Paul J. Reber - US grants

Nondeclarative memory is a type of long-term memory that has been shown to support a variety of nonconscious cognitive skill learning tasks. Nondeclarative memory is generally observed as improvements in task performance without awareness of what was learned. Declarative memory, conscious memory for facts and events, is distinct from nondeclarative memory and is known to depend on the medial temporal lobe (MTL) memory system (including the hippocampus and adjacent cortical areas). Damage to the MTL due to acute injury or progressive disease (e.g., Alzheimer's disease) leads to impaired declarative memory but intact nondeclarative memory. For a number of nondeclarative memory tasks, it is not known what brain areas support learning or are involved in the expression of memory. The component cognitive processes that support nondeclarative memory are also poorly understood, particularly in the area of cognitive skill learning. The proposed program of research will use functional magnetic resonance imaging (MRI) to identify the neural substrates of three nondeclarative skill learning phenomena: category learning through prototype abstraction (Specific Aim 1), artificial grammar learning (Specific Aim 2) and perceptual-motor sequence learning (Specific Aim 3). In addition to identifying the brain areas that support memory in these tasks, stimulus-level contrasts will be used to identify the types of processing associated with successful task performance. Comparisons of patterns of evoked activity across the tasks are expected to identify processing differences between tasks that depend on cortically mediated facility-driven processing (e.g., dot-pattern categorization and priming) and tasks that depend on subcortical- cortical interactions (e.g., perceptual-motor sequence learning) which appear to operate through a retrieval mechanism (Specific Aim 4). In addition, declarative and nondeclarative memory will be directly compared across all three tasks to identify consistent differences in the expression of these memory types (Specific Aim 5). For each task, additional studies using transfer paradigms will examine brain activity associated with generalization of nondeclarative memory to novel stimuli, to test the hypothesis that cortical nondeclarative memory mechanisms support flexible use of memory while subcortical nondeclarative memory is relatively inflexible (Specific Aim 6). Overall, this program of research aims to build a more complete map of memory function throughout the brain while also identifying the component processes and operating characteristics of multiple types of nondeclarative memory.

Coercion attacks that compel an authorized user to reveal his or her secret authentication credentials can give attackers access to restricted systems. The PIs are developing a new approach to preventing coercion attacks using the concept of implicit learning from cognitive psychology. Implicit learning refers to learning of patterns without any conscious knowledge of the learned pattern. Using a carefully crafted keyboard-based computer game the PIs plant a secret password in the participant's brain without the participant having any conscious knowledge of the trained password. This planted secret can be used for authentication, but participants cannot be coerced into revealing their secret since they have no conscious knowledge of it.

This project explores three directions for using implicit learning in computer security. First, the PIs are developing implicit learning tasks designed to be used in challenge-response authentication. Second, the PIs are experimenting with methods to demonstrate implicit knowledge by measuring electrical activity along the scalp using off the shelf EEG devices. Third, the PIs are conducting user experiments to demonstrate that participants are able to properly authenticate, but cannot consciously recognize the trained secret. This project is a collaboration between computer security researchers and cognitive psychologists. Ultimately, the project aims to understand how the brain represents implicit knowledge. This in turn will lead to new coercion resistant security mechanisms for high-security applications.