Why Some Like It Loud: Timing Power Attacks in Multi-tenant Data - - PowerPoint PPT Presentation

▶
why some like it loud
SMART_READER_LITE
LIVE PREVIEW

Why Some Like It Loud: Timing Power Attacks in Multi-tenant Data - - PowerPoint PPT Presentation

Dec 10, 2022 130 likes •451 views

Why Some Like It Loud: Timing Power Attacks in Multi-tenant Data Centers Using an Acoustic Side Channel Mohammad A. Islam, Luting Yang, Kiran Ranganath, and Shaolei Ren Acknowledgement: This work was supported in part by NSF under grants

slide-1
SLIDE 1

Why Some Like It Loud:

Timing Power Attacks in Multi-tenant Data Centers Using an Acoustic Side Channel

Mohammad A. Islam, Luting Yang, Kiran Ranganath, and Shaolei Ren

Acknowledgement: This work was supported in part by NSF under grants CNS-1551661 and ECCS-1610471.

slide-2
SLIDE 2

2

This talk is NOT about multi-tenant clouds; it’s about multi-tenant data centers!

slide-3
SLIDE 3

2

This talk is NOT about multi-tenant clouds; it’s about multi-tenant data centers!

Tenant = virtual machines

UPS

P D U P D U

vs

Tenant = physical servers

slide-4
SLIDE 4

Multi-tenant data centers are everywhere…

3

Apple houses 25% of its servers in multi-tenant data centers…

slide-5
SLIDE 5

Multi-tenant data centers are everywhere…

3

Google, Amazon, MS, Fb… :7.8% Enterprise: 53% Multi-tenant: 37%

Percentage of electricity usage by data center type (source: NRDC 2015)

slide-6
SLIDE 6

4

Generator UPS ATS P D U P D U

An overview of multi-tenant data center

Managed by operator

Utility

slide-7
SLIDE 7

4

Generator UPS ATS P D U P D U

An overview of multi-tenant data center

Managed by operator Managed by tenants

Utility

slide-8
SLIDE 8

4

Generator UPS ATS P D U P D U

Securing the cyberspace

  • DDoS attack, network intrusion, privacy

protection, etc.

[Mirkovic, Sigcomm’04][Zhang CCS’12][Moon CCS’15][Dong CCS’17]…

An overview of multi-tenant data center

Managed by operator Managed by tenants

Utility

slide-9
SLIDE 9

5

Generator UPS ATS P D U P D U

slide-10
SLIDE 10

How to attack the physical infrastructure?

5

Generator UPS ATS P D U P D U

slide-11
SLIDE 11

How to attack the physical infrastructure?

5

Hacking control systems Human intrusion

Power

Overload using server power

Generator UPS ATS P D U P D U

slide-12
SLIDE 12

How to attack the physical infrastructure?

5

Hacking control systems Human intrusion

Power

Overload using server power

Generator UPS ATS P D U P D U

Our focus

slide-13
SLIDE 13

6

Generator UPS ATS P D U P D U

slide-14
SLIDE 14

6

Generator UPS ATS P D U P D U

slide-15
SLIDE 15

6

Generator UPS ATS P D U P D U

Power attack: Well-timed power injection to overload the shared data center capacity, subject to all applicable usage constraints set by the operator

Malicious Tenant

Malicious load

slide-16
SLIDE 16

Cost analysis

7

15.6 8.7 3.5 5 10 15 20 25 Tier-II Tier-III Tier-IV

Million $/MW/year

Estimated cost based on 5% overloads and a data center of 1MW-10,00sqft

More likely to have an outage during overloads (e.g., risk increases by ~280 times for a Tier-IV data center )

slide-17
SLIDE 17

Cost analysis

7

15.6 8.7 3.5 5 10 15 20 25 Tier-II Tier-III Tier-IV

Million $/MW/year

Estimated cost based on 5% overloads and a data center of 1MW-10,00sqft

Annual cost > $2 billion

(if only 10% of the U.S. data centers are affected)

More likely to have an outage during overloads (e.g., risk increases by ~280 times for a Tier-IV data center )

slide-18
SLIDE 18

How to precisely time power attacks?

8

slide-19
SLIDE 19

How to precisely time power attacks?

  • Random attacks are unlikely to be successful, while constant full power is prohibited

8

slide-20
SLIDE 20

How to precisely time power attacks?

  • Random attacks are unlikely to be successful, while constant full power is prohibited
  • Coarse timing (e.g., based on “peak” hours) is ineffective

8

slide-21
SLIDE 21

Server power  Heat  Cold Airflow  Fan Speed  Noise

9

Dell PowerEdge servers

slide-22
SLIDE 22

Server power  Heat  Cold Airflow  Fan Speed  Noise

9

Dell PowerEdge servers

slide-23
SLIDE 23

Server power  Heat  Cold Airflow  Fan Speed  Noise

9

Dell PowerEdge servers

slide-24
SLIDE 24

10

There are challenges…!

slide-25
SLIDE 25

Suppressing the loud AC noise

11

Serves in a data center Serves noise

slide-26
SLIDE 26

Suppressing the loud AC noise

11

Serves in a data center Serves noise

A high-pass filter reveals the server noise pattern

slide-27
SLIDE 27

12

𝒀(𝑶×𝑳) 𝑩(𝑵×𝑶) 𝒁(𝑵×𝑳) 𝑭(𝑵×𝑳)

Observation Interest Unknown

slide-28
SLIDE 28

12

𝒀(𝑶×𝑳) 𝑩(𝑵×𝑶) 𝒁(𝑵×𝑳) 𝑭(𝑵×𝑳)

Observation Interest Unknown Solution: Blind source separation using non-negative matrix factorization (NMF)

slide-29
SLIDE 29

Experimental evaluation

13

  • Experimental settings
  • Run real workload traces in a university data center
  • True positive: % of attack opportunities detected
  • Precision: % of an attack being successful
slide-30
SLIDE 30

Experimental evaluation

13

  • Experimental settings
  • Run real workload traces in a university data center
  • True positive: % of attack opportunities detected
  • Precision: % of an attack being successful
slide-31
SLIDE 31

14

Physical co-residence and space sharing result in physical side channels Can be exploited to compromise data center physical security!

Thanks!