@@ -1162,46 +1162,25 @@ <h3>Content Distribution Networks</h3>
1162
1162
< h3 > Decoy Values</ h3 >
1163
1163
1164
1164
< p >
1165
- [=Issuer=] use of decoy values in status lists has been explored as
1166
- a mechanism to increase the privacy of [=subjects=] by further hiding both
1167
- the size of the group associated with a status list and when values in the
1168
- list change. While algorithms for employing decoy values are out of scope for
1169
- this specification, implementers are advised to heed the guidance in this
1170
- section when considering the use of decoy values.
1165
+ [=Issuer=] use of decoy values in status lists has been explored as a mechanism
1166
+ to increase the privacy of [=subjects=]. While algorithms for employing decoy
1167
+ values are out of scope for this specification, implementers are advised that
1168
+ the use of decoy values do not provide privacy gains and can harm privacy in
1169
+ most cases.
1171
1170
</ p >
1172
1171
< p >
1173
- Decoys might help privacy in some cases, but harm privacy in others. For
1174
- example, if status list entry indexes are allocated in a random fashion,
1175
- adding decoys harms privacy because it reduces the group privacy size by the
1176
- number of decoys added to the group. A random allocation of indexes inherently
1177
- hides the true group size, ensuring that decoys are not necessary unless a bulk
1178
- operation, such as setting every credential in a list to the same status
1179
- (such as "revoked"), could reveal the true size of the group to an adversary
1180
- that also knew that the bulk update was to be applied to all non-decoy indexes.```
1172
+ When status list entry indexes are allocated in a random fashion, which is the
1173
+ suggested mode of operation for this specification, adding decoys harms privacy
1174
+ because it reduces the group privacy size by the number of decoys added to the
1175
+ group. A random allocation of indexes inherently hides the true group size,
1176
+ ensuring that decoys are not necessary.
1181
1177
</ p >
1182
1178
< p >
1183
- If decoys are used, the proper number of decoys to use is a function of
1184
- at least the desired group privacy size, the randomness of the
1185
- distribution of entries in the set, and whether entities watching the list
1186
- can determine which entries are real and which are decoys as they change
1187
- through time. These variables often change based on the use case and
1188
- implementers are encouraged to carefully evaluate whether decoys would
1189
- help or harm privacy for their particular deployment scenario.
1190
- </ p >
1191
- < p >
1192
- When using decoy values, it is important to ensure that the decoys
1193
- behave like real entries in the group. For example, if a decoy is used for
1194
- revocation, it would be strange to flip the decoy value to "unrevoked" when that
1195
- is not how the revocation values associated with real entries
1196
- behave. Similarly, changing the decoy values at times and to values that do not
1197
- mimic the behavior of real entries can allow the real entries to be separated
1198
- from the decoy entries through statistical analysis.
1199
- </ p >
1200
- < p >
1201
- The use of decoys is discouraged for most use cases, as random allocation
1202
- of status list entry indexes provides adequate protection.
1203
- Decoy are only to be considered in deployments where bulk
1204
- operations that would reveal the group size can be performed on a status list.
1179
+ There might be use cases where decoy values provide benefits. Implementers are
1180
+ cautioned that no such use cases were clearly identified by the group that
1181
+ created this specification. As a result, the use of decoys is discouraged for
1182
+ most use cases, as random allocation of status list entry indexes provides
1183
+ adequate protection.
1205
1184
</ p >
1206
1185
</ section >
1207
1186
0 commit comments