- Joined
- Jan 27, 2011
- Location
- Beautiful Sunny Winfield
This is a H/W problem where certain memory access patterns (hammering on the RAM?) causes bits to flip in unrelated memory locations. More info at https://googleprojectzero.blogspot.com/2015/03/exploiting-dram-rowhammer-bug-to-gain.html. This problem is H/W related but apparently can be mitigated but perhaps not entirely prevented by increasing DRAM refresh rates. DDR4 has rowhammer mitigation built in.
I found source code for testing this on Linux at https://github.com/google/rowhammer-test and have tested my laptop, desktop and file server and detected no problems with any. There is an APK (drammer.apk) to test this on Android and it caused my Nexus 5X to restart when run with the most aggressive setting.
I did not find a test for this when running Windows. I suppose that since it is a H/W bug that Windows would be just as susceptible but perhaps no exploit has yet been produced or it does exist but is too valuable to make public.
Aside from the security implications, it seems like this would also compromise stability though the access pattern required to trigger the problem may be vanishingly unlikely unless it is intentionally performed.
I found source code for testing this on Linux at https://github.com/google/rowhammer-test and have tested my laptop, desktop and file server and detected no problems with any. There is an APK (drammer.apk) to test this on Android and it caused my Nexus 5X to restart when run with the most aggressive setting.
I did not find a test for this when running Windows. I suppose that since it is a H/W bug that Windows would be just as susceptible but perhaps no exploit has yet been produced or it does exist but is too valuable to make public.
Aside from the security implications, it seems like this would also compromise stability though the access pattern required to trigger the problem may be vanishingly unlikely unless it is intentionally performed.
