New Directions in Detection, Security and Privacy for RFID Leonid Bolotnyy and Gabriel Robins Department of Computer Science, UVa 1 Thesis Multi-tags, “yoking-proofs”, and physical unclonable functions can improve reliability, security, and privacy in radio frequency identification (RFID) systems. 2 Progress • L. Bolotnyy and G. Robins, Multi-Tag Radio Frequency Identification Systems, IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), pp. 83-88, 2005 • L. Bolotnyy and G. Robins, Randomized Pseudo-Random Function Tree Walking Algorithm for Secure Radio Frequency Identification, IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), pp. 43-48, 2005 • L. Bolotnyy and G. Robins, Generalized ‘Yoking-Proofs’ for a Group of RFID Tags, IEEE International Conference on Mobile and Ubiquitous Systems (Mobiquitous), 2006 • L. Bolotnyy and G. Robins, PUF-Based Security and Privacy in RFID Systems, IEEE International Conference on Pervasive Computing (PerCom), 2007 • Several additional papers in progress • NSF Cyber Trust proposal (submitted January 2007) • Deutsche Telekom (largest in EU) offered to patent our multi-tags idea 3 Introduction • RFID • Tags types: passive semi-passive active • Frequencies: Low (125KHz), High (13.56MHz), UHF (915MHz) • Coupling methods: signal Reader antenna signal Reader antenna Inductive coupling Backscatter coupling 4 History • Radar invented - 1935 • EAS invented - early 1960’s • First RFID patent filed - 1973 • First RFID book published - 1999 • Auto-ID Center formed - 1999 • EPCglobal formed - 2004 • First RFID game marketed - 2006 5 Thesis Proposal • Improve tag detection • Improve security and privacy Auditing algorithms for RFID “Yoking-Proofs” Inter-tag communication Definition of privacy PUF-based security Algorithms PUF design 6 Why Multi-Tag RFID? • Bar-codes vs. RFID – line-of-sight – scanning rate • Unreliability of tag detection – radio noise is ubiquitous – liquids and metals are opaque to RF • milk, water, juice • metal-foil wrappers – Wal-Mart experiments (2005) • 90% tag detection at case level • 95% detection on conveyor belts • 66% detection of individual items inside fully loaded pallets – Our preliminary experiments support data above 7 Applications of Multi-Tags 8 The Power of an Angle • Inductive coupling: voltage ~ sin(β), distance ~ (power)1/6 • Far-field propagation: voltage ~ sin2(β), distance ~ (power)1/2 B-field • Optimal Tag Placement: 4 Expected angle (in Degrees) 1 β 3 2 65 61.86 58.11 60 55 47.98 50 4 2 [ x(2 cos x) dx ( 0 4 2 x)(2 cos x) dx] / 45 40 35 32.7 [ x(2 cos x)dx] /(2 ) 2 0 30 1 2 3 Number of Tags 4 9 Benefits and Costs of Multi-Tags • PROS – – – – – – – – increases expected induced voltage on tag increases operational range of system increases memory per object improves availability improves reliability improves durability provides potential security enhancement new applications • CONS – increases system cost – modestly complicates manufacturing – potentially increases tags’ interrogation time 10 Experimental Apparatus and Experiments with Multi-Tags • Equipment • Experiments – Measure detection of ~20 multi-tagged objects • With/without metals and liquids – Rotate multi-tagged object mixes • 1, 2, 3, & 4 tags per object – Vary tag, reader, and antenna types – Vary distances, geometry, power – Multi-tags vs. multiple readers 11 Preliminary Experimental Results 1 Average Detection Probability 0.9 0.8 0.7 0.6 0.5 0.4 0.3 2 Readers, 1 Reader, 2 Readers, 1 Reader, Δ= 4.0% 0.2 Δ=18.7% 0.1 Δ= 6.1% 0 1 2 3 4 5 6 7 8 9 10 11 12 13 2 Tags 2 Tags 1 Tag 1 Tag 86.6% 82.6% 63.9% 57.8% 14 16 15 Δ=22.7% Δ=24.8% 17 18 19 20 Object Number 12 Security and Privacy in RFID • Privacy A B C Alice was here: A, B, C privacy 13 Security and Privacy in RFID • Privacy: difficult to track tags • Security – Secure Identification – Tag Authentication – Message Authentication f(c) f(r, ID) c σ (m) m – Ownership Transfer – Auditing 14 “Yoking-Proofs” • Yoking: joining together / simultaneous presence of multiple tags • Key Observation: Passive tags can communicate with each other through reader • Problem Statement: Generate proof that a group of passive tags were identified nearly-simultaneously • Applications – verify that: – medicine bottle sold together with instructions – tools sold together with safety devices – matching parts were delivered together – several forms of ID were presented – a group of people was present at a meeting 15 Assumptions and Goals • Assumptions – – – – – Tags are passive Tags have limited computational abilities Tags can compute a keyed hash function Tags can maintain some state Verifier is trusted and powerful • Solution Goals – – – – Allow readers to be adversarial Make valid proofs improbable to forge Allow verifier to verify proofs off-line Detect replays of valid proofs • Timer on-board a tag – FCC regulations: protocol termination < 400ms – Capacitor discharge can implement timeout 16 Generalized “Yoking-Proof” Protocol Idea: construct a chain of mutually dependent MACs 1 2 3 5 4 Anonymous Yoking: tags keep their identities private Speedup yoking protocols by splitting chain into arcs 17 Inter-Tag Communication in RFID • • • • • • • Idea: heterogeneity in ubiquitous computing “Yoking proofs” Battery-less sensing Tags as mailboxes Tags as proxies Location access control Tags partitioned into groups – Group leader in charge of authentication and access control • Subordinate reader-tag authentication 18 PUF-Based Security and Privacy • Digital crypto implementations require 1000’s of gates • Low-cost alternatives – Pseudonyms / one-time pads – Low complexity / power hash function designs – Hardware-based solutions • Definition of privacy that incorporates hardware attacks • PUF definition • Security is based on: – wire delays – gate delays – quantum mechanical fluctuations • PUF characteristics – uniqueness – reliability – unpredictability 19 PUF-Based Algorithms • Identification Sequence: ID, p(ID), …, pk(ID) • It is important to have – – – – a reliable PUF no loops in PUF chains no identical PUF outputs no impersonation attacks • Authentication Pairs: c1, p(c1), c2, p(c2), ..., cn, p(cn) • Verify that at least the desired fraction of challenge-response pairs is correct • MAC based on PUF – Motivation: “yoking-proofs”, signing sensor data – large keys – cannot support arbitrary messages • Large message set • Small message set 20 PUF-Based Ownership Transfer • Ownership Transfer • To maintain privacy we need – ownership privacy – forward privacy • Physical security is especially important • Solutions – – – – public key cryptography knowledge of owners sequence trusted authority short period of privacy 21 Comparison of PUF With Digital Hash Functions algorithm MD4 MD5 SHA-256 AES Yuksel PUF # of gates 7350 8400 10868 3400 1701 545 • Reference PUF: 545 gates for 64-bit input – 6 to 8 gates for each input bit – 33 gates to measure the delay • Low gate count of PUF has a cost – – – – probabilistic outputs difficult to characterize analytically non-unique computation extra storage • Different attack target for adversaries – model building rather than key discovery • Physical security – hard to break tag and remain undetected 22 PUF Design • Attacks on PUF – – – – • impersonation modeling hardware tampering side-channel Weaknesses of existing PUF reliability • New PUF design – – • no oscillating circuit sub-threshold voltage Compare different non-linear delay approaches 23 Conclusion and Research Plan • Contributions – Multi-Tags • tag objects with multiple tags to improve detection – Security and Privacy • Yoking proofs • Inter-tag communication • Hardware-based security – PUFs • Plan for the next 5 months – – – – finish multi-tag experiments define privacy w.r.t. physical attacks design / evaluate improved PUF circuits publish more papers 24 • Bolotnyy and Robins, Multi-Tag Radio Frequency Identification Systems,IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), pp. 83-88, 2005 • Bolotnyy and Robins, Randomized Tree Walking Algorithm for Secure RFID, IEEE Workshop on Automatic Identification Advanced Technologies (AutoID), pp. 43-48, 2005 • Bolotnyy and Robins, Generalized ‘Yoking-Proofs’ for a Group of RFID Tags, IEEE International Conference on Mobile and Ubiquitous Systems (Mobiquitous), 2006 • Bolotnyy and Robins, PUF-Based Security and Privacy in RFID Systems, IEEE International Conference on Pervasive Computing (PerCom), 2007 Back Up Slides 26 Related Work on Multi-Tags • Two-antennas per tag to determine location • Four tags per object to determine movement direction • Multiple tags to increase reliability (for visually impaired) • Random placement of two tags on playing cards • Splitting tag ID into Class ID and Pure ID • Up to three tags to determine object-person interaction 27 Types of Multi-Tags • Redundant Tags • Complimentary Tags • Dual-Tags – Own Memory Only – Shared Memory Only – Own and Shared Memory • Triple-Tags • n-Tags 28 Detection Distance with Multi-Tags Expected Factor of Distance Increase 1.7 1.57 Increase Factor 1.6 1.5 1.63 1.37 1.4 Far-Field Propagation 1.3 Inductive Coupling 1.2 1.1 1 1.06 1.08 1.09 2 3 4 1 1 Num ber of Tags 29 Effects of Multi-Tags on Anti-Collision Algorithms Algorithm Redundant Tags Dual-Tags Binary No Affect No Affect Binary Variant No Affect No Affect Randomized Doubles Time** No Affect* STAC Causes DOS No Affect* Slotted Aloha Doubles Time** No Affect* *If Dual-Tags communicate to form a single response **Assuming an object is tagged with two tags 30 Related Work on “Yoking-Proofs” • Juels [2004] – protocol is limited to two tags – no timely timer update (minor/crucial omission) • Saito and Sakurai [2005] – – – – solution relies on timestamps generated by trusted database violates original problem statement one tag is assumed to be more powerful than the others vulnerable to “future timestamp” attack • Piramuthu [2006] – – – – discusses inapplicable replay-attack problem of Juels’ protocol independently observes the problem with Saito/Sakurai protocol proposed fix only works for a pair of tags violates original problem statement 31 Speeding Up The Yoking Protocol Idea: split cycle into several sequences of dependent MACs starting / closing tags Requires – multiple readers or multiple antennas – anti-collision protocol 32 Related Work on PUF • Optical PUF [Ravikanth 2001] • Silicon PUF [Gassend et al 2002] – design, implementation, simulation, manufacturing – authentication algorithm – controlled PUF • PUF in RFID – off-line reader authentication using public key cryptography [Tuyls et al 2006] 33 PUF-Based Authentication Reader Tag GetID probv(n) ID GetResponse(c1) p(c1) n n μi(1-μ)n-i probv = 1 - ∑ i i=t+1 . . . probf(n) GetResponse(cn) p(cn) n probf = 1 - ∑ j=t+1 n τj(1- τ)n-j j α < probv ≤ 1 and probf ≤ β ≤ 1 0 ≤ t ≤ n-1 34 PUF-Based Identification Algorithm • Tag stores its identifier: ID • Database stores: ID, p(ID), …, pk(ID) • Upon reader’s query, the tag – responds with p(ID) – updates its ID with p(ID) • It is important to have – a reliable PUF – no loops in PUF chains – no identical PUF outputs • Assumptions – passive adversaries (otherwise, denial of service possible) – physical compromise of tags not possible – reliable PUF 35 PUF-Based MAC Algorithms • MAC = (K, τ, υ) • valid signature σ : υ K(M, σ) = 1 • forged signature σ’ : υ K(M’, σ’) = 1, M = M’ • Need to protect against replay attacks • MAC based on PUF – large keys – cannot support arbitrary messages – Motivational example: buyer/seller • Large message set σ (m) = c, r1, ..., rn, pc(r1, m), ..., pc(rn, m) • Small message set σ (m) = c, pc(1)(m), ..., pc(n) (m), ..., c+q-1, pc+q-1(1)(m), pc+q-1(n)(m) Using PUF to Detect and Restore Privacy of Compromised System s1,0 s2,0 s3,0 s1,1 s2,1 s3,1 s3,2 s2,2 s3,3 s3, s3,5 s1,2 s2,3 s3,6 s2,4 s3,7 s3,8 s3,9 s2,5 s3,10 4 1. Detect potential tag compromise 2. Update secrets of affected tags 37