SELF-VERIFICATION SYSTEM FOR AUTOMATION TECHNOLOGY
Patent Information
- Application Number
- DE502020009637
- Authority / Receiving Office
- DE · DE
- Patent Type
- Patents
- Current Assignee / Owner
- Priority Date
- 2019-09-18
- Filing Date
- 2020-08-20
- Publication Date
- 2024-11-21
- Estimated Expiration
- 2040-08-20
AI Technical Summary
Current anomaly detection systems in industrial automation technology are reactive, prone to false positives, lack transparency, and are centrally managed, leading to inefficiencies and increased risk of failures due to their inability to proactively address anomalies and prevent unauthorized changes or attacks.
A decentralized automation technology system utilizing a distributed ledger database with participant nodes that autonomously check for rule violations, allowing dynamic negotiation of rules and incorporating AI-based algorithms for anomaly detection, enabling proactive and reliable anomaly detection without a central unit.
This approach minimizes false positives, provides transparent and deterministic security, and ensures reliable anomaly detection and response, reducing the risk of failures and enhancing system trust by allowing participant nodes to dynamically define their properties and relationships.
Description
[0001] The invention relates to a system of automation technology.
[0002] Field devices are already known from the state of the art and are used in industrial plants. They are widely employed in process automation as well as in manufacturing automation. In principle, field devices are defined as all devices used close to the process that provide or process process-relevant data or information. Thus, field devices are used to acquire and / or influence process variables. Measuring devices or sensors are used to acquire process variables. These are used, for example, for measuring pressure and temperature, conductivity, flow rate, pH, level, etc., and acquire the corresponding process variables such as pressure, temperature, conductivity, pH value, level, and flow rate. Actuators are used to influence process variables.These include, for example, pumps or valves that can influence the flow of a liquid in a pipe or the fill level in a container. In addition to the aforementioned measuring devices and actuators, field devices also include remote I / Os, radio adapters, and generally any devices located at the field level.
[0003] A large number of such field devices are produced and distributed by the Endress+Hauser Group.
[0004] In modern industrial plants, field devices are typically connected to higher-level units via communication networks such as fieldbuses (Profibus®, Foundation® Fieldbus, HART®, etc.). These higher-level units are usually control units, such as a PLC (programmable logic controller) or a DCS (distributed control system). The higher-level units are used, among other things, for process control and commissioning of the field devices. The measured values acquired by the field devices, especially sensors, are transmitted via the respective bus system to one (or possibly several) higher-level units, which may process the measured values further and forward them to the plant's control center. The control center is used for process visualization, process monitoring and control, as well as diagnostics and data storage, etc., via the higher-level units.In addition, data transmission from the higher-level unit via the bus system to the field devices is required, particularly for configuring and parameterizing field devices and for controlling actuators. Such fieldbuses are increasingly being replaced by Ethernet-based communication networks.
[0005] Field devices generate a wide variety of data. This data includes, in addition to the aforementioned sensor measurement data, which provides plant operators with information about the current process values at their plant's measuring points, control data such as that used for actuator regulation. Furthermore, the data includes diagnostic, historical, and / or status data, which informs the plant operator about problems with the field devices or their current status, as well as calibration / parameterization data.
[0006] Today's anomaly detection systems in networked automation systems, which consist of multiple participants (for example, in the OT world of a process environment, i.e., field devices, control units, etc.), rely on historical data and recognize normal or abnormal behavior of the systems being monitored using pattern recognition, for example, by employing AI ("artificial intelligence") algorithms. Normal and abnormal behavior in such a networked system is determined across the entire state space of participant properties, their relationships (e.g., the number of connections of a network participant or the services used), and network properties. Such properties include, for example, the amount of data transferred, the number and duration of a (network) connection, the published services of a network participant, etc.
[0007] The large variance of the state space across the number of properties and their non-standardization leads to a large, opaque number of false positives. The system therefore loses its primary function, resulting in a lack of trust, particularly regarding the transmitted process variables, and placing a significant burden on the operator.
[0008] Current measures taken in response to anomalies are predominantly reactive. This offers no benefit to the operator. The entities being monitored currently lack the capability to address the individual characteristics of all participants. Furthermore, there is no reliable way to prevent the entities being monitored from being shut down or to modify / disable individual monitoring functions during operation. Current anomaly control systems are centrally managed, increasing the risk of a single point of failure. This centralized management prevents participants from dynamically defining or negotiating their properties and relationships. Anomalies generally cannot be prevented; only a response can be implemented after the fact.
[0009] The object of the present invention is therefore to offer reliable, proactive and safety-oriented anomaly detection in an automation technology system.
[0010] DE10 2016 118613 A1 relates to a system and a method for determining or monitoring a process variable in an automation system with at least one higher-level unit, PLC, and a plurality of field devices, wherein the field devices generate data, wherein the field devices are in communication connection with each other, and wherein the higher-level unit and the field devices represent participant nodes, TK, according to a distributed ledger or blockchain technology.
[0011] The problem is solved by an automation technology system according to claim 1.
[0012] The advantage of the system according to the invention lies in the reliable detection of anomalies occurring within the database. Anomalies include malfunctions of individual participant nodes, as well as changes in the behavior of participant nodes caused by attacks or unauthorized access to participant nodes. The database is a distributed ledger database. At least some of the database's participant nodes are automation components. These components are used in automation technology to collect, transmit, and process data originating from and / or used to control a process.Accordingly, the automation components include predefined field devices, but also remotely located units such as cloud platforms or IT components like network switches. The participating nodes communicate with each other via the internet or using non-internet-based telecommunications technologies (e.g., satellite). It may be possible for the automation components to be connected to an additional network, such as a fieldbus. In this case, the data must be transmitted synchronously over both networks (internet and fieldbus) to reliably detect anomalies.
[0013] By implementing the rule set in each participating node, the system can operate autonomously, as each node checks for rule violations. A central unit for anomaly detection is no longer required. The system exhibits high reliability, also due to redundancies. The rule set is dynamically negotiated a priori by each individual participating node. This drastically minimizes the state space, resulting in fewer false-positive anomalies and creating a fast, deterministic, reactive safety system with clearly defined and transparent rules for every participating node.
[0014] By using a distributed ledger database, the system is transparent, decentralized, and self-governing. Since all data from every participating node is stored at all times, detected rule violations or anomalies are undeniable for every participating node at any time.
[0015] According to a first variant of the system according to the invention, the decentralized database comprises only the automation components as participant nodes.
[0016] According to a second embodiment of the system according to the invention, the decentralized database includes additional participant nodes besides the automation components. These participant nodes can thus also be integrated into an existing database. The additional participant nodes can be, for example, IT components such as servers, etc., which do not necessarily have to be located near the plant.
[0017] According to an advantageous embodiment of the system according to the invention, the rule set is stored as program code on each of the participating nodes, which program code is then executed by the participating nodes. In the context of blockchain-based databases, such program code is referred to as "smart contracts." Smart contracts, in particular, represent contracts or verify or support the technical negotiation or execution of a contract, but can be extended as needed so that the rule set can be represented.
[0018] According to an advantageous further development of the system according to the invention, it is provided that the rule set includes algorithms which, when executed on the participant node, perform a check for an impending violation of the rule set.
[0019] According to an advantageous embodiment of the system according to the invention, the algorithms include AI-based evaluations for checking for an impending violation of the rules. Suitable examples include algorithms that utilize neural networks or DeepL learning-based algorithms. This enables continuous improvement of the rules.
[0020] According to a preferred embodiment of the system according to the invention, the results obtained from an AI evaluation of a participant node are communicated to the algorithms of the other participant nodes. In this way, not every participant node "learns" separately, but rather exchanges the findings, so that the collective of participant nodes learns together, builds up a wealth of experience, and continuously improves the rule set.
[0021] According to an alternative advantageous embodiment of the system according to the invention, the algorithms include threshold-based evaluations for checking for an impending violation of the rule set. This alternative is particularly suitable for participant nodes that have only limited resources available in terms of performance, memory, and / or energy. While the learning aspect is eliminated here, new insights can be incorporated by uploading a new version of the rule set.
[0022] According to an advantageous embodiment of the system according to the invention, the rule set comprises different types of algorithms, and the participating nodes are configured to select and execute at least one of the algorithms depending on their available resources. For example, certain high-performance participating nodes can use Kt-based algorithms and improve the rule set, while other, less powerful participating nodes use threshold-based algorithms.
[0023] According to a preferred embodiment of the system according to the invention, the rule set has a dynamic and a static component, wherein the content of the static component is unchangeable and wherein the content of the dynamic component is changeable and / or expandable.
[0024] According to an advantageous embodiment, the system according to the invention is designed such that the content of the dynamic component is modified and / or extended if a predetermined proportion of the participating nodes agree. For example, a permissible value range for a field device acting as a participating node is to be changed. This change to the rule set is proposed by the field device acting as a participating node, or, for example, by the user's operating PC, which is also a participating node. During the change process, all participating nodes are checked to ensure that the change pertains to a permissible category and / or that the participating node proposing the change can be authenticated. Only if, for example, more than half of the participating nodes can confirm this is the dynamic component of the rule set modified accordingly.
[0025] According to an advantageous embodiment of the system according to the invention, the set of rules comprises the following rules: Services through which participant nodes are permitted to communicate with other participant nodes according to the ruleset; A permitted value range within which measured values contained in the transactions of the participant nodes are specified; A maximum packet rate and bandwidth of a respective participant node; A maximum number of simultaneous connections of a participant node; A maximum ratio of time to new connections of a participant node; A maximum time for existing connections to or from a participant node; Error rate in connections between participant nodes; Resource utilization of a participant node; Measurement and control value behavior; Automation communication patterns (e.g., behavior of acyclic and cyclic communication of the real-time bus); Parameter settings; Data and communication heuristics or patterns; Metacommunication data; Interface utilization of a participant node.
[0026] The invention is explained in more detail with reference to the following figures. They show Fig. 1 : an explanation of a database designed according to distributed ledger technology; and Fig: 2 : an embodiment of the system according to the invention.
[0027] Fig. 1 This document illustrates a decentralized database (DB) designed according to distributed ledger technology. In this case, the database is based on blockchain technology. Blockchain technology became known as the backbone of the internet currency "Bitcoin." A blockchain, i.e., a chain of linked data blocks BL1, BL2, BL3, allows for high data integrity. The following is a brief explanation of how a database (DB) used for this invention functions.
[0028] A data block BL1, BL2, BL3 typically consists of at least two components: First, there is a data field DF. This data field DF stores data in the form of transactions TA. A transaction TA is the transmission of data from a first participant node TK1, TK2, ..., TK6 to a second participant node TK1, TK2, ..., TK6 in a communication network, such as the internet. A transaction TA contains a transmitted value, for example, data from the field device FG, as well as the sender and receiver of the transaction TA. Participant nodes TK1, TK2, ..., TK6 are all devices that constitute the database or are connected to it and enable distributed ledger functionality.
[0029] A data field DF of a data block BL1, BL2, BL3 contains at least one transaction TA, more often several transactions TA.
[0030] Secondly, a data block BL1, BL2, BL3 contains a checksum #1, #2, #3. Such a checksum #1, #2, #3 is a hash value and is generated through sometimes complex calculations. For this purpose, all transactions TA of the data field within a block BL1, BL2, BL3 are calculated to arrive at an intermediate value. For example, the Merkle root of the total number of transactions TA is calculated. The exact operating principle will not be discussed here. For further information, see, for example, https: / / en.wikipedia.org / wiki / Merkle_tree.
[0031] This calculated intermediate value is then combined with the checksum #1, #2, #3 of the previous data block BL1, BL2, BL3 to obtain the checksum #1, #2, #3 of the current data block BL1, BL2, BL3. For example, the one in Fig. 1The data block BL2 shown contains a checksum #2. This checksum #2 was calculated from the transactions TA stored in the data field DF of data block B2 and the checksum #1 of the preceding data block BL1. Similarly, the one in Fig. 1 The data block BL3 shown has a checksum #3. This checksum #3 was calculated from the transactions TA stored in the data field DF of data block B3 and the checksum #2 of the preceding data block BL2.
[0032] The integrity of the data, i.e., the protection of the data against subsequent manipulation, is thus ensured by storing the checksums #1, #2, #3 of the preceding data block BL1, BL2 in the respective subsequent data block BL2, BL3. A blockchain therefore consists of a series of data blocks BL1, BL2, BL3, each containing one or more transactions (TA) and marked with the checksums #1, #2, #3. A change to the data generates a modified intermediate value, which also changes the checksums #1, #2, #3 of the respective data block BL1, BL2, BL3. The subsequent data block BL1, BL2, BL3 is therefore no longer compatible with the preceding data block BL1, BL2, BL3. Once successfully validated, the data in a data block BL1, BL2, BL3 is thus no longer modifiable by an attacker.
[0033] New data blocks BL1, BL2, BL3 are created at regular intervals. All transactions (TA) created after the creation time of the last data block BL1, BL2, BL3 are stored in the data field of the new data block BL1, BL2, BL3.
[0034] The complexity of block creation can be increased by requiring the generated checksums #1, #2, #3 to adhere to a predefined format. For example, it might be specified that the checksum must be 24 digits long, with the first four digits being 0. For this purpose, in addition to the intermediate value of the transaction TA and the checksum of the previous data block, a sequence of numbers, called a "nonce," of a defined length is used to calculate the checksums #1, #2, #3 of the current data block BL1, BL2, BL3. Calculating the new checksums #1, #2, #3 takes correspondingly longer because there are only a few nonces available that can be used to calculate a checksum #1, #2, #3 with the specified criteria. Finding such a suitable nonce is what causes the additional time expenditure described above.
[0035] After the checksums #1, #2, #3 of a new data block BL1, BL2, BL3 have been generated, the data block is transmitted to all participant nodes TK1, TK2, ..., TK6. The validation-capable participant nodes TK1, TK2, TK3, TK4 then verify the checksums #1, #2, #3 of the new data block BL1, BL2, BL3. Only after successful validation is the data block BL1, BL2, BL3 stored in all participant nodes TK. Specifically, this requires successful validation by more than half of all validation-capable participant nodes TK1, TK2, TK3, TK4. An attacker would therefore have to manipulate or control a large number of validatable participant nodes TK1, TK2, TK3, TK4 to inject / create a malicious data block BL1, BL2, BL3 and successfully validate the injected data block. With an increasing number of validatable participant nodes TK1, TK2, TK3, TK4, this becomes virtually impossible.
[0036] Validating a data block BL1, BL2, BL3 requires significantly less effort than creating the data block itself. The checksums #1, #2, #3 are calculated backwards, the intermediate value of the transactions TA, or the checksums #1, #2, #3 of the previous data block BL1, BL2, BL3, are recovered, and compared with the actual intermediate value or the actual checksums #1, #2, #3 of the previous data block BL1, BL2, BL3. If these values match, the data block BL1, BL2, BL3 is successfully validated.
[0037] The following describes how a self-monitoring system for an automation technology plant can be set up using such a database: In Fig. 2The system according to the invention is shown schematically. The database DB, which is structured as described above, consists of six participant nodes TK1, TK2, TK3, TK4, TK5, and TK6, which are configured as full nodes. Each of these stores an image AB of the contents of the database DB, i.e., the chain of data blocks DB1, DB2, and DB3. The participant nodes TK1, TK2, TK3, and TK4 are automation components, for example, field devices (participant nodes TK1, TK2, and TK3). By executing distributed ledger software or a software stack, these automation components can act as participant nodes, load the image of the contents of the database DB, write data to the database, and verify transactions. However, they cannot verify transactions themselves. Examples of such field devices are mentioned in the introductory part of the description.The field devices collect, for example, measured values of a process engineering process and transmit them within the database, for example to a control unit (participant node TK4).
[0038] The participant nodes TK5 and TK6 are IT components; in particular, participant node TK5 is a cloud server and participant node TK6 is a user's operating PC. If these two participant nodes are integrated into the plant's OT network or have process-relevant operating applications, they also fall under the definition of an automation component.
[0039] Each of the participating nodes TK1, ..., has a rule set integrated as program code. This set contains information about the number and identity of all participating nodes TK1, ..., TK6, as well as rules regarding the actions, properties, and states of each of the participating nodes TK1, ..., TK6. Upon execution, a participating node TK1, ..., TK6 can verify the transactions of the participating nodes according to the rule set RG, thereby enabling anomaly detection. In the context of the blockchain technology used as a database in this example, such a rule set can be implemented as a smart contract.
[0040] A violation of a rule set, or a resulting anomaly, can be detected using two different methods. Both are implemented via corresponding algorithms, which can be implemented on the participant nodes TK1, ..., TK6 depending on their available resources.
[0041] Firstly, the algorithms enable threshold-based evaluations. Here, the contents of transactions from participant nodes TK1, ..., TK6 are compared with the data contained in the rule set RG. If these differ from each other, or deviate from each other by a certain value (threshold), this represents an anomaly or a rule violation.
[0042] Alternatively, the algorithms provide AI-based evaluations to check for an impending violation of the RG rule set. Suitable algorithms include those using neural networks or deep learning-based algorithms. The results generated by the algorithms after an AI evaluation of a participant node TK1, ..., TK6 can be shared with the other participant nodes TK1, ..., TK6 to establish a knowledge base.
[0043] The rulebook also includes a catalog of measures linked to detected anomalies, thereby initiating actions to rectify malfunctions and / or actions to prevent further errors / attacks.
[0044] The following describes a first example of implementation in which the RG rule set is used to determine whether a violation of the RG rule set or an anomaly exists: A new, unknown device attempts to join the database DB as a new participant node. However, joining the database is only possible if the RG rule set allows a potential participant to be assigned a unique identification identifier that is also defined as an authorized participant within the RG rule set.
[0045] Here, a distinction can be made between three different cases, which trigger different reactions from the participant nodes.
[0046] Firstly, the access request may be unplanned and unwanted. In this case, the requesting device is not authorized. Its identification information is therefore not listed in the RG rule set, so it cannot authenticate or identify itself. The participant nodes TK1, ..., TK6 receive the access request as a transaction, check its content, in this case the sender identification, and detect a rule violation (possible via threshold-based evaluation and / or an AI algorithm). As a resulting measure linked to this rule violation, any non-compliant communication or transaction within the database DB with this device is blocked (Layer 7 or Edge / Switch can block).
[0047] In a second scenario, the access request is planned and intentional. The new potential participant node in the database submits the access request. The participant nodes TK1, ..., TK6 receive the access request as a transaction, check its content—in this case, the sender identification—and verify that it exists in the rule set RG. The device thus has valid authorization and can authenticate or identify itself. Subsequently, the device is added to the database DB as a new participant node.
[0048] In a second scenario, the access request is unplanned but intentional. In this case as well, the requesting device lacks authorization. Its identification information is not listed in the rule set RG, preventing it from authenticating or identifying itself. The participating nodes TK1, ..., TK6 receive the access request as a transaction, examine its content (in this case, the sender identification), and detect a rule violation. As a countermeasure, the existing participating nodes TK1, ..., TK6 agree to add the requesting device's new identification information to the rule set RG. This can be achieved by the existing participating nodes TK1, ..., TK6 verifying the device's identification and its rule set to be added, or by negotiating a new rule set based on the device's new information and the existing information in the rule set RG.In both cases, the device is either accepted or rejected as a new participating node in a majority vote. If the device is accepted as a new participating node, it must also agree to the newly negotiated rules.
[0049] The device can also be added as a new participant node by submitting its identification information and agreeing to the existing rule set RG without introducing any new rules itself. The existing participant nodes TK1, ..., TK6 verify the identification information and add the device as a new participant node after checking whether the existing rule set RG can be complied with given the new participant node's properties. The majority procedure or voting process described above can also be used for this.
[0050] Finally, a further embodiment of the method according to the invention is described. The participant nodes TK1, TK2, and TK3, acting as field devices, continuously collect measured values from a process engineering operation and transmit these to the higher-level unit (participant node TK4). A range of values, within which the measured values may fall, has been defined through calibration and verification and stored in the rule set RG. The participant nodes TK1, ..., TK6 check the measured values contained in the transactions of participant nodes TK1, TK2, and TK3 according to the rule set RG. If the value range is exceeded or fallen below, they generate a warning to the plant operator. Detection of the exceedance or fall below the range can be achieved through threshold-based evaluation or by means of AI algorithms. In the event that a participant node TK1, ...Using TK6 AI algorithms, it can also predict an overshoot or undershoot of the value range that has not yet occurred but is foreseeable ("Predictive Maintenance"). List of reference symbols
[0051] AK1, ..., AK4 Automation components BL1, BL2, BL3 Data block BO Dashboard DB decentralized database EL Electronic unit FG Field device KN Communication network KS Communication interface RG Rules TA transaction TK1, ..., TK6 Participant node #1, #2, #3 Hash values of the data blocks
Claims
1. An automation system, comprising: - A decentralized, distributed database (DB) based on distributed ledger technology, in particular a blockchain, comprising a number of participating nodes (TK1, ... , TK6), wherein the participating nodes (TK1, ... , TK6) are configured to share data or information with each other via transactions (TA), wherein the database (DB) is configured to save the transactions (TA), in particular in interlinked data blocks (BL1, BL2, BL3); - A policy (RG) which is implemented on each of the participating nodes (TK1, ... , TK6), wherein the policy (RG) comprises information about the number and identity of all participating nodes (TK1, ... , TK6) as well as rules regarding actions, properties, and statuses of each of the participating nodes (TK1, ... , TK6); and A number of automation components (AK1, ... , AK4), which are participating nodes (TK1, ... , TK6) of the decentralized database (DB), wherein each of the participating nodes (TK1, ... , TK6) is configured to check and validate transactions (TA) between the participating nodes (TK1, ..., TK6) at all times based on the policy (RG), wherein each of the participating nodes (TK1, ... , TK6) is configured to implement at least one of the following measures if a violation of the policy (RG) is detected: i. Blocking communication within the database (DB) with an affected participating node (TK1, ... , TK6); ii. Blocking communication outside the database (DB) with an affected participating node (TK1, ... , TK6); iii. Closing ports; iv. Stopping the services of each participating node (TK1, ... , TK6).
2. The system as claimed in claim 1, wherein the decentralized database (DB) only comprises the automation components (AK1, ... , AK4) as participating nodes (TK1, ... , TK6).
3. The system as claimed in claim 1, wherein the decentralized database (DB) comprises other participating nodes (TK1, ... , TK6) in addition to the automation components (AK1, ... , AK4).
4. The system as claimed in at least one of the preceding claims, wherein the policy (RG) is saved as a program code on each of the participating nodes (TK1, ... , TK6), said program code being executed by the participating nodes (TK1, ... , TK6).
5. The system as claimed in at least one of the preceding claims, wherein the policy (RG) contains algorithms which check whether the policy (RG) is imminently going to be violated when executed on the participating nodes (TK1, ... , TK6).
6. The system as claimed in claim 5, wherein the algorithms comprise AI-based assessments for the purpose of checking whether the policy (RG) is imminently going to be violated.
7. The system as claimed in claim 6, wherein the results generated by an AI assessment of a participating node (TK1, ... , TK6) are shared with the algorithms of the other participating nodes (TK1, ... , TK6).
8. The system as claimed in claim 5, wherein the algorithms comprise threshold-based assessments for the purpose of checking whether the policy (RG) is imminently going to be violated.
9. The system as claimed in at least one of claims 5 to 8, wherein the policy (RG) has different kinds of algorithms and wherein the participating nodes (TK1, ... , TK6) are configured to select at least one of the algorithms and execute it or them depending on the resources available to each participating node (TK1, ... , TK6).
10. The system as claimed in at least one of the preceding claims, wherein the policy (RG) has a dynamic section and a static section, wherein the content of the static section cannot be changed and wherein the content of the dynamic section can be changed and / or expanded.
11. The system as claimed in claim 10, which is configured in such a way that the content of the dynamic section is changed and / or expanded if this is allowed by a predefined number of participating nodes (TK1, ... , TK6).
12. The system as claimed in at least one of the preceding claims, wherein the policy (RG) comprises the following rules: - Services via which the participating nodes (TK1, ... , TK6) are permitted to communicate with other participating nodes (TK1, ... , TK6) in accordance with the policy; - A permitted range of values in which measured values included in the transactions between the participating nodes (TK1, ... , TK6) are provided; - A maximum packet rate and bandwidth for each participating node (TK1, ... , TK6); - A maximum number of simultaneous connections for a participating node (TK1, ... , TK6); - A maximum ratio between time and new connections for a participating node (TK1, ... , TK6); - A maximum time for existing connections from or to a participating node (TK1, ... , TK6); - Error rate for connections between participating nodes (TK1, ... , TK6); - Resource usage of a participating node (TK1, ... , TK6); - Behavior of measured values and control values - Automation communication patterns (e.g. behavior of the acyclic and cyclic communication of the real-time bus) - Parameter settings - Data and communication heuristics or patterns - Meta communication data; - Interface usage of a participating node (TK1, ... , TK6).
