Interest-Based Access Control in CCN Cesar Ghali, Marc A. - - PowerPoint PPT Presentation

interest based access control in ccn
SMART_READER_LITE
LIVE PREVIEW

Interest-Based Access Control in CCN Cesar Ghali, Marc A. - - PowerPoint PPT Presentation

Interest-Based Access Control in CCN Cesar Ghali, Marc A. Schlosberg, Gene Tsudik, Christopher A. Wood Department of Computer Science University of California Irvine woodc1@uci.edu Confidential Agenda 1. Introduction and Access Control


slide-1
SLIDE 1

Confidential

Interest-Based Access Control in CCN

Cesar Ghali, Marc A. Schlosberg, Gene Tsudik, Christopher A. Wood

Department of Computer Science University of California Irvine woodc1@uci.edu

slide-2
SLIDE 2

Agenda

  • 1. Introduction and Access Control Overview
  • 2. IBAC Security Model
  • 3. IBAC via Name Obfuscation
  • 4. Security Considerations
  • 5. Experimental Assessment
  • 6. Conclusions and Recommendations

2

slide-3
SLIDE 3

CCN Elevator Pitch

  • Content is named and transferred through the network from

producers to consumers upon request

  • Any consumer can ask for content by name
  • Producers are responsible for access control

3

slide-4
SLIDE 4

Notation


 name of a Content Object content object with name
 
 set of consumers authorized to access content with name complement of the above group of consumers

4

U(N)

¯ U(N)

N

CO [N]

N

N

G

slide-5
SLIDE 5

The Access Control Problem

Question: How to ensure that only authorized users access a content

  • bject?
  • 1. Content-based: Ensure that only authorized consumers can

decrypt content they retrieve

  • 2. Interest-based: Ensure that consumers can only retrieve content

they are authorized to access

5

slide-6
SLIDE 6

Content-Based Access Control

Main Idea: If then should not be able to decrypt

  • A preliminary specification was first introduced in [1]
  • Many variations based on different public-key cryptographic algorithms have been

proposed (see [2]):

  • Broadcast encryption
  • Attribute-based encryption
  • Proxy-based re-encryption
  • … etc.

6

CO [N]

[1] Smetters, Diana, Philippe Golle, and Jim Thornton. CCNx Access Control Specifications. Technical report, PARC, 2010. [2] Kurihara, Jun, C. Wood, and Ersin Uzun. “An Encryption-Based Access Control Framework for Content-Centric Networking.” IFIP, 2015.

Cr / ∈ U(N)

Cr

slide-7
SLIDE 7

Content-Based AC in Pictures

7

Header [/a/b/c] Payload ValidationSection Content Object

slide-8
SLIDE 8

Content-Based AC in Pictures (cont’d)

8

Header [/a/b/c] Payload ValidationSection Content Object Header [/a/b/c] Payload ValidationSection Content Object Enc(k, payload)

slide-9
SLIDE 9

Content-Based AC in Pictures (cont’d)

9

Header [/a/b/c] Payload ValidationSection Content Object Header [/a/b/c] Payload ValidationSection Content Object Enc(k, payload) Header [/a/b/c/k] Payload ValidationSection Content Object

slide-10
SLIDE 10

Interest-Based Access Control

Main Idea: If then should not be able to construct a correct interest for Implication: Interest names should depend on a secret that only authorized consumers know

10

CO [N]

Cr / ∈ U(N)

Cr

slide-11
SLIDE 11

Interest-Based AC in Pictures

11

Header [/a/b/c] Payload ValidationSection Content Object

slide-12
SLIDE 12

Interest-Based AC in Pictures (cont’d)

12

Header [/a/b/c] Payload ValidationSection Content Object Header [/a/b/c] Payload ValidationSection Content Object Enc(k, name)

  • bfuscated

suffix

slide-13
SLIDE 13

… why not do both?

13

slide-14
SLIDE 14

Two Dimensions of AC

14

Header [/a/b/c] Payload ValidationSection Content Object Header [/a/b/c] Payload ValidationSection Content Object Enc(k2, payload) Header [/a/b/c/k] Payload ValidationSection Content Object Header [/a/b/c]

  • bfuscated

suffix Header [/a/b/c/k]

  • bfuscated

suffix Enc(k1, name)

slide-15
SLIDE 15

Security Model

IBAC is about obfuscating the name (the payload may also be encrypted…) Security means: an adversary without the (group) secret cannot generate the same obfuscated name Let be the set of all routers on the path between and Assume who can deploy and compromise any unauthorized consumer or any router

15

Path(Cr, P)

Cr

P

R / ∈ Path(Cr, P)

Adv

  • On-path adversaries can see the names in interest and content
  • …will consider this later
slide-16
SLIDE 16

IBAC via Name Obfuscation

The goal of IBAC is to make the name of a content object available under the name for some obfuscation function At least two ways to do this:

  • Encryption-based
  • Hash-based

Note: the obfuscation function only masks the suffix of a name — not the routable prefix

16

N

N 0 = f(N)

f

slide-17
SLIDE 17

Encryption-based Obfuscation

where is the private key associated with an authorized

17

N 0 = Enc(k, N)

k

Cr

slide-18
SLIDE 18

Supporting Multiple Groups

Question 1: What if we want group-based access control, i.e., where consumers in the same group generate the same obfuscated name? (One) Answer: Consumers in group share the encryption key

18

kG

G

slide-19
SLIDE 19

Supporting Multiple Groups

Question 1: What if we want group-based access control, i.e., where consumers in the same group generate the same obfuscated name? (One) Answer: Consumers in group share the encryption key Question 2: How does a producer identify the correct decryption key for content? (One) Answer: Include the group identifier in the payload of each interest, e.g.,

19

kG

G

IDG = H(kG)

P

slide-20
SLIDE 20

Supporting Multiple Groups (cont’d)

Question 3: How to prevent likability of multiple interests with the same ? (One) Answer: Encrypt the identifiers using the publisher’s public key

20

IDG

pkP

IDG = Enc(pkP , H(kG))

P

slide-21
SLIDE 21

Hash-based Obfuscation

where is the same shared group key


21

N 0 = H(k, N)

k

slide-22
SLIDE 22

Hash-based Obfuscation

where is the same shared group key Introduces more state since a producer must be able to invert to discover

22

N 0 = H(k, N)

k

H

N

slide-23
SLIDE 23

What about on-path attackers?…

23

slide-24
SLIDE 24

Replay Attacks

24

C R P 1) issue interest I for IBAC-protected content with name N 2) Cache IBAC-protected content CO[N] 3) Consume content CO[N] I[N] I[N] CO[N] CO[N]

slide-25
SLIDE 25

Replay Attacks (cont’d)

25

slide-26
SLIDE 26

Replay Attacks (cont’d)

26

slide-27
SLIDE 27

Replay Attacks in Detail

Any (on-path) adversary can observe an obfuscated interest, replay it, and get the same content Replay prevention:

  • Nonces and timestamps help prevent replays
  • … in addition to consumer authentication information

27

Payload = ⇣ IDG, r, t, σ = Signsks

G (N 0||IDG||r||t)

slide-28
SLIDE 28

Interest Authentication

Question: How can a router check if a given (cached) content object should be returned in response to an interest? Answer: Verify an authenticator in interests (e.g., a digital signature)

28

slide-29
SLIDE 29

Interest Authentication

Question: How can a router check if a given (cached) content object should be returned in response to an interest? Answer: Verify an authenticator in interests (e.g., a digital signature) Question: How does a router know what key(s) to use for verification? Answer: Follow the authorized content key binding (ACKB) rule:

29

ACKB: Cached content protected under IBAC must reflect the verification key associated with the authorization policy.

slide-30
SLIDE 30

ACKB in Action (Part 1)

30

P

verification key

slide-31
SLIDE 31

ACKB in Action (Part 2)

31

slide-32
SLIDE 32

Verification Procedure

32

slide-33
SLIDE 33

Handling Policy Changes

  • Policy changes include adding and removing users from groups
  • Adding users to groups is easy (give them the right key)
  • Removing users is hard:
  • Generate and distribute new group keys
  • Cached content may still exist in the network

33

slide-34
SLIDE 34

Experimental Assessment

  • Without authorization checks,

routers incur no added overhead

  • With authorization checks, routers

must:

  • Manage and verify nonces and

timestamps

  • Verify interest signatures (!)

34

slide-35
SLIDE 35

DoS Issues

35

Fraction I

slide-36
SLIDE 36

Recommendations

  • If replay attacks are not a concern, consumers use name
  • bfuscation and include their group identity in interests.
  • Otherwise, name obfuscation must be used and authorization

information must be included in interests.

  • If replay attacks are plausible but name privacy is not a concern,

authorization information is sufficient.

36

slide-37
SLIDE 37

Conclusion

  • 1. Motivated content- and interest-based access control
  • 2. Two ways to enforce IBAC
  • 3. One way to handle replay attacks
  • 4. Experimental assessment
  • 5. Recommendations for using IBAC

37

slide-38
SLIDE 38

38

Questions?…