Automata Learning and Model Checking Techniques

Automata learning and model checking have emerged as complementary pillars of formal verification, each addressing the synthesis and analysis of system models from distinct perspectives. Automata learning focuses on inferring abstract representations of system behaviour—often as deterministic finite automata—from observations or queries posed to a system under learning. These techniques range from classical query-based algorithms that assume a minimally adequate teacher to neural and grammar-inference approaches that extract structure from raw data. Model checking, by contrast, exhaustively explores state-space models to verify whether they satisfy specified temporal or logical properties. Together, these methods enable the reverse engineering of black-box systems, the validation of legacy components, and the automated generation of reliable controllers. Recent advances have addressed scalability, non-determinism and probabilistic behaviour, widening the applicability of both learning and checking to protocols, embedded devices and interactive software. The interplay between active learning, conformance testing and symbolic model checking is now central to rapid prototyping, security analysis and trustworthy AI systems.

Research from Nature Portfolio

No recent Nature Portfolio content available.

Automata Learning and Model Checking Techniques publication trend

The graph below shows the total number of articles in automata learning and model checking techniques across all publications each year (not limited to Nature Index journals).

Technical terms

Deterministic finite automaton (DFA): An abstract machine that recognises regular languages through a finite set of states, transitions and an accepting condition. Active automata learning: A process by which a learner infers a model of a system by posing membership and equivalence queries to that system. Membership query: A request to determine whether a particular input sequence is accepted by the target system. Equivalence query: A request to verify whether a hypothesised model matches the target system’s behaviour, often answered by a counterexample. Model checking: A systematic method for exploring all states of a finite model to verify compliance with formal specifications. Visibly pushdown automaton (VPA): A pushdown automaton whose input alphabet dictates stack operations, enabling precise learning of nested or recursive input structures.

References

1. DLIQ: A Deterministic Finite Automaton Learning Algorithm through Inverse Queries. Information Technology And Control (2022).
2. Learning minimal automata with recurrent neural networks. Software and Systems Modeling (2024).
3. V-Star: Learning Visibly Pushdown Grammars from Program Inputs. Proceedings of the ACM on Programming Languages (2024).
4. Fingerprinting and analysis of Bluetooth devices with automata learning. Formal Methods in System Design (2022).
5. Interface protocol inference to aid understanding legacy software components. Software and Systems Modeling (2020).