Modelling and verification of security requirements and stealthiness in security protocols

Singh, Rajiv Ranjan ORCID: 0000-0003-1808-3433 (2022). Modelling and verification of security requirements and stealthiness in security protocols. University of Birmingham. Ph.D.

Preview

Singh2022PhD.pdf
Text - Accepted Version
Available under License All rights reserved.
Download (2MB) | Preview

Abstract

Traditionally, formal methods are used to verify security guarantees of a system by proving that the system meets its desired specifications. These guarantees are achieved by verifying the system's security properties, in a formal setting, against its formal specifications. This includes, for example, proving the security properties of confidentiality and authentication, in an adversarial setting, by constructing a complete formal model of the protocol. Any counterexample to this proof implies an attack on the security property. All such proofs are usually based on an ordered set of actions, generated by the protocol execution, called a trace. Both the proofs and their counterexamples can be investigated further by analysing the behaviour of these protocol traces. The attack trace might either follow the standard behaviour as per protocol semantics or show deviation from it. In the latter case, however, it should be easy for an analyst to spot any attack based on its comparison from standard traces.

This thesis makes two key contributions: a novel methodology for verifying the security requirements of security protocols by only modelling the attacks against a protocol specification, and, secondly, a formal definition of ‘stealthiness’ in a protocol trace which is used to classify attacks on security protocols as either ‘stealthy’ or ‘non-stealthy’.

Our first novel proposal tests security properties and then verifies the security requirements of a protocol by modelling only a subset of interactions that constitute the attacks. Using this both time and effort saving methodology, without modelling the complete protocol specifications, we demonstrate the efficacy of our technique using real attacks on one of the world's most used protocols-WPA2. We show that the process of modelling the complete protocol specifications, for verifying security properties, can be simplified by modelling only a subset of protocol specifications needed to model a given attack. We establish the merit of our novel simplified approach by identifying the inadequacy of security properties apart from augmenting and verifying the new security properties, by modelling only the attacks versus the current practice of modelling the complete protocol which is a time and effort intensive process. We find that the current security requirements for WPA2, as stated in its specification, are insufficient to ensure security. We then propose a set of security properties to be augmented to the specification to stop these attacks. Further, our method also allows us to verify if the proposed additional security requirements, if enforced correctly, would be enough to stop attacks.

Second, we seek to verify the ‘stealthiness’ of protocol attacks by introducing a novel formal definition of a ‘stealthy’ trace. ‘Stealthy’ actions by a participating entity or an adversary in a protocol interaction are about camouflaging fraudulent actions as genuine ones by fine-tuning their actions to make it look like honest ones. In our model, protocols are annotated to indicate what each party will log about each communication. Given a particular logging strategy, our framework determines whether it is possible to find an attack that produces log entries indistinguishable from normal runs of the protocol, or if any attack can be detected from the log entries alone. We present an intuitive definition of when an attack is ‘stealthy’, which cannot be automatically checked directly, with regard to some logging strategy. Next, we introduce session IDs to identify unique sessions. We show that our initial intuitive definition is equivalent to a second definition using these session IDs, which can also be tested automatically in TAMARIN. We analyse various attacks on known vulnerable protocols to see, for a range of logging strategies, which can be made into stealth attacks, and which cannot. This approach compares the stealthiness of various known attacks against a range of logging strategies.

Type of Work:

Thesis (Doctorates > Ph.D.)

Award Type:

Doctorates > Ph.D.

Supervisor(s):