Privacy in Library RFIDAttacks and Proposals

David MolnarDavid Wagner

{dmolnar, daw}@eecs.berkeley.edu

Privacy in Libraries

• Must protect what patrons are reading• Library only source of info for many• FBI Library Awareness Program

– 1973-1988, official policy to monitor “suspicious” persons’ reading habits

– Library privacy laws passed as backlash– Even with PATRIOT act, need court order

• Privacy adversaries not limited to FBI – Marketers, Scientologists, pick your favorite…

RFID & Library Overview• RFID = Radio Frequency IDentification• One RFID tag per book• Each RFID tag has ``bar code” ID

number– Unique to each book, may identify library

• Exit gates read RFID for anti-theft• 13.56MHz passive RFID

– ISO 15693, Checkpoint, TAGSYS C220– Read range depends on antenna size

• Deployed in Oakland, Santa Clara, 130+

Why RFID?

• Speedy self-checkout– reduce library employee RSI (carpal tunnel)

• Security devices– ensure checkout occured

• Inventory Tracking

Pictures courtesy Santa Clara City Library

Privacy and Ubiquitous Readers

• Read range not whole privacy story• Even full in-view readers can be

problem– Scan at airport security, events, etc. – Like metal detectors now– Not clear what read or how used

• Readers easy to camouflage– RFID reader looks like store anti-theft gate

Library RFID ArchitectureLibrary database

•No authentication between reader and tag•Database maps bar code (title, status)

Bar code

Attack: Book Scanning• Can Mallory scan me and tell what I am

reading?– No reader – tag authentication– Anyone can read tag data

• Most deployments data limited to bar code– Some vendors suggest more

• Need library database• In CA, database protected by law

– Varies by state

Attack: Hotlisting and Profiling

• Hotlisting is book on special list?– It’s real – FBI and almanacs

• Profiling – bar code prefix identifies library– Is library in predominantly minority area?

• Bar code never changes so hotlisting easy– Walk into library, read bar code– See the book again, recognize book– Does not need library database

Attack: Book Tracking

• Bar code never changes• Can link different sightings• Track book movement

– Spatial movement– Combine w/video for person-to-person

• “This person checked out same book as terrorist”

• Does not need library database

“Security Bit” Denial of Service

• RFID used for anti-theft• Some vendors store “security bit” on tag

– Security bit = checked out/not checked out– Bit re-written each checkout

• ISO 15693 tags have “write, then lock” – No way to unlock data, no password on lock

• Adversary can lock security bit data page• Can’t change security bit tag useless

Collision Avoidance and Privacy

• Collision avoidance protocols identify tag• Example: ISO 15693 mandates MFR ID

• Read passwords,changing ID,etc. don’t help• Privacy requires attention to all layers

MaskDoes mask match MFR ID?Respond if yes

RFID Limitations

• RFID powered only when near reader– No precomputation, no caching

• RFID have few gates (< 5,000 for security)

• Randomness difficult on RFID• “Cryptography” extremely hard on RFID

– Best we can do is a few XOR• Future generation tags focus on price,

not on security features

Problem: Private Authentication

• Reader does not know tag ID• Authentication must preserve privacy• Privacy and authentication in tension

Random Transaction IDs• Required: rewritable tags• Attacker model: outside the library• On checkout

– Obtain random # r– Write (r, D) to DB– Erase D & Write r to tag

• On checkin– Use r to lookup D– Write D to tag

Attacks Against Random IDs

• Tracking– Possible– Only for checkout duration

• Hot-listing– Not Possible

• Comparison-based– Not possible

Password Enhancement

• Eavesdropping– Not the same in the two channels– Tag to Reader is Harder

Hello

r

cmd, p=r s

Good and Bad of Passwords

• Good– low computation cost– s remains secure (info-theoretically!)– r is independent of book info

• cannot be tracked

• Bad– Requires randomness on tag

Private Authentication

• Every tag has a secret– DB has all (secret, ID) pairs

• Basic ID– Reader sends a nonce– Tag sends new nonce– Tag sends ID f(s, 0, nonce 1, nonce 2)– Reader checks the whole DB

• Problems?

Tree-based

• Set it all up as a binary balanced tree

• log(n) rounds– Check if the secret is on the left or right– Get down to a single leaf

• Advanced version– 1 million tags– 168 bits of communication

Summary

• Library RFID is here now• All today’s technology has privacy

flaws• Privacy is achievable efficiently• Work still ongoing

Acknowledgements• Many, many people to thank!

In no particular order:

Peter Warfield, Karen Duffy (Santa Clara City Library), Karen Saunders (Santa Clara City Library), Susan Hildreth (San Francisco Public Library), Al Skinner (Checkpoint), Paul Simon (Checkpoint),Doug Karp(Checkpoint), Rebekah E. Anderson (3M), Jackie Griffin(Berkeley Public Library), Elena Engel (BPL), Alicia Abramson(BPL)Lee Tien (Electronic Frontier Foundation), Dan Moniz (EFF), Laura Quliter (Boalt Hall School of Law, UC-Berkeley), Jennifer Urban(Boalt), Nathaniel Good (SIMS), Samuelson Technology and PolicyLaw Clinic at Boalt Hall School of Law, Elizabeth Miles (Boalt),John Han (SIMS), Ross Stapleton-Gray, Eric Ipsen, Oleg Boyarsky(Library Automation/FlashScan), Laura Smart (Library RFIDWeblog/Cal State Pomona), Craig K. Harmon (ISO 18000 committee),Justin Chen (SVCWireless RFID SIG), Steve Halliday(ISO 18000 committee), Zulfikar Ramzan (NTT DoCoMo), Craig Gentry (NTTDoCoMo), Hoeteck Wee, Matt Piotrowski, Jayanth Kumar Kannan, Kris Hildrum, David Schultz, and Rupert Scammell(RSA Security).

Questions?