Abstract The unsolved Andrews-Curtis conjecture in group theory gives rise to many famously challenging theorem-proving problems. Automated reasoning is one of the best state-of-the-art approaches to solving them. We have developed a new classifier inspired by group theory and a new search algorithm orientated at using for the Andrews-Curtis conjecture. We combine them with automated reasoning and compare their performance with automated reasoning.

Export control (EC) compliance is a critical yet labour-intensive process within research institutions, where the classification of sensitive technologies and cross-border disclosures often depends on expert interpretation of complex legal frameworks.This paper investigates the potential of large language models (LLMs), specifically in this study ChatGPT-4o and LLaMA-3.3, to support EC screening through a multistage, expert-in-the-loop framework.The methodology includes prompt variation, regulatory conditioning, reflective reasoning, and expert-informed evaluation to simulate real-world compliance workflows.Using a curated dataset of UK research project descriptions and the UK Strategic Export Control List, we assess model performance across over 1,400 outputs.Results show that while both models benefit from domain-specific grounding, ChatGPT-4o consistently produces more stable and interpretable classifications.Prompt sensitivity, bias behaviour, and ambiguity handling are also examined to highlight model limitations.The findings suggest that LLMs can support early stage EC assessment but require structured prompting and human oversight to ensure regulatory alignment.

Abstract We demonstrate how a generic automated theorem prover can be applied to establish the non-orderability of groups. Our approach incorporates various tools such as reasoning from the first principles, positive cones, torsions, generalised torsions and cofinal elements.

Parkinson’s disease (PD) is a neurodegenerative disorder characterized by tremors as its most typical symptom. Wearable accelerometer sensors, along with corresponding machine learning algorithms, can effectively assist in the diagnosis of PD tremors. However, due to the variations in disease progression and symptoms caused by individual differences among PD patients, it is challenging for existing algorithms to eliminate label noise and accurately identify and extract disease-related features across diverse patient data. In this study, we propose a Learnable Discriminative Instance Pool (LDIP) algorithm based on multi-instance learning, which integrates the concept of learnable shapelets. This method transforms the traditional DIP algorithm into a learnable instance pool that can be adaptively adjusted according to discriminative criteria, thereby enhancing the separability between different classes after bag mapping. We evaluated the proposed method on two clinical datasets using three different machine learning classifiers, achieving a maximum 73% accuracy for 5-class classification. The experimental results demonstrate that our proposed method consistently outperforms current baselines across various settings.

This paper proposes a novel multi-agent framework for penetration testing that aims to enable efficient and adaptive collaboration of specialised agents. This framework uses the Blackboard system for communication between Scout, Attack, and Analysis Agents. The combined efforts of these agents are aimed at assessing the security available on the networks, and a Decision-Making Agent (DMA) controls it all by making crucial decisions using collected information. This leads to a much better scalability and sophistication of security analyses, thus enhancing the penetration testing process. To guarantee the reliability and robustness of our framework, we use the Abstract State Machine (ASM) as a method to develop a formal model expressing this framework. The model developed is then validated and verified against some defined constraints and properties in order to demonstrate safety, free-deadlock, liveness, and reachability of the elaborated framework.

Untangling a braid is a typical multi-step process, and reinforcement learning can be used to train an agent to untangle braids. Here we present another approach. Starting from the untangled braid, we produce a dataset of braids using breadth-first search and then apply behavioral cloning to train an agent on the output of this search. As a result, the (inverses of) steps predicted by the agent turn out to be an unexpectedly good method of untangling braids, including those braids which did not feature in the dataset.

Recent research in computational topology found sets of counterexamples demonstrating that several recent mathematical articles purporting to describe a mathematical concept of realizable Gauss diagrams contain a mistake. In this study we propose several ways of encoding Gauss diagrams as binary matrices, and train several classical ML models to recognise whether a Gauss diagram is realizable or unrealizable. We test their accuracy in general, on the one hand, and on the counterexamples, on the other hand. Intriguingly, accuracy is good in general and surprisingly bad on the counterexamples. Thus, although human mathematicians and AI perceive Gauss diagrams completely differently, they tend to make the same mistake when describing realizable Gauss diagrams.

The shipping industry is an important component of the global trade and economy. In order to ensure law compliance and safety, it needs to be monitored. In this paper, we present a novel ship type classification model that combines vessel transmitted data from the Automatic Identification System, with vessel imagery. The main components of our approach are the Faster R-CNN Deep Neural Network and a Neuro-Fuzzy system with IF-THEN rules. We evaluate our model using real world data and showcase the advantages of this combination while also compare it with other methods. Results show that our model can increase prediction scores by up to 15.4% when compared with the next best model we considered, while also maintaining a level of explainability as opposed to common black box approaches.

While an abundance of real-time maritime information exists and is readily available to monitoring authorities, there are still many instances in which ships are found to be engaged in dangerous or illegal activities. In order to prevent such activities, authorities employ Vessel Traffic Services systems since they promote safety at sea while also assisting in management of ports. In this paper we report on research done in cooperation with Denbridge Marine Ltd., a global provider of maritime solutions, and present an application integrated in a Vessel Tracking Services system that allows the detection of normal vessel activity as well as dangerous or illegal situations in real-time, using information from the Automatic Identification System, a radar sensor and other information. We use a set of phenomena representing maritime activities of interest in the language of Phenesthe, our Complex Event Processing engine, and detect them on real maritime data streams from the area of Liverpool, United Kingdom. We evaluate our application and show that our system is capable of detecting and visualising maritime activities on the map in real time. Finally, we study and demonstrate the significance of using data from the Automatic Identification System along with radar data for maritime monitoring.

Several previous studies have focused on Distributed Denial of Service (DDoS) attacks, which are a crucial problem in computer network security.In this paper we explore the applicability of a a time series method known as a matrix profile to the anomaly based DDoS attacks detection.The study thus examined how the matrix profile method performed in diverse situations related to DDoS attacks, as well as identifying those features that are most applicable in various scenarios.Based on reported empirical evaluation the matrix profile method is shown to be efficient against most of the considered types of DDoS attacks.

Abstract We introduce a framework for the verification of protocols involving a distinguished machine (referred to as a leader) orchestrating the operation of an arbitrary number of identical machines (referred to as followers) in a network. At the core of our framework is a high-level formalism capturing the operation of these types of machines together with their network interactions. We show that this formalism automatically translates to a tractable form of first-order temporal logic. Checking whether a protocol specified in our formalism satisfies a desired property (expressible in temporal logic) then amounts to checking whether the protocol’s translation in first-order temporal logic entails that property. Many different types of protocols used in practice, such as cache coherence, atomic commitment, consensus, and synchronization protocols, fit within our framework. First-order temporal logic also facilitates parameterized verification by enabling us to model such protocols abstractly without referring to individual machines.

In this paper, we present a SDDB scheme regarding document-based store that satisfies three security requirements: confidentiality, flexible access control, and querying over encrypted data. The scheme is inspired by PIRATTE and CryptDB concepts. PIRATTE is a proxy for sharing encrypted files through a social network between the data owner and the number of users and the files are decrypted on user side with the proxy key, whereas in CryptDB, it is proxy between a database and one user to encrypt or decrypt data based on user’s queries. The scheme also improves CryptDB security and provides the possibility of sharing data with multi-users through PIRATTE concept which is used to verify authentication on the proxy side.

This work tries to evaluate the existing approaches used to benchmark the performance of machine learning models applied to network-based intrusion detection systems (NIDS). First, we demonstrate that we can reach a very high accuracy with most of the traditional machine learning and deep learning models by using the existing performance evaluation strategy. It just requires the right hyperparameter tuning to outperform the existing reported accuracy results in deep learning models. We further question the value of the existing evaluation methods in which the same datasets are used for training and testing the models. We are proposing the use of an alternative strategy that aims to evaluate the practicality and the performance of the models and datasets as well. In this approach, different datasets with compatible sets of features are used for training and testing. When we evaluate the models that we created with the proposed strategy, we demonstrate that the performance is very bad. Thus, models have no practical usage, and it performs based on a pure randomness. This research is important for security-based machine learning applications to re-think about the datasets and the model's quality.

Copyright © 2018 by SCITEPRESS – Science and Technology Publications, Lda. All rights reserved. We present an approach to execution of queries on encrypted graph databases. The approach is inspired by CryptDB system for relational DBs (R. A. Popa et al). Before processing a graph query is translated into encrypted form which then executed on a server without decrypting any data; the encrypted results are sent back to a client where they are finally decrypted. In this way data privacy is protected at the server side. We present the design of the system and empirical data obtained by experimentation with a prototype, implemented for Neo4j graph DBMS and Cypher query language, utilizing Java API. We report the efficiency of query execution for various types of queries on encrypted and non-encrypted Neo4j graph databases.

Traditional penetration testing relies on the domain expert knowledge and requires considerable human effort all of which incurs a high cost. In this paper, we propose an automated penetration testing approach based on the belief-desire-intention (BDI) agent model, which is central in the research on agent based processing in that it deals interactively with dynamic, uncertain and complex environments. Penetration testing actions are defined as a series of BDI plans and the BDI reasoning cycle is used to represent the penetration testing process. The model is extensible and new plans can be added, once they have been elicited from the human experts. We report on the results of testing of proof of concept BDI-based penetration testing tool in the simulated environment.

We present a verification methodology for analysing the decision-making component in agent-based hybrid systems. Traditionally hybrid automata have been used to both implement and verify such systems, but hybrid automata based modelling, programming and verification techniques scale poorly as the complexity of discrete decision-making increases making them unattractive in situations where complex logical reasoning is required. In the programming of complex systems it has, therefore, become common to separate out logical decision-making into a separate, discrete, component. However, verification techniques have failed to keep pace with this development. We are exploring agent-based logical components and have developed a model checking technique for such components which can then be composed with a separate analysis of the continuous part of the hybrid system. Among other things this allows program model checkers to be used to verify the actual implementation of the decision-making in hybrid autonomous systems.

Feature ordering is a significant data preprocessing method in incremental attribute learning (IAL), where features are gradually trained according to a given order. Previous research showed feature ordering is crucial to the IAL performance. It is relevant to each feature's discrimination ability, which can be calculated by single discriminability (SD). However, when feature dimensions increase, feature discrimination ability should also be calculated incrementally, because discrimination ability in lower dimensional spaces is different from that in higher spaces. Thus based on SD, accumulative discriminability (AD), a new statistical metric for incremental feature discrimination ability estimation, is designed. Moreover, a criterion that summarizes all the produced values of AD is employed to obtain the optimum feature ordering for classification problems based on neural networks by means of IAL. In addition, in order to reduce the time consumption, an effective feature ordering approach is developed. Compared with the feature ordering obtained by other approaches, the method outlined in this paper obtained good final classification results, which indicates that, firstly, feature discrimination ability should be incrementally estimated in IAL; and secondly, feature ordering derived by AD and its corresponding approaches are applicable with IAL.

In this paper we deal with verification of safety properties of term-rewriting systems. The verification problem is translated to a purely logical problem of finding a finite countermodel for a first-order formula, which is further resolved by a generic finite model finding procedure. A finite countermodel produced during successful verification provides with a concise description of the system invariant sufficient to demonstrate a specific safety property. We show the relative completeness of this approach with respect to the tree automata completion technique. On a set of examples taken from the literature we demonstrate the efficiency of finite model finding approach as well as its explanatory power.

Intrusion Detection Systems (IDS) aim to detect the actions that attempt to compromise the confidentiality, availability, and integrity of a resource by monitoring the events occurring in computer systems and/or networks. Stream data processing is a database technology applied to flows of data. Temporal Logic is a formalism for representing change over time. This paper proposes the development of a network intrusion detection system by combining temporal formalisms for representing attack patterns with stream processing for intruder detection. The experimental results show that this combination successfully was able to detect all the attacks of that type in the test data. Additionally, the solution provides a concise and unambiguous way to formally represent attack signatures and it is extensible and scalable.

The semantics of fixed point constructions is commonly defined in terms of iteration sequences. For example, the least fixed point of a monotone operator consists of all points which eventually appear in the approximations computed iteratively. We take this temporal reading as the starting point and develop a systematic approach to temporal definitions over iteration sequences. As a result, we propose an extension of first-order predicate logic with an iterative operator, in which iteration steps may be accessed by temporal logic formulae. We show that proposed logic FO+TAI subsumes virtually all known deterministic fixed point extentions of first-order logic as its natural fragments. On the other hand we show that over finite structures FO+TAI has the same expressive power as FO+PFP (FO with partial fixed point operator), but in many cases providing with more concise definitions. Finally, we show that the extension of modal mu-calculus with the temporal access leads to the more expressive logic closed under assume-guarantee specifications operator.

Querying its own history is an important mechanism in the computations, especially those interacting with people or other computations such as transaction processing, electronic data interchange. John McCarthy in his Elephant programming language proposal suggested exploiting the referring to the past as the main programming primitive. In this paper we study the computational power of such primitive. In order to do that we propose a refined formal model, History Dependent Machine (HDM), which uses querying the history as its sole computational primitive. Our main result may be spelled in general terms as: a model with a single agent wandering around a pool of resources and having ability to check its own history for simple temporal properties has a universal computational power. Moreover, HDM can simulate any multicounter machine in real time. Then we show that the computations of HDM may be specified in the extension of propositional linear temporal logic by flexible constants, the abstraction operator and equality. We use then universality of HDM model to show that the above extension with a single flexible constant is not recursively axiomatizable.

In this paper we consider the specification and verification of infinite-state systems using temporal logic. In particular, we describe parameterised systems using a new variety of first-order temporal logic that is both powerful enough for this form of specification and tractable enough for practical deductive verification. Importantly, the power of the temporal language allows us to describe (and verify) asynchronous systems, communication delays and more complex liveness and fairness properties. These aspects appear difficult for many other approaches to infinite-state verification.

Querying its own history is an important mechanism in the computations, especially those interacting with people or other computations such as transaction processing, electronic data interchange. In this paper we study the computational power of referring to the past primitive. To do that we propose a refined formal model, history dependent machine (RDM), which uses querying the history as its sole computational primitive. Our main result may be spelled in general terms as: a model with a single agent wandering around a pool of resources and having ability to check its own history for simple temporal properties has a universal computational power. Moreover, RDM can simulate any multicounter machine in real time. Then we show that the computations of RDM may be specified in the extension of propositional linear temporal logic by flexible constants, the abstraction operator and equality. We use then universality of RDM model to show that the above extension with a single flexible constant is not recursively axiomatizable

Supercompilation, or supervised compilation is a technique for program specialization, optimization and, more generally, program transformation. We present an idea to use supercompilation for verification of parameterized programs and protocols, present a case study and report on our initial experiments.

A predicate linear temporal logic LTL/sub /spl lambda/=/ without quantifiers but with predicate /spl lambda/-abstraction mechanism and equality is considered. The models of LTL/sub /spl lambda/=/ can be naturally seen as the systems of pebbles (flexible constants) moving over the elements of some (possibly infinite) domain. This allows to use LTL/sub /spl lambda/=/ for the specification of dynamic systems using some resources, such as processes using memory locations, mobile agents occupying some sites, etc. On the other hand we show that LTL/sub /spl lambda/=/ is not recursively axiomatizable and, therefore, fully automated verification of LTL/sub /spl lambda/=/ specifications via validity checking is not, in general, possible. The result is based on computational universality of the above abstract computational model of pebble systems, which is of independent interest due to the range of possible interpretations of such systems.