일 | 월 | 화 | 수 | 목 | 금 | 토 |
---|---|---|---|---|---|---|
1 | 2 | 3 | 4 | 5 | ||
6 | 7 | 8 | 9 | 10 | 11 | 12 |
13 | 14 | 15 | 16 | 17 | 18 | 19 |
20 | 21 | 22 | 23 | 24 | 25 | 26 |
27 | 28 | 29 | 30 | 31 |
- 캐시사이드채널
- cache side channel
- depth-wise convolution
- down-sampling
- clflush
- 의존성
- Logical address
- down sampling
- 프라임 프로브
- cache side-channel
- depthwise separable convolution
- 부채널 공격
- 캐시 사이드 채널
- side-channel
- dilated convolution
- 로스함수
- min pooling
- mean pooling
- out-of-order
- cpu
- depthwise convolution
- 프라임프로브
- depth-wise separable convolution
- PRIME+PROBE
- PRIME+ABORT
- PRMIE ABORT
- prime abort
- transient execution attack
- 캐시 사이드채널
- 사이드채널
- Today
- Total
컴공과블로그
Prime+Abort : A Timer-Free High-Precision L3 Cache Attack using Intel TSX 본문
Prime+Abort : A Timer-Free High-Precision L3 Cache Attack using Intel TSX
LYHyoung 2023. 11. 29. 00:07Disselkoen, Craig, et al. "{Prime+ Abort}: A {Timer-Free}{High-Precision} L3 Cache Attack using Intel {TSX}." 26th USENIX Security Symposium (USENIX Security 17). 2017.
https://www.usenix.org/conference/usenixsecurity17/technical-sessions/presentation/disselkoen
Prime+Abort: A Timer-Free High-Precision L3 Cache Attack using Intel TSX | USENIX
Open Access Media USENIX is committed to Open Access to the research presented at our events. Papers and proceedings are freely available to everyone once the event begins. Any video, audio, and/or slides that are posted after the event are also free and o
www.usenix.org
2017년 USENIX SECURITY에 실린 논문이다.
PRIME+ABORT란?
기존에 나온 FLUSH+RELOAD, FLUSH+FLUSH, PRIME+PROBE와 같은 Cache Side-channel 공격이다.
그러나 이 공격은 이전 공격들과 다르게 'Timer-free' 하다는 장점이 있다.
Timer-free하다는 장점을 알기 위해선 공격이 어떻게 이루어지는지 알아야하기에 차근차근 들어가보겠다.
공격 Transaction이 진행되는 것. 즉 공격자가 공격을 진행하기 위해선 PRIME+PROBE와 같이 cache set을 미리 준비시킵니다(prime 시키다).
그 후 victim이 cache set에 접근하게 된다면, Intel의 메모리 보호기법 중 하나인 TSX로 인하여 victim의 접근이 중단됩니다(Abort가 이루어짐)
Prime후 Abort가 일어날 때까지 기다리는 공격이기 때문에 이름도 Prime+Abort이다.
Main Contribution
1. Intel SGX를 이용한 Cache Side-channel Attack
이 공격의 주요한 기여점은 무엇일까? 우선 Intel SGX를 이용한 side-channel이라는 것이다.
Intel SGX는 특정한 코드나 데이터를 보호하는 기법으로 아무나 사용하지 못하도록 hardware 측면에서 보호하는 기법이다.
특정 이용자가 SGX를 사용하려고 하면 이용자의 프로세스를 중단시켜버린다.
Prime+Abort는 이러한 점을 사용하여 공격자가 지정한 Data를 Prime시켜두고 이를 SGX로 보호되게 한다. 그 후 다른 이용자(Victim)이 Abort를 시킨다면 Prime한 부분에 접근했음을 알 수 있다.
2. Timer를 사용하지 않음.
다른 Cache Side-channel(Flush+reload, Prime+Probe 등)은 Timer가 필요하다. (rdtsc)
하지만 Prime+Abort는 특정한 타이머를 사용하지 않고 Prime시킨 이후에 Abort가 이루어질 때까지 기다리기만 하면 되는 공격으로 Time을 측정할 필요가 없다.
Detecting Memory Access
위 사진은 Event들의 Detection을 Victim access에 따라 만든 그래프이다.
위의 사진을 통해 Prime+Probe보다 3배는 더 빠른 공격임을 확인할 수 있고, 그 이유는 Timer를 사용할 필요가 없기 때문이다. 그렇기에 Timer를 사용하지 않음으로 더 빠른 공격이 될 수도 있었다.