This thesis represented a research on “Security in Delay/Disruption Tolerant Networks”. Over the past few years, networks with characteristics of long delays, high disruptions, asymmetric data rates and/or low delivery ratio etc, have gained popularity. Different approaches have been researched in the past to improve performance of networks under these challenged conditions e.g. modifying TCP behaviour suitable for a selected set of networks including Performance Enhancing Proxies (PEPs) based satellite networks and by proposing complete new networking architecture such as Delay/Disruption Tolerant Networking (DTN).
The presence of PEPs on satellite links has disadvantages, e.g. splitting the TCP connection is not compliant with standard internet security mechanism IPsec as IPsec encrypts the traffic which can be only viewed at end nodes only. A new dynamic ML-IPsec protocol is discussed as a solution for TCP/IP based challenged networks. The ML-IPsec can solve the interworking issues between intermediate devices such as PEPs and IPsec. The new dynamic ML-IPsec can be applied to a vast domain of applications by making application more flexible to break IP datagram into different zones. It also provides the support to break down IP datagram up to 15 levels while degrading the performance of the system to some extent. The proposed method also improves the efficiency and reduces complexity to encapsulate the zones information into ESP payload. Some results of our analysis relating to security policy enforcement and network performance evaluation with different network bandwidth and traffic load are shown. It is observed that dynamic ML-IPsec gives almost same performance as IPsec performs when network bandwidth is more than 3 Mbps. However, when network bandwidth is low, then there is small performance reduction as compared to IPsec.
The other paradigm. Delay and Disruption Tolerant Networking (DTN), is an overlay networking architecture; evolved from a focus on deep space networks to a broader class of heterogeneous networks e.g. wireless adhoc networks, wireless local area networks etc. Lots of
A new ESKTS scheme provides a way to transport the symmetric key using public key cryptography. The symmetric key generated at a DTN node can be transported to other communicating body securely along with the data. The ESKTS is scalable, communication efficient and compliant with Bundle Security Protocol (BSP) semantics. It is observed via simulations that when network size is changed, the performance of ESKTS in presence of security is very close to SW performance without security. However, there is a 9.75% performance degradation in ESKTS with DSA security compared to SW without security performance in very densely deployed network configuration. This behaviour is almost the same as the behaviour observed in the traffic load analysis when the number of forwarding copies of SW routing protocol is increased. The performance degradation in optimal configurations of number of forwarding copies scenario was 5% and when all optimal configurations for different network sizes is simulated, the performance degradation has increased to 10% for very densely deployed networks. In summary, we can say that the simulation results have demonstrated that ESKTS is indeed a viable and lightweight solution for transporting the symmetric keys securely in DTN environment.
Standard PKI validation and revocation mechanism is enhanced by a new scheme, which enables the applications to build Certificate Revocation List (CRL) of reduced size. Furthermore it also increases the efficiency to search through the list while providing communication efficiency to distribute in the network due to its reduced size. It is observed via simulations that successful delivery ratio degrades with increase in CRL size. The decrease in the delivery ratio is due to the increase in traffic load a network can handle and computational processing to verify signatures and re-computing hashes. However, we can see that new CRL based revocation mechanism gives slightly better performance as compared to X.509 CRL based revocation method. The delivery ratio decreases more rapidly as the traffic load increases beyond a specific threshold (60,000, 80,000 number of revoked certificates) and before that threshold there is a gradual decrease in the delivery ratio when number of revoked certificates are less than 60,000. In summary, we can say that above simulation results have demonstrated that new proposed CRL based revocation mechanism is indeed a viable and lightweight solution for distributing CRLs in DTN environment while remaining in compliance with bundle security architecture. The new proposed revocation mechanism gives slightly better performance as compared X.509 CRL based revocation mechanism and besides that it remains in compliance with BSP specifications while X.509 based revocation mechanism will not be viable to work in compliance with BSP specifications due to the fact of verifying CA certificate via OCSP method.
New key management architecture is proposed to establish a shared state between communicating parties dynamically. The shared state defines the security services; the cryptographic algorithms and the keys. The new key management architecture and framework provides a way to build a shared state by establishing a secure channel between communicating parties by exchanging keys and negotiating cipher suites to use for communication. It also helps the communicating nodes to setup their security policy information dependent upon a shared secret state. A way to notify about certain events or to send error messages between communicating parties is also defined in the new key management architecture and framework.