7.1 Discussion and Analysis of the Protection Mechanisms 144
7.1.3 Context-Aware Multifactor Authentication Scheme Based on Dynamic PIN:
This section discusses and analyses the Context-Aware Multifactor Authentication Scheme Based on Dynamic PIN presented in Chapter 6.
7.1.3.1 Explanatory example
The example follows the risk-based rules mechanism described in section 6.6.1. The mechanism is used to parameterise and dynamically generate an image-based challenge based on runtime contextual information and client device’s constraints.
Consider an authentication transaction where a user chooses to authenticate using a smartphone already registered with the authentication server and consider the following policies:
1. Client device constraint rule:
, 5 , 20
2. Contextual rule:
, , 0.5
Rule 1 adds a constraint to the size of the challenge ( 20) given the size of the smartphone’s screen; while rule 2 is contextual and associates a given time and location to a level of risk. These rules become applicable within the context of a given transaction.
3. Challenge rules:
a. 6, 20, , 0.5
b. 4, 20, , 0.5
c. 5, 20, , 0.7
157
a. 2, 0.5
As explained before, the level of assurance depends on the strength of the challenge (see 6.6.1.3) and on the number of additional authentication factors (see 6.4.1) required during an authentication. This corresponds to rules 3 (a, b and c) and 4, respectively. The challenge rules above associate the size of the challenge (i.e. p and q) and mode (i.e. ordered or unordered) to a given level of assurance. For the three rules above the assurance level value was calculated based on the analysis presented in section 6.6.1.3 and the number of combinations and permutations of Table 13 (see the case for rules 3a and 3b which provide same order of magnitude in the number of possible permutations; therefore, they were assigned the same level of assurance 0.5 in this example). The authentication factor rule associates a number of authentication factors (e.g. a pin and a token would be two factors) to a level of risk. Intuitively, number of factors and level of risk an inversely proportional.
The authentication process consists of the following main steps:
Assurance Policy Configuration Process
Step 1. The authentication sever verifies that policies (1) and (2) hold true and searches for applicable challenge rules (3) and authentication factor rules (4). In this case, policies (3a), (3b), and (3c) are applicable since p=20, the same as in policy (1), and the assurance level value is equal to or higher than the level of risk in policy (2). Policy (4) is also applicable since
is equal or less in value to that of policy (2).
Step 2. Since there are three applicable challenge rules, the authentication server selects the best one in terms of usability. In this case policy (3a) has the best usability as it defines unordered mode.
Step3. The authentication server defines the assurance level policy for the authentication transaction consisting of policies (3a) and (4), or:
5. 6, 20, , 0.5 ∧
2, 0.5
The authentication server generates the challenge based on policy (3a) and send it to the client device along with the RPS, and the IDs of the 2 authentication factors required by policy (4). In this case, suppose one authentication factor is the challenge itself and the other it is the IMEI and IMSI numbers of the smartphone.
158
Step 4. The client device retrieves the authentication token associated to the IMSI/IMEI of the device and retrieves information about the last successful authentication attempt (i.e. ); and uses this information to parameterise the unordered mode indexing policy:
6. ∑ ∑
∑ 240
Policy Evaluation Process
Step 5. The user responds to the challenge, in this case an image-grid, by recognising the secret images. The client device retrieves the required seeds including those associated to the user input and evaluates the indexing policy.
Step 6. Once evaluated, the indexing policy (6) is used to configure the cryptographic transformation function to be used in the system which is given by
Dynamic PIN generation and validation Process
Step 7. The user input in response to the challenge is transformed into the that is used as input along with the RPS value to the cryptographic transformation function
in order to produce the Dynamic PIN.
Step 8. The Dynamic PIN is sent to the authentication server for validation.
7.1.3.2 Abstract and executable policies
The proposed authentication system distinguishes the following policies: client device constraints rules (1), contextual rules (2), challenge rules (3a, 3b, and 3c), authentication factor rules (4), assurance level policy (5), and the indexing policy (6).
Policies (1), (2), (3a), (3b), (3c), (4), and (5) are abstract policies while policy (6) is executable.
Client device constraint rules (1) and contextual rules (2) are used to select applicable challenge rules (3) and authentication factor rules (4). Client device constraint rules provide parameters about the device that are used to determine the existing constraints that affect the characteristics of the challenge to be constructed and also the type of authentication factors
159
that may be available. Contextual rules (2) provide runtime information about the context of execution of the authentication transaction, in the running example location and time, and the associated level of risk.
Challenge rules (3) and authentication factor rules (4) take into account the level of risk and reflect the level of authentication assurance required for the given transaction while considering the trade-off usability vs. security. As described in the running example, first, the level of assurance required is determined and applicable policies are selected; and then based on the applicable policies a priority selection approach is used to determine the best rules in terms of usability. Here the result is the selection of policies (3a) and (4).
The assurance level policy ultimately defines how the challenge will be constructed and the authentication factors required to be presented by the client device as part of the authentication process.
The indexing policy (6) is the executable policy since it determines what cryptographic transformation function to use and how to configure it to produce the Dynamic PIN.
7.1.3.3 Protection mechanism and adaptation
In the proposed protection mechanism, security decisions are driven by the different policies in the system. The authentication system consists of four main functional processes that introduce dynamic adaptation: Assurance Policy Configuration Process (steps 1-3), Cryptographic Transformation Policy Configuration Process (step 4), Policy Evaluation (steps 5 and 6), and Dynamic PIN Generation and Validation Process (steps 7-8).
Figure 59 shows how these three processes fit into the overall system. The greyed components indicate the components/processes that induce adaptation into the system.
The Assurance Policy Configuration Process occurs at the authentication server and consists of two sub-processes: Policy Selection Process and Policy Combination Process. Policy Selection process takes as input client device constraint rules (1) and contextual rules (2) and via a matching mechanism determines applicable challenge rules (3) and authentication factor rules (4). As explained before, policy selection takes into account contextual information of the execution and device constraints. The policy combination takes as input the applicable policies and based on risk level and usability combines them to produce a new executable policy, i.e. the assurance policy (5), that reflects the level of assurance required during authentication. The assurance policy introduces adaptation into the system as it defines the characteristics of the challenge produced at runtime. The challenge influences how the user
160
interacts with the system. It defines how many secret objects the user needs to recognise and the recognition mode, whether it is unordered or ordered. The authentication factors required also influence how the user interacts with the client device or how the client executes actions. For example, the user could be asked to provide a fingerprint as a biometric factor. In the running example, the IMEI/IMSI values force behaviour by the client device in order to provide the adequate token. The cognitive step between user and the client device is not predefined in any way and it entirely depends on the assurance policy generated dynamically at runtime that takes into account the abovementioned factors.
Figure 59 Dynamic PIN Authentication and adaptive decision-making process
The Cryptographic Transformation Policy Configuration Process takes place at the client device (see Figure 59) and consists of two sub-processes: the generation of required authentication tokens and fetching of execution state information such as the last successful authentication attempts, and the parameterisation of the indexing policy (6).
The parameterised indexing policy is the executable policy that is taken as input into the Policy Evaluation module (steps 5 and 6).
During policy evaluation, adaptation is introduced into the system via the indexing policy primarily due to the fact that this function depends on the user input and how he/she interacts with the challenge and the device. More precisely, depending on the user input the indexing policy will define a specific cryptographic transformation to produce the Dynamic PIN. In other words, the user input determines the cryptographic transformation that will reconfigure the system dynamically at runtime. The system will not know what cryptographic transformation to use except until the user provides its input and a parameterised indexing policy is defined. Client Device Cryptographic Transformation Policy Configuration Process Assurance Policy Configuration Process Dynamic PIN Validation Process Dynamic PIN Generation Process Authentication Server Policy Selection Process Policy Combination Process Context and device constraints Policy Evaluation
161
As shown in Figure 59, the Dynamic PIN Generation and Validation Process is the last functional process and it occurs at both sides: client device and authentication server, that is, Dynamic PIN Generation Process and Dynamic PIN Validation Process, respectively. These processes correspond to the enforcement aspect of the system, which is essentially the enforcement of the indexing policy (6) that if correctly parameterised will cause a valid Dynamic PIN to be produced and verified.
7.1.3.4 Specialisation of the protection mechanism
Specialisation of the authentication system takes place at both sides: authentication server and client device. The challenge is produced based on a specific set of images associated to seeds stored on the client device, the client device constraints, and available authentication factors. All these elements are considered in the usability vs. assurance trade-off aspect and as mentioned before they scope the new behaviour induced into the system. Recall that the challenge is derived from a generic model characterised by the values p and q. Specialisation is achieved by making the values p and q dependent on the client device that the user chooses to use in a given transaction and on the device’s unique characteristics. There are parameters specific to a given authentication transaction and device that scope the execution such and , which are here considered as execution state elements.
7.1.3.5 Security model: Enhancing adaptation using the indexing policy
Challenges are generated dynamically in order to suit a particular user, a particular device, and contextual information such as, for example, location and time. The combination of all these factors is taken into account in order to produce a challenge that provides the right level of authentication assurance given the implicit levels of risk and trust involved in the particular context of execution.
The indexing policy (6) used to configure the cryptographic transformation function that generates the dynamic PIN is in essence a security model that determines what crypto-function to use (out of the 240 possible options). The indexing policy is a modular component within the policy transformation process that ultimately determines what S-Box to apply. However, a different indexing policy (representing another model) could be used instead. Therefore, the indexing policy is ultimately a model that acts as an extension and supports security decisions within the policy transformation process. Recall that the purpose of the indexing policy is to increase the pseudo-randomness of the authentication scheme.
162