Cooperation between ASes? ▫Governing ‘body’ for BGP disputes

18. YouTube – What We’ve Learned ISP Routing Policy “Routing Registry” – RIPE Certificate-based approvals BGP not substitute for ACLs! Exploitation of protocol “trust” Rapid replication Extreme vulnerability

19. Protocol Security MD5 & other encryption ▫Hard to standardize between all ASes ▫Vendor agreement issues “Reinventing” the protocol ▫Secure BGP ▫PGBGP ▫Revisions to existing BGP

20. Secure BGP (SBGP) Public key infrastructure (PKI) ▫Authentication/ownership of IP address space ▫AS identity verification ▫Encrypting BGP Update messages Implementation ▫Vendor support must be unanimous ▫All ASes must agree to adopt SBGP, or any other protocol-level change

21. Secure BGP (SBGP) Doesn’t prevent human error ▫“Encrypting garbage” Origins ▫NSA/DoD initial support (1997) ▫DARPA Next Steps ▫PKI infrastructure, CA ▫Oversight organization for PKI? Hosting?

22. Pretty Good BGP (PGBGP) Cautiously accepting/updating routes ▫Suspicious updaters ▫Quarantine routes ▫Time-delay updates Implementation – Adapt PGBGP logic? ▫Vendor support could vary – depends on route- accepting algorithms ▫Introduce PGBGP logic into existing BGP environment.

23. Layered Security Analysis PHYSICAL SECURITY SOCIAL ENGINEERING HUMAN ERROR INADEQUATE CORP /ORG / ISP POLICY AVAILABLE “TCP-STYLE” VECTORS SOURCE / SENDER AUTHENTICATION BGP PROTOCOL WEAKNESSES SINGLE POINT OF FAILURE (INTERNET)

24. Layered Solution Protocol level ▫SBGP (PKI) + PGBGP (update logic) = a more secure solution ▫Limits peer trust, introduces authentication and encryption ▫AS identity verification ▫Slower route change replication throughout the Internet ▫Not the end-all solution!

25. Layered Solution Implement stringent ISP routing policy Implement SBGP + PGBGP logic into existing protocol ▫Attain vendor agreement on implementation Reduce human error Enforce proper-use of BGP (ACL example) Router security / minimize vectors Physical security, etc.

26. Q/A ? Matthew Nickasch nickaschm@uwplatt.edu