1) For authenticated disk encryption, there are extra data such as IV or tag that need to be stored for all the alternative mods that are placed in Table 1. For these data, it will be necessary to use a sector other than the sector where the encrypted data is located. Consumed space by extra storage of tags or iv is relatively small problem. The main problem here is that you have to process at least two sectors to write or read data in a sector and this directly cuts performance in half or less. What would you recommend for the management of these tag and IV sectors?
For example, consider the structure of "| Data-1 || tag-1 || Data-2 || tag-2 ||…|| Data-n || tag-n |" where "|...|" represents a disk sector. When it is kept in the form, the storage area is directly halved but tag management is easy. On the other hand, if we use | Data-1 || Data-2 || Data-3 ||…. ||Data-n || tag-1, tag-2, tag-3… tag-n | form, the storage space required for tags is less in percentage while tag management is more complex. For instance, if we want to delete the data in the sector of "|Data-2|", it is sufficient to directly deleting two sectors in the first option -"|Data-2| and | tag-2 |"- In the second option it will be necessary to delete "|Data-2|" and update the tag sector. Could you explicate these -or related- methods from the aspect of OS? Which one would you prefer in what situation? Do you have any additional recommendations?
2) The problem has mentioned in the previous part also has another side: Let's encrypt the data and write the authentication tag to the other sector. If we want to read it later, how do we know that the data and tag have not been altered? We are faced with a requirement such as integrity of integrity. Alternative solutions were presented for this problem under the cathegory of incremental authentication schemes in this thesis that I read. How applicable are these solutions? When a sector is updated, it will be necessary to update more sectors than the previous cases. So, performance will suffer more severely than alternatives in the first part. By considering this security-performance trade-off, does it make sense to
concern about "integrity of integrity" and apply these methods?
