Supervisory Control of Discrete Event Systems for Small Language Under Cyber Attacks

Xiaojun Wang, Shaolong Shu, Feng Lin · 2025

Read Supervisory Control of Discrete Event Systems for Small Language Under Cyber Attacks by Xiaojun Wang, Shaolong Shu, Feng Lin (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2508-02083.txt.

Introduction

Xiaojun Wang, Shaolong Shu, and Feng Lin present a 2025 paper on controlling automated systems whose signals travel over networks an attacker can corrupt. In such discrete event systems, attacks scramble what the supervisor sees and does, so the behavior the plant may show becomes uncertain. The smallest set of behaviors that is guaranteed anyway is called the small language.

A supervisor must keep at least that much behavior alive for the system to do its job. The team defines two properties, tuned to attacks, covering whether the plant can be steered and whether it can be watched well enough. They prove the control task is solvable exactly when the target behavior has both properties. When it does not, they give conditions for the best workable substitute.

Catalog record

CREATORXiaojun Wang, Shaolong Shu, Feng Lin
YEAR2025
COLLECTIONPAPERS
FORMATTXT
FILE SIZE60 KB
VERSION1.0
LICENSEhttp://creativecommons.org/licenses/by/4.0/
RIGHTSCC BY
SOURCEarXiv

Source: https://arxiv.org/abs/2508-02083

Browse the full library · About the Tower of Babel

Full text

The opening of the work, from the library's hosted text:

Supervisory Control of Discrete Event Systems for Small Language Under Cyber Attacks Xiaojun Wang, Shaolong Shu, Feng Lin Abstract Cyber attacks are unavoidable in networked discrete event systems where the plant and the supervisor communicate with each other via networks. Because of the nondeterminism in observation and control caused by cyber attacks, the language generated by the supervised system becomes nondeterministic. The small language is defined as the lower bound on all possible languages that can be generated by the supervised system, which is needed for a supervised system to perform some required tasks under cyber attacks. In this paper, we investigate supervisory control for the small language. After introducing CA-S-controllability and CA-S-observability, we prove that the supervisory control problem of achieving a required small language is solvable if and only if the given language is CA-S-controllable and CA-S-observable. If the given language is not CA-S controllable and/or CA-S-observable, we derive conditions under which the infimal CA-S-controllable and CA-S-observable superlanguage exists and can be used to design a supervisor satisfying the given requirement. I Introduction With the rapid development of computer, communication and control technologies, wired or wireless communication networks are used to exchange information between controllers and plants, which have the advantages of high efficiency, low cost, and high transmission capability. Since information is sent through networks, it is unavoidable that the controlled system may suffer from cyber attacks. Therefore, control of systems under cyber attacks has become an important research direction in systems and control, for both continuous-variable systems and discrete event systems. In this paper, we investigate supervisory control of discrete event systems (DES), where the plant and the supervisor are communicated via networks. Specifically, the communication channel from the plant to the supervisor is called observation channel and the communication channel from the supervisor to the plant is called control channel. In observation channels, an attacker can delete, insert, and/or change some observable events, which may cause the supervisors to make incorrect decisions. In control channels, cyber attacks may alter disablement or enablement of some controllable events, which will change the behavior of the system. Within the DES framework, cyber attacks are investigated extensively [ 1 , 2 , 3 , 4 , 5 ] . For example, [ 2 ] shows that an intruder may interfere with the feedback performance of the system, causing the controllers in a supervisory control system to fail. To solve the problem, the authors propose a framework for modeling supervisory control systems to estimate how much damage that cyber attacks might cause. In [ 3 ] , four types of cyber attacks are discussed: sensor insertion attacks (event insertion attacks), sensor erasure attacks (event deletion attacks), actuator enablement attacks (event enablement attacks), and actuator disablement attacks (event disablement attacks). It is shown that a diagnoser can be constructed to detect attacks. Once an attack is detected, the supervisor will disable all controllable events. The researchers categorize cyber attacks into the following three types: sensor attacks, actuator attacks, and joint sensor-actuator attacks. Sensor attacks in observation channels are considered in [ 6 , 7 , 8 , 9 , 10 , 11 , 12 , 13 ] . Among them, [ 7 ] investigates the problem of synthesizing active sensor attackers against initial-secret of supervisory control systems. Based on an all attack structure which records state estimates for both the supervisor and the attacker, the authors present algorithms for synthesizing successful attack strategies. [ 8 ] investigates the problem of state estimation under attacks which may erase some events that have occurred and/or insert some events that have not actually occurred. The authors solve the problem by constructing a joint estimator which contains all the possible attacks. In [ 9 ] , the authors propose a new attack detection mechanism in which the supervisor only needs to keep track of the last observable event received to solve the problem of detecting stealthy sensor attackers in cyber-physical discrete event systems. The authors of papers [ 12 ] and [ 13 ] propose to model sensor attackers with the following steps. First, sensor attack constraints are modeled as a finite state automaton AC , which describes the attack capabilities. It is required that the sensor attack action (insertion, deletion, and replacement) initiated by the sensor attacker is instantaneous. Second, the sensor attack over attack constraint is modeled as a finite state automaton A , which is the attack that they aim to synthesize. Third, a fixed unit time interval, i.e., one tick , is used to model the observation channel. Actuator attacks in control channels are investigated in [ 14 , 15 , 16 , 17 ] . Among them, [ 14 ] proposes the resiliency automata, based on which the authors develop a polynomial method to design a resilient supervisor such that the plant under control does not reach any unsafe state if the number of actuator attacks is less than the required safety level. In [ 15 ] , the problems of estimation and prevention of actuator attacks are studied. Based on the proposed notions of strong and weak actuator enablement estimabilities, the authors design an estimator and a prevention module to mislead an intruder’s attack estimation by using the reverse sensor functions to modify sensor readings. [ 16 ] considers the problem of intrusion detection and prevention in supervisory control systems, where the attacker has the ability to enable vulnerable actuator events that are disabled by the supervisor. To solve the problem, the authors present a mathematical model and propose a defense strategy. Joint sensor-actuator attacks in both observation and control channels are investigated in [ 18 , 19 , 20 , 21 , 22 , 23 , 24 , 25 ] . In [ 18 ] , the authors consider a supervisor that guarantees the safety of the system even when sensor readings and actuator commands are compromised. Their solution methodology reduces the problem of synthesizing a robust supervisor against deception attacks to a conventional supervisory control problem. In [ 19 ] , the authors develop an attack structure computed as the parallel composition of the attacker observer and the supervisor under attack. It is used to select attacks that cause the closed loop system to reach an unsafe state. In [ 20 ] , the authors synthesize resilient supervisors against combined actuator and sensor attacks. A constraint-based approach for the bounded synthesis of resilient supervisors is developed by reducing the problem to a quantified Boolean formula problem. In [ 21 ] and [ 22 ] , the authors investigate how to find a powerful joint sensor and actuator attack policy, not how to synthesize a supervisor to tame attacks. For more information on cyber attacks in discrete event systems, the reader is referred to a tutorial paper [25]. We have also investigated supervisory control of DES under joint sensor-actuator attacks. For sensor attacks, we propose a new attack model, called ALTER (Attack Language for Transition-basEd Replacement) model [ 26 , 27 , 28 , 29 ] . In the ALTER model, an attackable transition can be replaced by any strings in the corresponding attack language. The ALTER model is both general and specific. It is general in the

Read the complete text