RFID Security

RFID Security and Privacy

References

1. 6.857: RFID Security and Privacy, Massachusetts Institute of Technology, Computer Science and Artificial Intelligence Laboratory, 6.857 Lecture, November 2nd, 2004

2. Privacy in pervasive computing What can technologists do?, David Wagner, U.C. Berkeley, In collaboration with David Molnar, Andrea Soppera, Ari Juels

Abstract and Outline

• Abstract: What is RFID, how does it affect security and privacy, and what can we do about it?

• Outline– RFID Introduction, History, and Applications– Security Threats and Adversarial Model– Countermeasures– Protocols for private identification– The challenge of scalability; trees of secrets

The tide is turning...

Pervasive computing is coming...

It’s time to get serious about privacy.

What is RFID?

• Radio Frequency Identification: Identify physical objects through a radio interface.

• Many different technologies called “RFID”.

• Others types of auto-ID systems include:– Optical barcodes– Radiological tracers– Chemical taggants

RFID System Primer

Three Main Components:

• Tags, or transponders, affixed to objects and carry identifying data.

• Readers, or transceivers, read or write tag data and interface with back-end databases.

• Back-end databases correlate data stored on tags with physical objects.

RFID Adhesive Labels

4 cm

An RFID “Smart Shelf”

Reader

Any Shelf becomes a Smart Shelf by means

of the    Flexible Panasonic Antenna

System.

System Interface

Reader

01.203D2A.916E8B.8719BAE03C

Tag Database

Reader

Network

DataProcessing

RFID Advantages

• Non-line-of-sight nature

• Dynamic memory : real-time addition of data without removing, changing or replacing tag

• Tags can be read through substances– Snow -Paint– Fog -Crusted grime– Ice

RFID History• Earliest Patent: John Logie Baird (1926)• “Identify Friend or Foe” (IFF) systems developed

by the British RAF to identify friendly aircraft. • Both sides secretly tracked their enemy’s IFF.• How do you identify yourself only to your friends?

Don’t shoot! We’re British!

Oh. We’re British too!

Digression #1: Related Military Applications

• IFF still used today for aircraft and missiles. Obviously classified.

• Could envision an IFF system for soldiers.

• Lots of military interest in pervasive networks of cheap, RFID-like sensors.

• Monitoring pipelines, detecting biological agents, tracking munitions, etc.

Commercial Applications

• Early Applications:– Tracking boxcars and shipping containers.– Cows: RFID ear tags.– Bulky, rugged, and expensive devices.

• The RFID Killer Application?

Supply-Chain Management(Not Gum)

• First Universal Product Code scanned was on a pack of Juicy Fruit gum in 1976.

• Every day, over five billion barcodes are scanned around the world.

• But barcodes are slow, need line of sight, physical alignment, and take up packaging “real estate”.

• Over one billion RFID tags on the market.• Example: Gillette’s “shrinkage” problem.

Example applications:• Electronic passports• ID cards and badges• Proximity cards, building access control• Automatic payment systems (Fastrak, EZPass)• Item tagging & tracking, inventory management

Key technologies:• RFID• Contactless smart card

Identification systems

Challenge: privacy (and security) for ID systems

Modern RFID Applications

• Supply-Chain Management– Inventory Control– Logistics– Retail Check-Out

• Access Control: MIT Proximity Cards.

• Payment Systems: Mobil SpeedPass.

• Medical Records: Pet tracking chips.

Prada's RFID Closet

MIT Prox Card

Tag Power Source

• Passive: – All power comes from a reader’s interrogation signal.– Tag’s are inactive unless a reader activates them.– Passive powering is the cheapest, but shortest range.

• Semi-Passive: – Tags have an on-board power source (battery).– Cannot initiate communications, but can be sensors.– Longer read range, more cost for battery.

• Active:– On-board power and can initiate communications.

RFID tags are passive, powered by reader, carry identity

Privacy issues: Unwanted tracking of people and items

Introduction to RFID

Power

Identity

Reader Tag

• Tags might lack writable non-volatile memory• Takes more energy to permanently write bits• Thus, state might only last as long as tag is powered

• Cryptography is expensive• Public-key out of reach for all but priciest tags• AES within reach for mid-class tags? [Feldhofer]• Can’t take random number generation for granted

• Readers might not be network-connected

RFID systems are resource-limited

Functionality ClassesClass Nickname Memory Power

SourceFeatures

0 Anti-Shoplift Tags

None Passive Article Surveillance

1 ElectronicProduct Code

Read-Only Passive Identification Only

2 Electronic Product Code

Read/Write

Passive Data Logging

3 Sensor Tags Read/Write

Semi-Passive Environmental Sensors

4 Smart Dust Read/Write

Active Ad Hoc Networking

Operating Frequencies

Range Class LF HF UHF

Frequency Range

120-140 KHz 13.56 MHz 868-956 MHz

Maximum Range?

3 meters 3 meters 10 meters

Typical Range 10-20 centimeters

10-20 centimeters

3 meters

RFID Established Standards

• 13.56 MHz– ISO 15693– ISO 14443

e-commerce; lower power; shorter ranges

• 120-140 kHz– ISO 7810

RFID Technical Standards

• ISO 18000-1-Generic Parameters for Air Interface for Global Interface• ISO 18000-2-Parameters for Air Interface <135 kHz• ISO 18000-3-Parameters for Air Interface at 13.56 MHz• ISO 18000-4 -Parameters for Air Interface at 2.45 GHz• ISO 18000-5 -Parameters for Air Interface at 5.8 GHz• ISO 18000-6 -Parameters for Air Interface at 860-930 MHz*• ISO 18000-7 -Parameters for Air Interface at 433.92 MHz**

*Proposed Name Change-UHF**In Development

ISO/IEC JTC 1/SC 31/WG 4/SG 3RFID for Item Management Air Interface

Intended read range

Com

pu

tati

on

ISO 14443 E-passports, ID cardsUS$5

ISO 15693Library booksUS$0.50

EPCWalMartUS$0.20

10cm

3DES,RSA

sym.-keycrypto

no crypto

1m 3m

RFID technologies vary widely

FrequenciesFREQUENCY BAND

CHARACTERISTICS TYPICAL APPLICATIONS

LOW

100-500 KHz

SHORT TO MEDIUM READ RANGE

INEXPENSIVE

LOW READ SPEED

ACCESS CONTROL

ANIMAL IDENTIFICATION

INVENTORY CONTROL

HIGH

10-15MHz

850-950MHZ

SHORT TO MEDIUM READ RANGE

POTENTIALLY INEXPENSIVE

MEDIUM READING SPEED

ACCESS CONTROL

SMART CARDS

ULTRA-HIGH

2.4-5.8 GHZ

LONG READ RANGE

HIGH READING SPEED

LINE OF SIGHT REQUIRED

EXPENSIVE

RAILROAD CAR MONITORING

TOLL COLLECTION SYSTEMS

VEHICLE IDENTIFICATION

normalreader

(10cm / 3m)

maliciousreader

(50cm / 15m)

eavesdropon tag(???)

Read range?

eavesdropon reader

(50m / ???)

Asymmetric Channels

Reader Tag Eavesdropper

Forward Channel Range (~100m)

Backward Channel Range (~5m)

Security Risks: Espionage

• Corporate Espionage:– Identify Valuable Items to Steal– Monitor Changes in Inventory

• Personal Privacy– Leaking of personal information

(prescriptions, brand of underwear, etc.).– Location privacy: Tracking the physical

location of individuals by their RFID tags.

Espionage Case Study

• The US Food and Drug Administration (FDA) recently recommended tagging prescription drugs with RFID “pedigrees”.

• Problems:– “I’m Oxycontin. Steal me.”– “Bob’s Viagra sales are really up this month.”– “Hi. I’m Alice’s anti-fungal cream.”

Security Risks: Forgery

• RFID casino chips, Mobil SpeedPass, EZ-Pass, FasTrak, prox cards, €500 banknotes, designer clothing.

• Skimming: Read your tag, make my own.

• Swapping: Replace real tags with decoys.

• Producing a basic RFID device is simple.

• A hobbyist could probably spoof most RFID devices in a weekend for under $50.

Security Risks: Forgery

• Mandel, Roach, and Winstein @ MIT• Took a “couple weeks” and $30 to figure out how

produce a proximity card emulator.• Can produce fake cards for a few dollars.• Can copy arbitrary data, including TechCash.• Could read cards from several feet.

(My card won’t open the door past a few inches.)• Broke Indala's FlexSecur “data encryption”.

(Just addition and bit shuffling. Doh.)

Security Risks: Sabotage

• If we can’t eavesdrop or forge valid tags, can simply attack the RFID infrastructure.

• Wiping out inventory data.

• Vandalization.

• Interrupting supply chains.

• Seeding fake tags – difficult to remove.

Adversarial Model

• Can classify adversaries by their access.

• Three levels of read or write access:– Physical: Direct access to physical bits.– Logical: Send or receive coherent messages. – Signal: Detect traffic or broadcast noise.

• Can further break down into Forward-only or Backward-only access.

Adversarial Model: Attacks

• Long-Range Passive Eavesdropper: – Forward-Only Logical Read Access.– No Write Access.

• Tag Manufacture/Cloning:– No Read Access/Physical Read Access.– Physical Write Access.

• Traffic Analysis: Signal Read Access.

• Jamming: Signal Write Access.

Adversarial Model: Countermeasures

• Countermeasures will degrade an adversary’s access. For example:

• Encryption degrades logical read access to signal read access.

• Authentication degrades logical write to signal write access.

• Tamper resistance can degrade physical read to logical read access.

Is it really that bad?

• Maybe Not. • Tags can only be read from a few meters.*• Will mostly be used in closed systems like

warehouses or shipping terminals.• Can already track many consumer purchases

through credit cards.• Difficult to read some tags near liquids or metals.• Can already track people by cell phones,

wireless MAC addresses, CCTV cameras, etc.

But…the customer is always right.

• The public perception of a security risk, whether valid or not, could limit adoption and success.

• Similar to Pentium III’s unique ID numbers.• Successful boycott of Benetton. • Privacy advocates have latched on:

– “…e-mails sent to the RFID Journal…hint at some of the concerns. ‘I'll grow a beard and f--k Gillette,’ wrote one reader”, Economist Magazine, June 2003.

– “Auto-ID: The worst thing that ever happened to consumer privacy”, CASPIAN website.

Digression #2:RFID Public Relations

• The industry never misses a chance to shoot itself in the foot.

• “Track anything, anywhere”.

• “Wal-Mart Caught Conducting Secret Human Trials Using Alien Technology!”

• Lesson: If you don’t want people to negatively spin your technology, don’t make their jobs easier.

Security Challenge

• Resources, resources, resources.

• EPC tags ~ 5 cents. 1000 gates ~ 1 cent.

• Main security challenges come from resource constraints.

• Gate count, memory, storage, power, time, bandwidth, performance, die space, and physical size are all tightly constrained.

• Pervasiveness also makes security hard.

Example Tag SpecificationStorage 128-512 bits of read-only storage.

Memory 32-128 bits of volatile read-write memory.

Gate Count 1000-10000 gates equivalents.

Security Gate Budget 200-2000 gate equivalents.

Operating Frequency UHF 868-956 MHz.

Forward Range 100 meters.

Backward Range 3 meters.

Read Performance 100 read operations per second.

Cycles per Read 10,000 clock cycles.

Tag Power Source Passively powered via RF signal.

Power Consumption per Read

10 μWatts

Features Anti-Collision SupportRandom Number Generator

Resource Constraints

• With these constraints, modular math based public-key algorithms like RSA or ElGamal are much too expensive.

• Alternative public-key cryptosystems like ECC, NTRU, or XTR are too expensive.

• Symmetric encryption is also too costly. We can’t fit DES, AES, or SHA-1 in 2000 gates.

• (Recent progress made with AES.)

Hash Locks

• Rivest, Weis, Sarma, Engels (2003).

• Access control mechanism: – Authenticates readers to tags.

• “Only” requires OW hash function on tag.

• Lock tags with a one-way hash output.

• Unlock tags with the hash pre-image.

• Old idea, new application.

Hash Lock Access Control

Reader Tag

metaID ← hash(key)

Store (key,metaID)

metaID

Store metaID

Locking a tag(metaID)

Querying a locked tag

Unlocking a tag(Who are you?)

keymetaID = hash(key)?

“Hi, my name is..”

Hash Lock Analysis

+ Cheap to implement on tags: A hash function and storage for metaID.

+ Security based on hardness of hash. + Hash output has nice random properties.+ Low key look-up overhead.- Tags respond predictably; allows tracking.

Motivates randomization.

Randomized Hash Lock

Reader Tag: IDk

Knows tag ID1,…, IDn

R,hash(R, IDk)

Query?

Select random R

Unlocking a tag

IDk

Search hash(R, IDi)

Randomized Hash Lock Analysis

+ Implementation requires hash and random number generator

• Low-cost PRNG.• Physical randomness.

+ Randomized response prevents tracking.- Inefficient brute force key look-up.- Hash is only guaranteed to be one-way.

Might leak information about the ID. (Essentially end up with a block cipher?)

Blocker Tags

• Juels, Rivest, Szydlo (2003).

• Consumer Privacy Protecting Device: – Hides your tag data from strangers.

• Users carry a “blocker tag” device.

• Blocker tag injects itself into the tag’s anti-collision protocol.

• Effectively spoofs non-existent tags.

• (Only exists on paper.)

Other Work

• Efficient Implementations for RFID:– Feldhofer, Dominikus, and Wolkerstorfer.– Gaubatz, Kaps, and Yüksel.

• Secure Protocols:– Ari Juels.– Inoue and Yasuura – Gildas Avoine.

• Privacy Issues:– Molnar and Wagner.– Henrici and Müller.

Limited Bibliography:

crypto.csail.mit.edu/~sweis/rfid/

RFID Policy

• Policy can address a lot of privacy issues.• RSA Security is proposing a “privacy bit”:

– Sort of like a “do not disturb” sign. – Doesn’t stop someone from reading a tag.– More bits could encode various access policies

• Garfinkel has proposed an RFID Bill of Rights. • Other fair information practices proposed by

EPIC, EFF, CASPIAN, etc.

Simson’s Bill of Rights

• The RFID Bill of Rights:1)The right to know whether products contain

RFID tags.2)The right to have RFID tags removed or

deactivated when they purchase products.3)The right to use RFID-enabled services

without RFID tags.4)The right to access an RFID tag’s stored data. 5)The right to know when, where and why the

tags are being read.

A New Idea: Humans and Tags

• Tags are dumb. But so are people.

• Hopper and Blum have human-oriented identification protocols that you can do in your head. Linked off www.captcha.net.

• Now adopting their protocol to RFID and securing it against stronger adversaries.

• (Papers in progress.)

Simple trick:Defeating eavesdropping on forward link

rm r

“go ahead”

wants tosend m

picksrandom r

Appears in EPC Gen II standards.

A first attempt at defeatingeavesdropping and unauthorized tag-reading

Ek(r, ID)

kk“pseudonym”

Problem: All tags and readers share the same key k• If any tag is compromised, all security is lost• If any reader is compromised, all security is lost

Risk: Massive data spills.

Take #2: Independently keyed tags

r, Fki(r)

Scans throughall keys to decode

ki

“pseudonym”

Problem: Doesn’t scale.• Takes O(N) work to decode each pseudonym

(k1, ID1) :(kN, IDN)

Private identification protocols

Goal: a tag <-> reader protocol, providing:• Identification: Authorized reader learns tag’s identity• Privacy: Unauthorized readers learn nothing

• Attacker cannot even link two sightings of same tag

• Authentication: Tag identity cannot be spoofed• Scalability: Can be used with many tags

A non-trivial technical challenge,with many possible applications.

A beautiful method for private identification

r, Fki(r), Fkij

(r)

ki, kij

pseudonym

More scalable: O(√N) work to decode each pseudonym• First, scan all ki to learn i• Then, scan all kij to learn j and thus tag identity

:(ki, i) :(i, kij, IDij) :

Decodes i, then j

The tree of secrets

Tag leaf of the tree.Each tag receives the keys on path from leaf to the root.Tag ij generates pseudonyms as (r, Fki

(r), Fkij(r)).

Reader can decode pseudonym using a depth-first search.

k0

k00 k01

k0

k00 k01

k1

k10 k11

Analysis: tree of secrets

Generalizations:• Use any depth tree (e.g., lg N)• Use any branching factor (e.g., 210)• Use any other identification scheme (e.g., mutual auth)

Theory A concrete exampleNumber of tags: N 220 tagsTag storage: O(lg N) 128 bitsTag work: O(lg N) 2 PRF invocationsCommunications: O(lg N) 138 bitsReader work: O(lg N) 2 210 PRF invocations

Privacy degrades gracefully if tags are compromised

Reducing trust in readers

r, Fki(r), Fkij

(r)

ki, kij

If readers are online, Trusted Center can do decoding for them, and enforce a privacy policy for each tag.No keys stored at reader => less chance of privacy spills.

TrustedCenter

r, Fki(r), Fkij

(r)

IDij

Reader (kij, Policyij)

Reducing trust: Delegation

r, Fki(r), Fkij

(r)

ki, kij

For offline or partially disconnected readers, can delegate power to decode pseudonyms for a single tag to designated readers.Reader workload: O(D) per pseudonym,where D = # of tags delegated to this reader.

TrustedCenter

IDij

kij

(kij, Policyij)

kij

Time-limited delegation

pseudonym

ctr, ki, kij

TrustedCenter

IDij, L, R

{keys}

Only good for decodingL-th through R-thpseudonyms from tag IDij

Even less trust: Reader gets access to the next 100 pseudonyms from this tag (say), and nothing more.

k0000

Enabling time-limited delegation

Use GGM at lower levels: (ks0, ks1) = G(ks)Tag uses leaves sequentially

Reader gets keys for a subset

k0

k00 k01

k0

k00 k01

k1

k10 k11

k000

k0001 k0010 k0011

k001

• Identification systems: an exciting research area• Privacy is central• Many non-trivial technical challenges, many opportunities for clever solutions• There’s still time to have an impact on deployments

• Research question: Private identification protocols• Tree schemes have useful properties• Can we do better? Can do without persistent state?

• Recent work: Controlling readers with Trusted Computing (to appear at WPES’05)

Conclusions