Carolyn Talcott

Benjamin Lion, Farhad Arbab, Carolyn Talcott

We propose a component-based semantic model for Cyber-Physical Systems (CPSs) wherein the notion of a component abstracts the internal details of both cyber and physical processes, to expose a uniform semantic model of their externally observable behaviors expressed as sets of sequences of observations. We introduce algebraic operations on such sequences to model different kinds of component composition. These composition operators yield the externally observable behavior of their resulting composite components through specifications of interactions of the behaviors of their constituent components, as they, e.g., synchronize with or mutually exclude each other's alternative behaviors. Our framework is expressive enough to allow articulation of properties that coordinate desired interactions among composed components within the framework, also as component behavior. We demonstrate the usefulness of our formalism through examples of coordination properties in a CPS consisting of two robots interacting through shared physical resources.

Deborah Shands, Carolyn Talcott

Many damaging cybersecurity attacks are enabled when an attacker can access residual sensitive information (e.g. cryptographic keys, personal identifiers) left behind from earlier computation. Attackers can sometimes use residual information to take control of a system, impersonate a user, or manipulate data. Current approaches to addressing access to residual sensitive information aim to patch individual software or hardware vulnerabilities. While such patching approaches are necessary to mitigate sometimes serious security vulnerabilities in the near term, they cannot address the underlying issue: explicit requirements for adequately eliminating residual information and explicit representations of the erasure capabilities of systems are necessary to ensure that sensitive information is handled as expected. This position paper introduces the concept of intentional forgetting and the capabilities that are needed to achieve it. Intentional forgetting enables software and hardware system designers at every level of abstraction to clearly specify and rigorously reason about the forgetting capabilities required of and provided by a system. We identify related work that may help to illuminate challenges or contribute to solutions and consider conceptual and engineering tradeoffs in implementations of forgetting capabilities. We discuss approaches to modeling intentional forgetting and then modeling the strength of a system's forgetting capability by its resistance to disclosing information to different types of detectors. Research is needed in a variety of domains to advance the theory, specification techniques, system foundations, implementation tools, and methodologies for effective, practical forgetting. We highlight research challenges in several domains and encourage cross-disciplinary collaboration to one day create a robust theory and practice of intentional forgetting.

Akos Vertes, Albert-Baskar Arul, Peter Avar et al.

Transcriptomics response of SK-N-AS cells to methamidophos (an acetylcholine esterase inhibitor) exposure was measured at 10 time points between 0.5 and 48 h. The data was analyzed using a combination of traditional statistical methods and novel machine learning algorithms for detecting anomalous behavior and infer causal relations between time profiles. We identified several processes that appeared to be upregulated in cells treated with methamidophos including: unfolded protein response, response to cAMP, calcium ion response, and cell-cell signaling. The data confirmed the expected consequence of acetylcholine buildup. In addition, transcripts with potentially key roles were identified and causal networks relating these transcripts were inferred using two different computational methods: Siamese convolutional networks and time warp causal inference. Two types of anomaly detection algorithms, one based on Autoencoders and the other one based on Generative Adversarial Networks (GANs), were applied to narrow down the set of relevant transcripts.

Mark-Oliver Stehr, Minyoung Kim, Carolyn L. Talcott et al.

In spite of the rapidly increasing number of applications of machine learning in various domains, a principled and systematic approach to the incorporation of domain knowledge in the engineering process is still lacking and ad hoc solutions that are difficult to validate are still the norm in practice, which is of growing concern not only in mission-critical applications. In this note, we introduce Probabilistic Approximate Logic (PALO) as a logic based on the notion of mean approximate probability to overcome conceptual and computational difficulties inherent to strictly probabilistic logics. The logic is approximate in several dimensions. Logical independence assumptions are used to obtain approximate probabilities, but by averaging over many instances of formulas a useful estimate of mean probability with known confidence can usually be obtained. To enable efficient computational inference, the logic has a continuous semantics that reflects only a subset of the structural properties of classical logic, but this imprecision can be partly compensated by richer theories obtained by classical inference or other means. Computational inference, which refers to the construction of models and validation of logical properties, is based on Stochastic Gradient Descent (SGD) and Markov Chain Monte Carlo (MCMC) techniques and hence another dimension where approximations are involved. We also present the Logical Imagination Engine (LIME), a prototypical implementation of PALO based on TensorFlow. Albeit not limited to the biological domain, we illustrate its operation in a quite substantial bioinformatics machine learning application concerned with network synthesis and analysis in a recent DARPA project.

Mark-Oliver Stehr, Peter Avar, Andrew R. Korte et al.

There is an abundance of complex dynamic systems that are critical to our daily lives and our society but that are hardly understood, and even with today's possibilities to sense and collect large amounts of experimental data, they are so complex and continuously evolving that it is unlikely that their dynamics will ever be understood in full detail. Nevertheless, through computational tools we can try to make the best possible use of the current technologies and available data. We believe that the most useful models will have to take into account the imbalance between system complexity and available data in the context of limited knowledge or multiple hypotheses. The complex system of biological cells is a prime example of such a system that is studied in systems biology and has motivated the methods presented in this paper. They were developed as part of the DARPA Rapid Threat Assessment (RTA) program, which is concerned with understanding of the mechanism of action (MoA) of toxins or drugs affecting human cells. Using a combination of Gaussian processes and abstract network modeling, we present three fundamentally different machine-learning-based approaches to learn causal relations and synthesize causal networks from high-dimensional time series data. While other types of data are available and have been analyzed and integrated in our RTA work, we focus on transcriptomics (that is gene expression) data obtained from high-throughput microarray experiments in this paper to illustrate capabilities and limitations of our algorithms. Our algorithms make different but overall relatively few biological assumptions, so that they are applicable to other types of biological data and potentially even to other complex systems that exhibit high dimensionality but are not of biological nature.

Vivek Nigam, Carolyn Talcott, Abraão Aires Urquiza

Timed Intruder Models have been proposed for the verification of Cyber-Physical Security Protocols (CPSP) amending the traditional Dolev-Yao intruder to obey the physical restrictions of the environment. Since to learn a message, a Timed Intruder needs to wait for a message to arrive, mounting an attack may depend on where Timed Intruders are. It may well be the case that in the presence of a great number of intruders there is no attack, but there is an attack in the presence of a small number of well placed intruders. Therefore, a major challenge for the automated verification of CPSP is to determine how many Timed Intruders to use and where should they be placed. This paper answers this question by showing it is enough to use the same number of Timed Intruders as the number of participants. We also report on some preliminary experimental results in discovering attacks in CPSP.