1 files changed, 1 insertions, 0 deletions
diff --git a/attacksurface.md b/attacksurface.md
index 4a4ab37a..7eefb677 100644
--- a/attacksurface.md
+++ b/attacksurface.md
@@ -37,6 +37,7 @@ The UI is an unprivileged process running as the ordinary user for each user who
 
   - The manager service (above) calls `GetSecurityInfo(ATTRIBUTE_SECURITY_INFORMATION|LABEL_SECURITY_INFORMATION|SCOPE_SECURITY_INFORMATION|OWNER_SECURITY_INFORMATION|GROUP_SECURITY_INFORMATION|DACL_SECURITY_INFORMATION)` on itself. Then it examines all of the groups associated with the users token, and finds the first one that is of type `SE_GROUP_LOGON_ID`. It adds to the prior DACL one for this SID for only `PROCESS_QUERY_LIMITED_INFORMATION` permission. Then it passes the returned security attributes as the UI's process and thread attributes in the call to `CreateProcessAsUser`. This means that debugging the UI process (in order for another process to steal its handles and exfiltrate private keys) is only possible by processes that can debug the manager service.
   - Right now, we're removing the SACL/integrity level from the above, which means the UI process only requires medium integrity to access it. Requiring greater integrity level access prevents the process from running properly, unfortunately (`ERROR_PRIVILEGE_NOT_HELD`). It would be nice to require high integrity to open the process, without having to increase the privileges that the process has in its own token.
+    - Actually, for the time being, we're giving the UI a high integrity token. This is a bummer but seems unavoidable.
   - Perhaps due to the above and other reasons, it appears that it is possible for other processes to write into the UI process's message loop, which is bad, and might defeat the purpose of the above. On the other hand, the permissions of the above are fairly strict (`O:BAG:SYD:(A;;0x1fffff;;;SY)(A;;0x121411;;;BA)(A;;0x1000;;;S-logonsid)`).
   - It renders highlighted config files to a msftedit.dll control, which typically is capable of all sorts of OLE and RTF nastiness that we make some attempt to avoid.