Hello from the Other Side: SSH over Robust Cache Covert ... · SSH over Robust Cache Covert...

Post on 22-Jun-2020

19 views 0 download

Transcript of Hello from the Other Side: SSH over Robust Cache Covert ... · SSH over Robust Cache Covert...

www.iaik.tugraz.at

Hello from the Other Side:SSH over Robust CacheCovert Channels in the CloudClémentine Maurice, Manuel Weber, Michael Schwarz, Lukas Giner,Daniel Gruss, Carlo Alberto Boano, Stefan Mangard, Kay RömerGraz University of Technology

February 2017 — NDSS 2017

1 / 25

www.iaik.tugraz.at

Outline

cache covert channels

how do we get a covert channel working in the cloud?

how do we get a covert channel working in a noisy environment?

what are the applications of such covert channel?

2 / 25

www.iaik.tugraz.at

CPU cache

main memory is slow compared to the CPU

caches buffer frequently used data

every data access goes through the cache

caches are transparent to the OS and the software

3 / 25

www.iaik.tugraz.at

CPU cache

main memory is slow compared to the CPU

caches buffer frequently used data

every data access goes through the cache

caches are transparent to the OS and the software

3 / 25

www.iaik.tugraz.at

CPU cache

main memory is slow compared to the CPU

caches buffer frequently used data

every data access goes through the cache

caches are transparent to the OS and the software

3 / 25

www.iaik.tugraz.at

CPU cache

main memory is slow compared to the CPU

caches buffer frequently used data

every data access goes through the cache

caches are transparent to the OS and the software

3 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Caches on Intel CPUs

core 0

L1

L2

core 1

L1

L2

core 2

L1

L2

core 3

L1

L2 ringbus

LLCslice 0

LLCslice 1

LLCslice 2

LLCslice 3

L1 and L2 are private

last-level cache

divided in slices

shared across cores

inclusive

hash function maps aphysical address to a slice

4 / 25

www.iaik.tugraz.at

Set-associative caches0 16 17 25 26 31

Index OffsetAddress

Cache

5 / 25

www.iaik.tugraz.at

Set-associative caches0 16 17 25 26 31

Index OffsetAddress

Cache

Cache set

Data loaded in a speci�c set depending on its address

5 / 25

www.iaik.tugraz.at

Set-associative caches0 16 17 25 26 31

Index OffsetAddress

Cache

Cache set

way 0 way 3

Data loaded in a speci�c set depending on its address

Several ways per set

5 / 25

www.iaik.tugraz.at

Set-associative caches0 16 17 25 26 31

Index OffsetAddress

Cache

Cache set

way 0 way 3

Cache line

Data loaded in a speci�c set depending on its address

Several ways per set

Cache line loaded in a speci�c way depending on the replacement policy

5 / 25

www.iaik.tugraz.at

Timing differences

50 100 150 200 250 300 350 400

101

104

107

Access time [CPU cycles]

Numberofaccesses

cache hits cache misses

6 / 25

www.iaik.tugraz.at

Cache-based covert channels

cache attacks→ exploit timing differences of memory accesses

covert channel: two processes communicating with each other

not allowed to do so, e.g., across VMs

literature: stops working with noise on the machine

solution? “Just use error-correcting codes”

7 / 25

www.iaik.tugraz.at

Cache-based covert channels

cache attacks→ exploit timing differences of memory accesses

covert channel: two processes communicating with each other

not allowed to do so, e.g., across VMs

literature: stops working with noise on the machine

solution? “Just use error-correcting codes”

7 / 25

www.iaik.tugraz.at

Cache-based covert channels

cache attacks→ exploit timing differences of memory accesses

covert channel: two processes communicating with each other

not allowed to do so, e.g., across VMs

literature: stops working with noise on the machine

solution? “Just use error-correcting codes”

7 / 25

www.iaik.tugraz.at

Cache-based covert channels

cache attacks→ exploit timing differences of memory accesses

covert channel: two processes communicating with each other

not allowed to do so, e.g., across VMs

literature: stops working with noise on the machine

solution? “Just use error-correcting codes”

7 / 25

www.iaik.tugraz.at

Prime+Probe

attacker knows which cache set the victim accessed, not the content

works across CPU cores as the last-level cache is shared

does not need shared memory, e.g., memory de-deduplication

→ works across VM in the cloud, e.g., on Amazon EC2

8 / 25

www.iaik.tugraz.at

Prime+Probe

attacker knows which cache set the victim accessed, not the content

works across CPU cores as the last-level cache is shared

does not need shared memory, e.g., memory de-deduplication

→ works across VM in the cloud, e.g., on Amazon EC2

8 / 25

www.iaik.tugraz.at

Prime+Probe

attacker knows which cache set the victim accessed, not the content

works across CPU cores as the last-level cache is shared

does not need shared memory, e.g., memory de-deduplication

→ works across VM in the cloud, e.g., on Amazon EC2

8 / 25

www.iaik.tugraz.at

Prime+Probe

attacker knows which cache set the victim accessed, not the content

works across CPU cores as the last-level cache is shared

does not need shared memory, e.g., memory de-deduplication

→ works across VM in the cloud, e.g., on Amazon EC2

8 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

9 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

Step 1: Attacker primes, i.e., �lls, the cache (no shared memory)

9 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

Step 1: Attacker primes, i.e., �lls, the cache (no shared memory)

Step 2: Victim evicts cache lines while running

loads data

9 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

Step 1: Attacker primes, i.e., �lls, the cache (no shared memory)

Step 2: Victim evicts cache lines while running

loads data

9 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

Step 1: Attacker primes, i.e., �lls, the cache (no shared memory)

Step 2: Victim evicts cache lines while running

Step 3: Attacker probes data to determine if set has been accessed

fast access

9 / 25

www.iaik.tugraz.at

Prime+Probe

Victim address space Cache Attacker address space

Step 1: Attacker primes, i.e., �lls, the cache (no shared memory)

Step 2: Victim evicts cache lines while running

Step 3: Attacker probes data to determine if set has been accessed

slow access

9 / 25

www.iaik.tugraz.at

Why can’t we just use error correcting codes?

1 0 0 1 1 0Sender

1 0 0 1 1 0Receiver

(a) Transmission without errors

1 0 0 1 1 0Sender

1 1 0 1 1 0Receiver

(b) Noise: substitution error

1 0 0 1 1 0Sender

1 0 0 0 0 0 1 1 0Receiver

(c) Sender descheduled: insertions

1 0 0 1 1 0Sender

1 0 0Receiver

(d) Receiver descheduled: deletions

10 / 25

www.iaik.tugraz.at

Why can’t we just use error correcting codes?

1 0 0 1 1 0Sender

1 0 0 1 1 0Receiver

(a) Transmission without errors

1 0 0 1 1 0Sender

1 1 0 1 1 0Receiver

(b) Noise: substitution error

1 0 0 1 1 0Sender

1 0 0 0 0 0 1 1 0Receiver

(c) Sender descheduled: insertions

1 0 0 1 1 0Sender

1 0 0Receiver

(d) Receiver descheduled: deletions

10 / 25

www.iaik.tugraz.at

Why can’t we just use error correcting codes?

1 0 0 1 1 0Sender

1 0 0 1 1 0Receiver

(a) Transmission without errors

1 0 0 1 1 0Sender

1 1 0 1 1 0Receiver

(b) Noise: substitution error

1 0 0 1 1 0Sender

1 0 0 0 0 0 1 1 0Receiver

(c) Sender descheduled: insertions

1 0 0 1 1 0Sender

1 0 0Receiver

(d) Receiver descheduled: deletions

10 / 25

www.iaik.tugraz.at

Why can’t we just use error correcting codes?

1 0 0 1 1 0Sender

1 0 0 1 1 0Receiver

(a) Transmission without errors

1 0 0 1 1 0Sender

1 1 0 1 1 0Receiver

(b) Noise: substitution error

1 0 0 1 1 0Sender

1 0 0 0 0 0 1 1 0Receiver

(c) Sender descheduled: insertions

1 0 0 1 1 0Sender

1 0 0Receiver

(d) Receiver descheduled: deletions

10 / 25

www.iaik.tugraz.at

Our robust covert channel

physical layer:

transmits words as a sequence of ‘0’s and ‘1’s

deals with synchronization errors

data-link layer:

divides data to transmit into packets

corrects the remaining errors

11 / 25

www.iaik.tugraz.at

Physical layer: Sending ‘0’s and ‘1’s

sender and receiver agree on one set

receiver probes the set continuously

sender transmits ’0’ doing nothing

→ lines of the receiver still in cache→ fast access

sender transmits ’1’ accessing addresses in the set

→ evicts lines of the receiver→ slow access

12 / 25

www.iaik.tugraz.at

Physical layer: Sending ‘0’s and ‘1’s

sender and receiver agree on one set

receiver probes the set continuously

sender transmits ’0’ doing nothing

→ lines of the receiver still in cache→ fast access

sender transmits ’1’ accessing addresses in the set

→ evicts lines of the receiver→ slow access

12 / 25

www.iaik.tugraz.at

Physical layer: Sending ‘0’s and ‘1’s

sender and receiver agree on one set

receiver probes the set continuously

sender transmits ’0’ doing nothing

→ lines of the receiver still in cache→ fast access

sender transmits ’1’ accessing addresses in the set

→ evicts lines of the receiver→ slow access

12 / 25

www.iaik.tugraz.at

Physical layer: Sending ‘0’s and ‘1’s

sender and receiver agree on one set

receiver probes the set continuously

sender transmits ’0’ doing nothing

→ lines of the receiver still in cache→ fast access

sender transmits ’1’ accessing addresses in the set

→ evicts lines of the receiver→ slow access

12 / 25

www.iaik.tugraz.at

Eviction set generation

need a set of addresses in the same cache set and same slice

problem: slice number depends on all bits of the physical address

physical address

cache tagcache setindex

cache lineoffset

2MB page offset

xxxx

we can build a set of addresses in the same cache set and same slice

without knowing which slice

13 / 25

www.iaik.tugraz.at

Eviction set generation

need a set of addresses in the same cache set and same slice

problem: slice number depends on all bits of the physical address

physical address

cache tagcache setindex

cache lineoffset

2MB page offset

xxxx

we can build a set of addresses in the same cache set and same slice

without knowing which slice

13 / 25

www.iaik.tugraz.at

Eviction set generation

need a set of addresses in the same cache set and same slice

problem: slice number depends on all bits of the physical address

physical address

cache tagcache setindex

cache lineoffset

2MB page offset

xxxx

we can build a set of addresses in the same cache set and same slice

without knowing which slice

13 / 25

www.iaik.tugraz.at

Eviction set generation

need a set of addresses in the same cache set and same slice

problem: slice number depends on all bits of the physical address

physical address

cache tagcache setindex

cache lineoffset

2MB page offset

xxxx

we can build a set of addresses in the same cache set and same slice

without knowing which slice

13 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 prime

#2#3#4

Cache Sets

S S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

S S S S S S S SR R R R R R R R

receivereviction sets

prime

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 probe

#2#3#4

Cache Sets

S S S S S S S SR R R R R R R R

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

S S S S S S S SR R R R R R R R

receivereviction sets

probe

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 prime

#2#3#4

Cache Sets

S S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

S S S S S S S S

R R R R R R R R

receivereviction sets

prime

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 probe

#2#3#4

Cache Sets

S S S S S S S S

R R R R R R R R

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

S S S S S S S S

R R R R R R R R

receivereviction sets

probe

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 prime

#2#3#4

Cache Sets

S S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache SetsR R R R R R R RS S S S S S S S

receivereviction sets

prime

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 probe

#2#3#4

Cache SetsR R R R R R R RS S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache SetsR R R R R R R RS S S S S S S S

receivereviction sets

probe

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 prime

#2#3#4

Cache Sets

S S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

R R R R R R R R

receivereviction sets

prime

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 probe

#2#3#4

Cache Sets

S S S S S S S S

receivereviction sets

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1#2#3#4

Cache Sets

R R R R R R R R

receivereviction sets

#1probe

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2#3#4

Cache Sets

receivereviction sets

#1

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2#3#4

Cache Sets

receivereviction sets

#1

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2#3#4

repeat!

receivereviction sets

#1

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2 3

#3#4

repeat!

receivereviction sets

#2#1

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2 3

#3 3

#4

repeat!

receivereviction sets

#3#2#1

14 / 25

www.iaik.tugraz.at

Jamming agreement

sendereviction sets

#1 3

#2 3

#3 3

#4 3

repeat!

receivereviction sets

#4#3#2#1

14 / 25

www.iaik.tugraz.at

Sending the �rst image

15 / 25

www.iaik.tugraz.at

Handling synchronization errors

deletion errors: request-to-send scheme that also serves as ack

3-bit sequence numberrequest: encoded sequence number (7 bits)

’0’-insertion errors: error detection code→ Berger codes

appending the number of ’0’s in the word to itself→ property: a word cannot consist solely of ’0’s

DataPhysical layer word

12 bits

SQN

3 bits

EDC

4 bits

16 / 25

www.iaik.tugraz.at

Handling synchronization errors

deletion errors: request-to-send scheme that also serves as ack

3-bit sequence numberrequest: encoded sequence number (7 bits)

’0’-insertion errors: error detection code→ Berger codes

appending the number of ’0’s in the word to itself→ property: a word cannot consist solely of ’0’s

DataPhysical layer word

12 bits

SQN

3 bits

EDC

4 bits

16 / 25

www.iaik.tugraz.at

Handling synchronization errors

deletion errors: request-to-send scheme that also serves as ack

3-bit sequence numberrequest: encoded sequence number (7 bits)

’0’-insertion errors: error detection code→ Berger codes

appending the number of ’0’s in the word to itself→ property: a word cannot consist solely of ’0’s

DataPhysical layer word

12 bits

SQN

3 bits

EDC

4 bits

16 / 25

www.iaik.tugraz.at

Synchronization (before)

17 / 25

www.iaik.tugraz.at

Synchronization (after)

18 / 25

www.iaik.tugraz.at

Synchronization (after)

18 / 25

www.iaik.tugraz.at

Synchronization (after)

18 / 25

www.iaik.tugraz.at

Data-link layer: Error correction

Reed-Solomon codes to correct the remaining errors

RS word size = physical layer word size = 12 bits

packet size = 212 − 1 = 4095 RS words

10% error-correcting code: 409 parity and 3686 data RS words

Data Parity

3686 RS-words 409 RS-words

Data SQN EDC

12 bits 3 bits 4 bits

Data-link layer packet

Physical layer word

19 / 25

www.iaik.tugraz.at

Data-link layer: Error correction

Reed-Solomon codes to correct the remaining errors

RS word size = physical layer word size = 12 bits

packet size = 212 − 1 = 4095 RS words

10% error-correcting code: 409 parity and 3686 data RS words

Data Parity

3686 RS-words 409 RS-words

Data SQN EDC

12 bits 3 bits 4 bits

Data-link layer packet

Physical layer word

19 / 25

www.iaik.tugraz.at

Error correction (after)

20 / 25

www.iaik.tugraz.at

Evaluation

Environment Bit rate Error rate Noise

Native 75.10KBps 0.00% –

Native 36.03KBps 0.00% stress -m 1Amazon EC2 45.25KBps 0.00% –Amazon EC2 45.09KBps 0.00% web server serving �les on sender VMAmazon EC2 42.96KBps 0.00% stress -m 2 on sender VMAmazon EC2 42.26KBps 0.00% stress -m 1 on receiver VMAmazon EC2 37.42KBps 0.00% web server on all 3 VMs, stress -m

4 on 3rd VM, stress -m 1 on senderand receiver VMs

Amazon EC2 34.27KBps 0.00% stress -m 8 on third VM

21 / 25

www.iaik.tugraz.at

Evaluation

Environment Bit rate Error rate Noise

Native 75.10KBps 0.00% –Native 36.03KBps 0.00% stress -m 1

Amazon EC2 45.25KBps 0.00% –Amazon EC2 45.09KBps 0.00% web server serving �les on sender VMAmazon EC2 42.96KBps 0.00% stress -m 2 on sender VMAmazon EC2 42.26KBps 0.00% stress -m 1 on receiver VMAmazon EC2 37.42KBps 0.00% web server on all 3 VMs, stress -m

4 on 3rd VM, stress -m 1 on senderand receiver VMs

Amazon EC2 34.27KBps 0.00% stress -m 8 on third VM

21 / 25

www.iaik.tugraz.at

Evaluation

Environment Bit rate Error rate Noise

Native 75.10KBps 0.00% –Native 36.03KBps 0.00% stress -m 1Amazon EC2 45.25KBps 0.00% –

Amazon EC2 45.09KBps 0.00% web server serving �les on sender VMAmazon EC2 42.96KBps 0.00% stress -m 2 on sender VMAmazon EC2 42.26KBps 0.00% stress -m 1 on receiver VMAmazon EC2 37.42KBps 0.00% web server on all 3 VMs, stress -m

4 on 3rd VM, stress -m 1 on senderand receiver VMs

Amazon EC2 34.27KBps 0.00% stress -m 8 on third VM

21 / 25

www.iaik.tugraz.at

Evaluation

Environment Bit rate Error rate Noise

Native 75.10KBps 0.00% –Native 36.03KBps 0.00% stress -m 1Amazon EC2 45.25KBps 0.00% –Amazon EC2 45.09KBps 0.00% web server serving �les on sender VMAmazon EC2 42.96KBps 0.00% stress -m 2 on sender VMAmazon EC2 42.26KBps 0.00% stress -m 1 on receiver VMAmazon EC2 37.42KBps 0.00% web server on all 3 VMs, stress -m

4 on 3rd VM, stress -m 1 on senderand receiver VMs

Amazon EC2 34.27KBps 0.00% stress -m 8 on third VM

21 / 25

www.iaik.tugraz.at

Building an SSH connection

Hypervisor

Last Level Cache (LLC)

VM 1

Covert Channel

Prime+Probe

TCP↔FileFile System

TCP Client(e.g. ssh)

Socket

VM 2

Covert Channel

Prime+Probe

TCP↔FileFile System

TCP Server(e.g. sshd)

Socket

22 / 25

www.iaik.tugraz.at

SSH evaluation

Between two instances on Amazon EC2

Noise ConnectionNo noise 3stress -m 8 on third VM 3Web server on third VM 3Web server on SSH server VM 3Web server on all VMs 3stress -m 1 on server side unstable

Telnet also works with occasional corrupted bytes with stress -m 1

23 / 25

www.iaik.tugraz.at

SSH evaluation

Between two instances on Amazon EC2

Noise ConnectionNo noise 3stress -m 8 on third VM 3Web server on third VM 3Web server on SSH server VM 3Web server on all VMs 3stress -m 1 on server side unstable

Telnet also works with occasional corrupted bytes with stress -m 1

23 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Conclusion

cache covert channels are practical

even in the cloud, even in presence of extraordinary noise

our robust covert channel supports an SSH connection

we extended Amazon’s product portfolio :)

24 / 25

www.iaik.tugraz.at

Hello from the Other Side:SSH over Robust CacheCovert Channels in the CloudClémentine Maurice, Manuel Weber, Michael Schwarz, Lukas Giner,Daniel Gruss, Carlo Alberto Boano, Stefan Mangard, Kay RömerGraz University of Technology

February 2017 — NDSS 2017

25 / 25