You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
This repository has been archived by the owner on Mar 24, 2021. It is now read-only.
CacheSideChannel should have the facility to export data to a CSV file or similar, and we should have a Python script to visualize the timings. Some of our debugging and work has gone with print statement debugging and eyeballing for differences in timing distributions, but a histogram would work far better. The export/visualization should come in especially handy when experimenting with new timing side channels and new platforms and perfecting the data exfiltration.
(Also, it looks good in presentations and stuff.)
Ideally I'd like to export the timing for each run, and whether it was supposed to be a cache hit or not, so that we can show a histogram of cache hit timings and cache miss timings (with different colors for each). We should see this clear bimodal distribution, with one mode for cache hits and one mode for cache misses. We can even do things like draw a vertical line where our cutoff was for scoring something as a cache miss, which would show how good that measure is. (At time of writing I think the formula is median_latency - (median_latency - hit_latency) / 2)
The text was updated successfully, but these errors were encountered:
Sign up for freeto subscribe to this conversation on GitHub.
Already have an account?
Sign in.
CacheSideChannel
should have the facility to export data to a CSV file or similar, and we should have a Python script to visualize the timings. Some of our debugging and work has gone with print statement debugging and eyeballing for differences in timing distributions, but a histogram would work far better. The export/visualization should come in especially handy when experimenting with new timing side channels and new platforms and perfecting the data exfiltration.(Also, it looks good in presentations and stuff.)
Ideally I'd like to export the timing for each run, and whether it was supposed to be a cache hit or not, so that we can show a histogram of cache hit timings and cache miss timings (with different colors for each). We should see this clear bimodal distribution, with one mode for cache hits and one mode for cache misses. We can even do things like draw a vertical line where our cutoff was for scoring something as a cache miss, which would show how good that measure is. (At time of writing I think the formula is
median_latency - (median_latency - hit_latency) / 2
)The text was updated successfully, but these errors were encountered: