Secure producer mobility in information-centric network Alberto - PowerPoint PPT Presentation

Secure producer mobility in information-centric network Alberto Compagno, Xuan Zeng, Luca Muscariello, Giovanna Carofiglio, Jordan Auge Cisco, SystemX,UPMC September 25, 2017 1

Mobility in 5G § 5G requirements on mobility: § Seamless (low latency, packet loss, etc) § Continuity over dense & heterogeneous access (LTE, wifi) § Calls for new and effective mobility solutions 2

Support mobility in ICN § Consumer mobility ->naturally supported § Producer mobility -> challenging § Tracing-based approach(kite, Mapme), promising: § Meet 5G requirements: low latency, loss, network head Security consideration are inadequate § 3

How does trace-based solution work? § Producer updates forwarding states(PIT or FIB) of a subset of routers Interest update(IU) R2 Interest flow R3 R4 producer FIB direction R1 4

How does trace-based solution work? § Producer updates forwarding states(PIT or FIB) of a subset of routers producer R2 Interest flow R3 R4 FIB direction R1 5

trace-based solution: prefix hijacking attack Q:what if IU is from attacker? Interest update producer Interest update R2 Interest flow R3 R4 FIB direction R1 6

trace-based solution: prefix hijacking attack producer pollute cache! black-holed! R2 Interest privacy flow R3 R4 FIB direction R1 7

Challenges to protect trace-based approach from prefix hijacking? 8

Challenges to prevent prefix hijacking (1/2) 1. Distributed Interest update producer 2. Lightweight 9

Challenges to prevent prefix hijacking (2/2) 3. Deal with an attacker that can compromise edge routers May allow to R2 generate valid IU R3 R4 FIB direction R1 10

Existing approaches § Signature based approach: § Expensive for hardware at network access § See evaluation section later § Session key based approach: § CellularIP and telemIP: shared network key stolen compromises whole network 11

Our prefix attestation protocol? 12

Prefix attestation protocol: high level view § Only entitled producer can generate valid interest updates § Distribute minimal crypto info to network § We call this crypto info security context § Validate IU locally Sec.context Registration server Sec.context Sec.context Sec.context producer Registration 13

Prefix attestation protocol: high level view § Only entitled producer can generate valid interest updates § Distribute minimal crypto info to network § We call this crypto info security context § Validate IU locally Sec.context Registration server Sec.context Sec.context Sec.context IU producer 14

How to design security context? 15

Security context requirements § Allow fast validation -> crypto hash § Allow to validate but not generate genuine IU, -> hash chain Prevent attacker R2 generating valid IU R3 R4 FIB direction R1 16

Security context using hash chain § hash chain(originally by Lamport) 1 st authen. A authenticates to B: B: H n (s) H n-1 (s) A Hash matches, OK 17

Security context using hash chain § hash chain(originally by Lamport) 2 nd authen. A authenticates to B: B: H n-1 (s) H n-2 (s) A Hash matches again, OK 18

prefix attestation protocol: leveraging hash chain § Producer: ith IU, send with H n-i (s) H n-i (s) producer Sec.context Sec.context H n-2 (s) Sec.context prefix seq. No sec. context H n-1 (s)producer /p 0 H n (s p ) Sec.context 19

Evaluation? 20

Evaluation: computation overhead !"# Analytical model: goodput = !"# ∗% &'()*++ ,#∗% -*'./0 η = fraction of interest update(%) 21

Evaluation: computation overhead § Optimal case: no verification on interest update § Goodput decreases anyway as IU take up resources 22

Evaluation: computation overhead § Signature verification § Goodput drops to 0 with small percent of IU(3%) 23

Evaluation: computation overhead § Hash chain: one hash per IU verification § Maintains 90% of optimal goodput (low overhead) 24

Evaluation: computation overhead § Hash chain: many hashes per IU verification § By ~200 hashes , similar results w.r.t signature verification. 25

Evaluation: storage overhead § Storage overhead vs No. of mobile producers § Hash chain: 50MB per router needed for Millions of Mobiles. More scalable. 26

Conclusion & future work § We propose an attestation protocol to secure trace-based producer mobility in ICN: § Initial results confirm it’s light weight § Run unchanged over different hardware § Future work: Evaluation on real hardware and workload § § Exploit routing to refresh sec. context. Thanks! xuan.zeng@irt-system.fr 27