SESSION SECURITY AND ALLIED TECHNOLOGIES Chair(s) TBA

6 Int'l Conf. Security and Management | SAM'11 |
are generated based on roles instead of individual users.
However the authorization decisions now have to be made
at OSDs instead of at the central metadata server. OSDs
also have to store the entire role-based ACL for each object,
which in additional introduces the possible access policy
synchronization among OSDs issue.
Other than capability-based access control, OSDs do not
provide data confidentiality protection at rest. Insider and
intrusion attacks incur high risks for sensitive data stored in
the clear on OSDs. Most of the systems use network layer
protocols to protect data in transit.
5. User Centric Data Protection Systems
In this section we examine two types of user centric
storage systems.
5.1 Cryptographic Storage System
Symmetric encryption is used in many systems, such as
SiRiUS[24], and CRUST [12]. SiRiUS provides file-level
encryption. File owners are responsible for file encryption,
key distribution and access policy specification. The key distribution
of SiRiUS does not scale well as a file encryption
key has to be wrapped to users’ public keys. Data integrity
protection is provided by using hash trees.
CRUST was designed to eliminate the key distribution
and scalability issues of SiRiUS. CRUST used the Leighton-
Micali key pre-distribution scheme [29]. However it requires
each user to share a long-term key encryption key with every
other user, which results in multiple key management issues.
Miller et al [34] developed a scheme for data secrecy and
integrity protection on NASD. They used a similar approach
to SiRiUS, but the encryption is at the data block level. An
encryption key is encrypted by each legitimate user’s public
key and stored in a key object associated with the file on
the metadata server.
Wrapping the encryption key into users’ public keys
creates a user access right revocation issue. There are three
possible solutions: 1) simply remove the user from the key
object (which wraps the encryption key in users’ public key
for encryption key distribution); however the user may still
cache the encryption key and be able to read the data; 2)
immediately re-encrypt the file with a new encryption key
and encrypt the new key with the public keys of those users
who should still have access to the file, which is slower, but
will ensure that the revoked user cannot access the file; 3)
apply the second solution lazily (lazy revocation); although
the revoked user continues to have access to the old data,
this prevents his/her access to any new data that is encrypted
with a different key. Integrity protection is achieved through
a non-linear check-sum of the unencrypted data which is
attached to the encrypted data.
Kher [27] and Storer el al. [42] provide detailed surveys
of other encryption file systems such as NCryptfs, Microsoft
EFS, Plutus, Cepheus, etc.
Recently, Attribute Based Encryption (ABE), a type of
asymmetric key encryption, has been applied to address finegrained
data access control and privacy protection. Introduced
by Sahai and Waters in [38], ABE extended Identity
Based Encryption (IBE) to design flexible and scalable
access control systems. There are two kinds of ABE: keypolicy
ABE (KP-ABE) [16] and ciphertext-policy ABE (CP-
ABE) [5] [7]. KP-ABE is a per-key based access control. In
KP-ABE, the ciphertext is associated with a set of attributes
and the secret key is associated with the access policy. The
encryptor defines the set of descriptive attributes necessary to
decrypt the ciphertext. The trusted authority who generates
user’s secret key defines the combination of attributes for
which the secret key can be used. In CP-ABE, the idea is
reversed: the ciphertext is associated with the access policy
and the encrypting party determines the policy under which
the data can be decrypted. The secret key is now associated
with a set of attributes. Therefore CP-ABE is a per-message
based access control. In order to address privacy of the
access control policy, anonymous ABE was introduced and
further improved by [36]. User accountability and illegal key
sharing are addressed in [31].
Secret sharing schemes provides data secrecy protection
without encrypting the data. Lakshmanan, et al. [28] proposed
a distributed store that uses secret sharing to provide
confidentiality at rest. Secret share replication is used to
improve performance and provide availability. A dissemination
protocol is used by servers to propagate new data
shares among replication servers. A share renewal protocol
is used to periodically generate new shares for long-term
confidentiality. The recoverability of secret sharing schemes
is leveraged by POTSHARDS [41] to protect data over
indefinitely long period of time. Approximate pointers in
conjunction with secure distributed RAID techniques are
used for availability and reliability. Other storage systems
such as PASIS, CleverSafe and GridSharing also use secret
sharing for data protection [42].
5.2 Cloud-based Storage
Cloud-based storage is gaining rapid interest. The uniqueness
of cloud storage over traditional or object-based storage
is its ability to leverage virtualization techniques to provide a
storage service composed of thousands of networked storage
devices, distributed file systems, and storage middleware.
This enables on-demand service, capacity, and management
to users anywhere via the Internet. There are many
cloud storage service providers, such as IBM, Amazon
(S3), Google (GFS), Microsoft (Azure), EMC (Atmos),
open source CloudStore, HDFS, etc. However the common
standards for cloud service are still in development.
In most existing cloud storage systems, data in transit is
protected through network layer security, such as SSL/TLS.
User access control is provided through authentication and
access control lists. Data privacy protection is not typically