ControlML: A domain-specific modeling language in support of assessing internal controls and the internal control system David Heisea,∗, Stefan Streckerb , Ulrich Franka a Enterprise b Information Modelling Research Group, University of Duisburg-Essen, 45141 Essen, Germany Systems Development Research Group, University of Hagen, 58084 Hagen, Germany Abstract In this paper, we refine and extend an earlier language design to introduce a domain-specific modeling language (DSML) for internal controls modeling as an extension to an enterprise modeling method. The language is aimed at supporting the assessment of a firm’s internal control system through the use of conceptual models of internal controls. In the paper, we report on the design of the modeling language, on its integration with the enterprise modeling method, present the language specification, and discuss language applications in the context of the assessment of an internal control system. Keywords: Internal control, Internal Control System, Assessment of Internal Controls, Conceptual Modeling, Enterprise Modeling, Language Design, Domain-Specific Modeling Language, Design Science Research 1. Introduction The task of assessing a firm’s internal control system is beset by remarkable challenges: In addition to the challenges posed by the sheer number of controls and their possible interactions, internal controls affect the organization at different levels, refer to a multitude of resources and assets, and address a variety of risks (Maijoor, 2000; Crawford and Stein, 2002). Hence, an assessment of the internal control system requires an in-depth understanding of a firm’s business, the risks it faces and the controls it has put in place to treat risk exposure at all relevant organizational levels. That understanding in turn implies an understanding of entity objectives, business processes, organizational resources, structures, roles, and responsibilities (Rikhardsson et al., 2006; Elder et al., 2010). Accordingly, the assessment task involves stakeholders with different professional backgrounds and perspectives on internal control matters including executives, line managers, process owners, risk managers, internal and external auditors (Spira and Page, 2002). In this respect, the inherent complexity is intensified by differing mindsets, diverging professional languages, and the resulting barriers to communication. ∗ Corresponding author: phone +49 201 183-2719; fax +49 201 183-4011 Email addresses: david.heise@uni-duisburg-essen.de (David Heise), stefan.strecker@fernuni-hagen.de (Stefan Strecker), ulrich.frank@uni-duisburg-essen.de (Ulrich Frank) Preprint submitted to International Journal of Accounting Information Systems July 19, 2013 To cope with these challenges, methods and instruments are required that purposefully reduce the complexity inherent in the assessment of an internal control system and that facilitate communication about internal control matters among diverse groups of stakeholders. Although business process models have long been used to support the task, it has repeatedly been established that present business process modeling approaches provide only limited support for assessing the internal control system. A comparative evaluation study reviews prevalent business process modeling notations in the light of key domain requirements and concepts and concludes that the studied approaches do not provide the required modeling constructs for the assessment and raises the demand for constructs that specifically express the semantics of internal controls (Carnaghan, 2006). Dunn (2006) in a commentary on the last study adds that present approaches do not provide support for the appropriate representation of the relevant organizational context of internal controls (i. e., entity objectives, business processes, organizational resources, structures, roles, and responsibilities). Such concepts are, however, common to enterprise modeling approaches such as ARIS (Scheer, 1992), SOM (Ferstl and Sinz, 1998) and MEMO (Frank, 2012) which provide further abstractions of the enterprise and its organizational action systems (i. e., modeling constructs to create conceptual models representing, for example, organizational resources, structures, roles, and responsibilities). While current enterprise modeling approaches do provide support for aspects of the enterprise in demand when supporting the assessment of internal controls, research on domain-specific modeling languages (DSMLs) for representing internal controls and internal control systems in the context of enterprise modeling approaches has—as far as we are aware—been limited to early language design prototypes yet. These observations have inspired our research on a dedicated DSML for internal controls modeling. The present work follows a design science research process to develop such a DSML, ControlML, as extension to an enterprise modeling method. In the paper, we refine and extend an earlier language design, report on its integration with the enterprise modeling method, present the language specification, and apply the language to demonstrate its practical use for assessing internal control systems. The next section relates the present research to prior work on internal controls modeling. Section 3 comments on the research method. Design objectives, requirements, and essential assumptions are established in Section 4. Section 5 outlines the theoretical and conceptual background informing the language design. The structural specification of the language, its meta model, is discussed in Section 6. A prototypical language application is devised in Section 7. The paper concludes with a discussion and an outlook on paths for future research in Section 8. 2. Related work Among the first research on internal controls modeling is that of Karagiannis et al. (2007). In their approach, three SOX-related modeling concepts—Control, Risk, and Account—are introduced as an extension 2 to an enterprise modeling approach and a corresponding modeling tool. Language concepts are specified by a meta model (see also Karagiannis, 2008, p. 1164) and related to further concepts for risk management such as Event and Action. Interpretation of the semantics of modeling concepts is, however, partly left to the language user, as the meta model does not show attributes and the accompanying documentation does not offer further details. Moreover, important aspects of the design of a DSML are only briefly discussed, for instance, a notation and corresponding diagram types. In a series of papers, Namiri and Stojanovic (2007a,b) identify additional domain concepts (e. g., Control Objective, Risk Assessment, Authority), and visualize conceptual relationships in a UML class diagram notation (Namiri and Stojanovic, 2007b, p. 63). Their “domain model” structures the technical terminology in the auditing domain with the intention to “formulate logical statements representing the controls constraining the behavior of a Business Process” (Namiri and Stojanovic, 2007b, p. 62). However, the domain model does not specify the details (i. e., syntax and semantics) of a DSML and nor does it account for further modeling concepts pertaining to the organizational action system surrounding internal controls. With a different focus on internal controls modeling, Sadiq et al. (2007) interpret controls in terms of rules and target rule-based compliance checking based on automated reasoning (for related approaches, see e. g., Governatori et al., 2006, 2008). In contrast to the present work, models of controls and business process models are kept separate for pragmatic reasons: “prematurely load[ing] business process models with compliance controls will be highly problematic from a practical standpoint” (Sadiq et al., 2007, p. 151). Their approach starts from textual descriptions of controls in terms of normative statements such as “The creation and approval of purchase orders must be undertaken by two separate purchase officers”. The textual descriptions of internal controls are transformed into formal representations based on a modal logic to be used to reason on compliance violation (Lu et al., 2007, 2008). A procedure to link the logic-based statements about internal controls to process models is described, so that clauses appear as annotations of a process model. Further rule-based approaches to modeling internal controls include the work on a PROLOG implementation by Bailey, Jr. et al. (2000) and on a Petri net-based online auditing tool aimed at rule-based compliance verification (van der Aalst et al., 2011). Petri nets have earlier been discussed as a means to document business processes and corresponding internal controls also aiming at algorithmic verification of compliance (Pitthan and Philipp, 1997; Chen and Lee, 2003). In a related approach, Weigand and Elsas (2012) propose to derive internal controls from REA models based on an axiomatic foundation. In summary, a rich body of literature contributes to modeling internal controls. However, prior work has not closely examined the syntax and semantics of domain-specific modeling concepts for internal controls modeling. Nor does any previous research deal with the intricacies of the reuse of concepts in the context of enterprise modeling and of extending present enterprise modeling approaches with domain-specific concepts for internal controls modeling. Based on the preliminary results of this ongoing research project (Strecker et al., 2010, 2011a), the present work refines and extends an earlier tentative language design to introduce a 3 DSML for internal controls modeling as extension to an enterprise modeling method. Present rule-based and formal approaches complement the conceptual modeling approach presented in this work with conceptual models seen as a prerequisite to further automation. This brief analysis of the current state-of-the-art guides our research on domain-specific modeling constructs in support of the assessment of internal controls. 3. Research method The artifact designed in this research is a (domain-specific) modeling language; a linguistic artifact consisting of an abstract syntax (represented as a meta model), a concrete syntax (a corresponding graphical notation) as well as explanations of the intended semantics of language concepts in natural language (Wand et al., 1995; Wand and Weber, 2002; Frank, 2002). The present work is based on a research method configured for the epistemological particularity of research on modeling languages (Frank, 2006b). The method is informed by fundamental thoughts rooted in the philosophy of language, such as the dilemmas arising when attempting to evaluate an (artificial) language (Frank, 2006a). It has been refined over the past 15 years (see Frank, 1998, for further historical context) and constitutes the methodological foundation for long running design science research projects of which the present work is part (see Frank, 2012, for an overview of projects and Strecker et al., 2011b, 2012, for recent research results). In the light of idealized design science research processes (e. g., Verschuren and Hartog, 2005; Peffers et al., 2007; Geerts, 2011), the present work reports on the activities of defining design objectives (Section 4), design and development (Sections 5 and 6), and demonstration (Section 7) besides the previous problem identification and motivation. The concluding section addresses further research activities, for example, those relating to the evaluation of the artifact. This work targets integration with an enterprise modeling method to benefit from reuse of existing and mature modeling concepts and procedural guidelines. Developing a DSML in this context presupposes reconstructing technical terms of the targeted domain and their (perceivable) semantics (Ortner, 2008; Frank, 2010). One widespread approach to conceptual reconstruction—and that followed here—is to review, analyze, and interpret pertinent literature in the field under consideration (e. g., Gelinas et al., 2004; Dunn et al., 2005a; Davis et al., 2007; Moeller, 2008; Elder et al., 2010). Reconstruction of a professional terminology is an iterative process involving more than the identification of candidate (meta) concepts, their attributes and relations. Instead it requires the identification and resolution of terminological ambiguity and truncation, for instance, which may imply the introduction of additional abstractions. That in turn requires the shaping of their intended semantics, and leads to designing abstractions (i. e., language concepts) appropriate for domain-specific analyses and applications based on deliberate design decisions. The underlying research method is therefore driven by devising and analyzing domain-specific use scenarios describing, among other things, prototypical model-based analyses in the intended use context (Frank, 2013). 4 4. Design objectives, requirements and assumptions The overarching objective of the design science research project in which ControlML and further DSMLs were developed is to devise a comprehensive modeling method for governance, risk, and compliance (GRC). The primary objective of the design of ControlML is to effectively and efficiently support stakeholders when performing an assessment of the internal control system. More specifically, this research is aimed at supporting stakeholders interpreting internal controls and the surrounding organizational action system when assessing the internal control system. Given that an understanding of the internal control system is educed in group processes (Damianides, 2005, p. 79), its purpose is to support group processes by reducing the complexity inherent in internal control systems and by providing abstractions tailored to the perspectives of stakeholders involved. Consequently, this work is aimed at fostering and facilitating communication and collaboration among stakeholders participating in the assessment, and particularly addressing interaction between auditor and auditee. It also aims to increase the transparency of internal control matters, specifically by visually representing internal controls as part of the organizational action systems, and by improving traceability of the controls in place to treat risk exposure. These high-level requirements— reducing complexity, fostering communication and collaboration, and improving transparency of internal control matters—are refined to establish the subsequent requirements a DSML aimed at supporting internal controls modeling should satisfy (see also Strecker et al., 2011b): R1 A DSML should link internal controls to the surrounding organizational action system composed of all organizational entities relevant to the assessment. This organizational context is provided by (at least) entity objectives, business processes, business risks, performance measures, organizational resources, structures, roles and their responsibilities (Carnaghan, 2006, p. 177). Rationale: The organizational context in which an internal control is designed to be used is of particular importance to its accurate interpretation (Sutton and Hampton, 2003; Spira and Page, 2002). Providing explicit and qualified relationships between controls and organizational context by way of a DSML is seen as both contributing to reducing complexity, and improving transparency of internal control matters. R2 A DSML should account for relationships among internal controls on different organizational levels, from IT operations to business processes to value chains to the organization as whole. Rationale: PCAOB Auditing Standard No. 5 and other regulations assume relationships between internal controls as expressed, for instance, by a control hierarchy (Gelinas and Dull, 2010, p. 241). Relations between internal controls need not, however, be strictly subordinate. Rather, controls relate to each other in often case-specific ways as exemplified, for example, by diverse taxonomies of controls (Dunn et al., 2005a, p. 455). Explicit and qualified relationships among controls as part of a DSML promise to improve the transparency of internal control systems and, specifically, the interrelations between internal controls. R3 A DSML should account for the multiplicity of means to achieve control objectives and of the resulting 5 variety of internal control implementation. Rationale: A wide spectrum of ways to achieve control objectives employing a multitude of different organizational measures including policies and procedures, manual and automated controls is discussed (Gelinas and Dull, 2010; Elder et al., 2010). Providing appropriate abstractions of control means by means of a DSML is seen as a contribution to improving the transparency of internal control matters, and to reducing the complexity of internal control systems. R4 A DSML should provide means for justifying the existence and relative importance of an internal control and for revealing assumptions underlying internal control justification. Rationale: It has repeatedly been suggested to foster communication and collaboration on internal control matters by annotating assertions underlying the control specification and its intended usage (Carnaghan, 2006, p. 177). It is assumed that by providing a traceable rationale of an internal control, accurate interpretation of internal controls will be fostered and communication barriers among stakeholders lowered. R5 A DSML should provide meaningful representations of internal controls and internal control systems to satisfy the needs of affected groups of prospective users. To foster an intuitive use, representations should, as far as possible, correspond with abstractions, concepts and (visual) representations known and meaningful to the targeted group of stakeholders. Moreover, a DSML should provide means to integrate the various perspectives with each other. Rationale: The assessment process involves stakeholders with different professional backgrounds. To foster communication among these stakeholders, a DSML needs to support the perspectives of different (groups of ) stakeholders into account. It is assumed that satisfying these requirements will ensure the high-level requirements and the design objective will be met. That in turn implies that hypotheses underlie the relationships between higher and lower level requirements. It is, for example, assumed that if a modeling language supports integrated perspectives tailored to the needs of stakeholders involved in the assessment of internal controls, communication barriers are lowered and collaboration among stakeholders with different professional backgrounds is facilitated. In other words, the language positively contributes to approaching the primary design objective. In this regard, the identified requirements and the design goals guide the design of the modeling language, ControlML. 5. Theoretical and conceptual background Research on ControlML builds on a method for enterprise modeling, the “Multi-Perspective Enterprise Modeling” (MEMO) method researched and applied since the early 1990s (Frank, 1994). The rationale for choosing the MEMO method (subsequently referred to as MEMO) over other enterprise modeling approaches such as, for example, ARIS (Scheer, 1992) is based on several considerations: (1) MEMO provides an extensive set of modeling constructs relevant to designing a language for internal controls modeling, including the modeling of organizational goals, units, roles, responsibilities, and resources; (2) MEMO is based on a 6 Meta Level / M2 Type Level / M1 Business Process name : String organisationalUnit : String Server name : String type : {application, database, file} < A/R Manager > Financial eTransaction Application Server Authorize Credit Instance Level / M0 < Smith, John D. > Internal Control designator : String description : String isPreventive : Boolean Internal Control: „IO 3.3: Prevent unauthorized refund“ Description: Demands a segregation of duties in the refund business process, i.e., a separate authorization of the refund isPreventive: true Conducted at: ... Responsible person: ... < Miller, J.-A. > Meta Type Type Instance < Smith, John D. > started: ... completed: ... Conducted at: 2012-09-24 13:45 Responsible person: Smith, John D. Appl. Server #01 Authorize Credit #6110840A4 Authorize Credit #.... Authorize Credit #7781085C9 started: ... completed: ... Serial-No.: ... Installed: ... Appl. Server #02 Serial-No.: PX118D-07D Installed: 2007-07-15 started: 2012-09-24 09:45 completed: 2012-09-24 10:04 Conducted at: ... Responsible person: ... instance of Figure 1: Essential levels of abstraction in the MEMO language architecture: Instance, Type and Meta Type [@Publisher: Please use colored figures for web and black/white-figures for print (the latter are provided in separate files marked with the extension “bw”)] meta modeling approach whose language architecture is extensible through DSMLs (Frank, 2011d) and is complemented by guidelines and a process model aimed at the design of the respective DSML (Frank, 2010); and (3) in contrast to proprietary approaches, the specification of the MEMO method—especially its language architecture, meta modeling language, and most language specifications—are publicly available (Frank, 2012). In MEMO, DSMLs are specified by a meta model, a corresponding graphical notation and accompanying descriptions of language concepts and of intended semantics of modeling concepts in natural language. Thus, a language specification is semi-formal as it comprises formalisms (i. e., the meta model and visual language) and informally specified artifacts (i. e., verbal descriptions for those aspects of the language design that should not or cannot be expressed (solely) through formal means based on a deliberate design decision). The MEMO Meta Modeling Language (MEMO MML) and the MEMO language architecture serve as meta modeling foundation for the design of ControlML. The language designer specifies modeling concepts of the DSML (i. e., meta types) at the meta level—referred to as the M2 level in OMG terminology (see Fig. 1). These meta types (e. g., Business Process) are instantiated by the language user (modeler) to types at the type level (M1 ). Note that the MEMO MML is defined at the meta-meta or M3 level omitted from Fig. 1 (Frank, 2011d). Building on the MEMO MML for defining language concepts of a DSML fosters the integration of modeling languages. DSMLs are specified using the same meta modeling language, which enables the 7 language designer to reuse meta types already specified in a MEMO modeling language. Sharing common concepts at the meta level (e.g., Business Process) leads to integrated models at type level, for example, a business process model integrated with an organization structure model, with a model of IT assets, and with a model of internal controls. Moreover, the reuse of existing modeling concepts promises to increase the productivity of both language design and language application: Language designers benefit from mature language concepts and notations and can focus on relevant additions and modifications at the meta level; modelers and language users benefit from the reuse of existing models at type level (e. g., reuse of business process models) and can focus on extending or refering to these models. Figure 2 illustrates the integration of type level models in MEMO and shows diagram types provided by MEMO for different perspectives on an organization from a strategic management perspective (a strategy net, goal system or value chain diagram) to a business operations perspective (business process map, business process diagram, organization chart diagram) to an IT operations perspective (an IT resource diagram or an object model). Each modeling language in MEMO provides a set of language concepts for the aspects the language focuses on and that can be reused in other MEMO modeling languages. Of particular importance for internal controls modeling are (1) concepts for modeling organizational structure (to assess internal control scope); (2) business processes and organizational resources (to determine the organizational action system surrounding an internal control); (3) organizational goals, objectives and performance indicators (to analyze internal controls with respect to entity objectives); (4) risk and impact in case of risk occurrence (to provide a rationale for control existence). In this regard, a kernel of DSMLs and concepts is provided by MEMO: The organization modeling language (MEMO OrgML; Frank, 2011c) specifies concepts for modeling business processes and organizational structures (e. g., Process, Event, and Organisational Unit) and dedicated diagram types (e. g., a business process diagram detailing the control flow of a process) (Frank, 2011a,b,c). The goal modeling language (MEMO GoalML; Köhling, 2012) provides concepts for modeling organizational goal systems (e. g., Abstract Goal, Accountability Relation). The strategy modeling language (MEMO SML) provides high-level abstractions commonly referred to at the senior management level, for example, concepts for creating value chain diagrams (e. g., Value Chain Activity). The IT modeling language (MEMO ITML; Frank et al., 2009) allows the modeling of IT assets (e. g., Information System) and associations to the business processes they are used in, among others. The MEMO kernel of languages has been extended by DSMLs to support further analyses in various application domains. Two extensions to the kernel are indicated at the right-hand side of Figure 2: The risk modeling language (MEMO RiskML) provides concepts such as Risk and Effect Specification and diagram types like the shown risk diagram which accounts for interrelations among risks (Strecker et al., 2011b). The performance measurement language (MEMO MetricML) introduces concepts such as Indicator and an indicator system diagram type in support of the reflective design and use of performance measurement 8 IT Resource Diagram Strategy Net Business Process Diagram Business Process Map Goal System Diagram 0 ..1 0 ..* 1 ..* Figure 2: Integration of DSMLs in the MEMO method 0 ..* u p d a t e () 0 ..1 0 ..* 0 ..* In t er fa c e d e s c r ip tio n : S tr in g s o u r c e ID : S t r in g p r o t o c o l: S tr in g 0 ..1 in p u t P a r a m L is t: X M L S tr in g r e s u lt : X M L S t rin g o b ta in s d a ta f r o m 0 ..1 C h a p t e rS lid e Text c o n te n t: X M L S t r in g g e tC o n t e n t() V id e o v id e o : V id e o d e s c r ip tio n g e t V id e o ( ) G ra p h ic s c a p tio n : S tr in g 0 ..1 T itle S lid e v a r ia n t o f v a r ia n t o f g e t T it le () : S tr in g C o m p o s e d F ig u re D ia g ra m G ra p h ic s fig u r e : G r a p h ic s g e tG ra p h ic s() 0 ..* F o rm u la C irc le R e c t a n g le 1 ..1 M o d e llin g L a n g u a g e n a m e : S t r in g 0 ..* C o n c ep t u a lD ia g ra m M odel st a tic R e p r e se n t a tio n : G r a p h ic s v ie w o n n a m e : S tr in g ty p e : D ia g r a m 0 ..* 1 ..* D a t a D ia g ra m v isu a lis a tio n T y p e : D ia g r a m g e tD ia g r a m () L in e st a r tP o sit io n : C o o r d in a t e e n d P o s it io n : C o o r d in a te R e g io n fillC o lo u r: C o lo u r r e lW id t h : P e r c e n t a g e r e lH e ig h t : P e r c e n t a g e r e lP o sit io n : C o o rd in a te P u b lica t io n tit le : S tr in g ... d e fin e s la y o u t o f C h o ic e 0 ..1 A ss ig n m e n t F ra m e S ty le ... In h e r it s fr o m 0 ..1 O p e n Q u e st io n G ra p h ic sO b je c t lin e C o lo u r: C o lo u r A d v ic e w a r n in g : B o o le a n e m p h a sis : In t e g e r C it a t io n A s sig n m e n t g r o u p W o r k : B o o le a n tim e F r a m e : In te g e r 0 ..* Q u e s t io n 0 ..1 Q u e s t io n F ra m e S t y le ... F ra m e S t y le ic o n : G r a p h ic s 0 ..* ... d e f in e s la y o u t o f S lid e S ty le h e a d lin e F o n t: F o n t 0 ..* f o o tlin e F o n t: F o n t 0 ..1 ... C h a p te r S t y le 0 ..* c r e a t e d : D a te c h a p te r F o n t: F o n t S ty le h e a d lin e F o n t: F o n t cre ate d : D ate f o o tlin e F o n t: F o n t r e g u la r B a c k g r o u n d : C o lo u r fo n t: F o n t 0 ..1 ... fo n t C o lo u r: C o lo u r P r e se n ta t io n S t y le t itle B a c k g r o u n d : C o lo u r c h a p te r B a c k g o u n d : C o lo u r t itle F o n t: F o n t r e g u la r F o n t: F o n t 0 ..1 h e a d lin e F o n t: F o n t f o o tlin e F o n t: F o n t ... d e f in e s la y o u t o f 0 ..1 R e g u la rS lid e 0 ..* t itle : S t rin g 0 ..* 1 ..1 1 ..1 d e f in e s la y o u t o f 1 ..1 0 ..* d e fin e s la y o u t o f r e p r e se n ts T a b le n u m b e r O f R o w s : In te g e r n u m b e r O f C o ls : In t e ge r g e t A r r a y () Lo go 0 ..1 0 ..* g e tT itle () : S t r in g 0 ..* A b st ra c t F ra m e r e lW id t h : P e r c e n ta g r e lH e ig h t: P e r c e n ta g e r e lP o sitio n : C o o rd in a te F o o t lin e A n im a tio n e f fe c t: A n im a t io n T y p e 0 ..1 o n E v e n t: E v e n t T y p e 1 ..* r e p r e s e n ts t itle o f 0 ..* S lid e t itle : S tr in g c r e a te d : D a te sid e R a tio : R a t io F ra m e M a s t e rF ra m e H e a d lin e in c lu d e s in c lu d e s 0 ..1 1 ..1 0 ..* r e p r e se n ts title o f 0 ..* S im p le P re se n t a t io n 1 ..1 0 ..* 0 ..* 1 ..1 C h ap ter P r o x y S lid e sta r ts w it h t it le : S t r in g 0 ..* 1 ..1 g e t S lid e N o ( ) : In te g e r 0 ..1 g e t C h a p te r N o ( ) : In t e g e r 0 ..1 0 ..1 1 ..1 su c c e s so r su c c e sso r o f of v a r ia n t o f 0 ..* S t ru c t u re d P r e se n ta t io n P re se n t a t io n d e sc r ip t io n : S t rin g c r e a te d cre ate d : D ate 0 ..* 1 ..* t it le : S t r in g su b T itle : S t r in g Object Model A u th o r la st N a m e : S t rin g fir st N a m e : S tr in g 0 ..1 Value Chain Diagram Organizational Chart starts with includes Diagrams of core MEMO languages starts with instance of 9 inherits from includes inherits from Indicator System Diagram Risk Diagram Diagrams of extensions to MEMO Meta Type – abstracts over a set of similar types and describes their structure in terms of attributes (Abstract Meta Type – The italic header characterizes the Meta Type as abstract, i.e., it cannot be instantiated) Language – The colored rectangle serves as unique identifier for the modeling language this meta types origins from; hence the specification of the meta type can be found in the specification of the originating modeling language. <EntityName> Meta-Association – describes a relation between the two connected meta types; is further qualified by a designator, multiplicities and (optional) roles Generalization/Specialization Attributes – Attributes are initialized upon instantiation of the meta type, i.e., they are assigned a value; can (optional) have a multiplicity that describes the amount of initializations Role – describes the participation of this meta type in the association further <EntityName> <role> <Designator> <AttributeName> ':' <EntityName>|<TypeName> <AttributeName><AMultiplicity> ':' <EntityName>|<TypeName> <Multiplicity> <Multiplicity> i <AttributeName> ':' <EntityName>|<TypeName> Intrinsic Attribute – are not initialized upon instantiation of the meta type like regular attributes, but one level further down <#> Surrogate – serves as a surrogate for various other meta types and, thereby, aims to reduce the complexity in the visual represenation of the meta model (e.g., by reducing the amount of meta-associations in the meta model); a surrogate should always be used with exemplary meta types the surrogate serves to abstract over <String> <EntityName> Multiplicity (also: Cardinality) – describes to how much instances of this meta type one instance of the distant meta type is at minimum/at maximum is connected to Constraint – describes particular constraints that have to be valid for all instances of the meta type; can be described in natural language or, e.g., using the OCL ‘<<surrogate>>‘ <EntityName> <EntityName> <EntityName> <EntityName> Surrogate Relation – serves to describe exemplary meta types the surrogate serves to abstract over; note that the list of meta types is always exemplary, that is, not complete Figure 3: Overview of key concepts and notation elements of the MEMO Meta Modeling Language (MEMO MML) systems (Strecker et al., 2012). These modeling languages are integrated with each other through common concepts at the meta level. For example, the concept of Process defined by the MEMO OrgML is reused in the GoalML (which goals does a business process pursue?), in the ITML (where is an information system used?), and in the RiskML (which business processes are affected by a certain risk type?). In Figure 2, the integration of the modeling languages by sharing meta types is exemplarily indicated by the edges drawn between the diagrams. In the meta models, the reuse of a concept originally specified in another MEMO modeling language is indicated by a color-coded language identifier attached to the meta type. Figure 3 offers an overview of key concepts and notation elements of the meta modeling language, MEMO MML. While at first, the MEMO MML looks similar to the UML MOF and UML class diagrams, it differs in three respects (for a detailed comparision of MEMO MML and UML MOF, see Frank, 2011d): First, the MEMO MML focuses on concepts necessary for meta modeling and, thus, does not come with the complexity of a multi-purpose approach as, for instance, UML MOF. Second, as a dedicated meta modeling language, the MEMO MML contains concepts which specifically provide possible solutions to known challenges in meta modeling. And third, it has a specific graphical notation to support a clear visual distinction between meta models on the one hand and models at type level (and, specifically, UML class diagrams) on the other. 10 6. Language Design: The MEMO Control Modeling Language (MEMO ControlML) 6.1. Terminological analysis In auditing literature and practice, the terms “control”, “internal control”, and “internal control system” are commonly used (Moeller, 2008). Despite their proliferation, the lack of a precise definition has repeatedly been commented upon (Maijoor, 2000). The term “control” is subject to a considerable diversity of uses and disciplines, for example, “management control, organisational control, internal controls, operational control and financial control, which all seem to revolve around the same concept” (Rikhardsson et al., 2005, p. 5). The auditing perspective on internal control is decisively influenced by the Committee of Sponsoring Organizations of the Treadway Commission (COSO) (COSO, 1992, 2004) and subsequent auditing standards such as the PCAOB Auditing Standard No. 5 (for a discussion see Maijoor 2000 and Rikhardsson et al. 2006): “COSO defines internal control as a process, effected by an entity’s board of directors, management and other personnel. This process is designed to provide reasonable assurance regarding the achievement of objectives in effectiveness and efficiency of operations, reliability of financial reporting, and compliance with applicable laws and regulations. [. . . ] Internal control is not merely documented by policy manuals and forms. Rather, it is put in by people at every level of an organisation. [. . . ]” (COSO 2010; adapted from COSO 1992). While this very broad conceptualization provides insights into essential domain-specific concepts (e. g., objectives, policy, reasonable assurance), it also points to terminological ambiguity: The concept of “internal control” includes both, procedural (i. e., dynamic) aspects (i. e., business processes, auditing and monitoring processes) and structural (i. e., static) aspects (e. g., policies, organizational structures, organizational roles). It also links to other concepts such as risk or control objectives: “Enterprise risk management is a process, [. . . ] designed to identify potential events that may affect the entity, [. . . ] to provide reasonable assurance regarding the achievement of entity objectives” (COSO, 2004, p. 2). In fact, the COSO framework breaks down “internal control” into five interrelated components: Control Environment, Risk Assessment, Control Activities, Information and Communication as well as Monitoring (Gelinas and Dull, 2010, pp. 224–225). A prevalent representation of “internal control” in auditing practice is the “control matrix” (Gelinas and Dull, 2010, p. 227) and its variants, for example, a (process and) control matrix which maps control objectives to relevant control means grouped by business processes (Gelinas and Dull, 2010, p. 649). A key domain concept is the control objective, sometimes also termed the control goal (Gelinas et al., 2004, p. 249). It represents a desired state of an enterprise (e. g., “Prevent unauthorized refunds”) with respect to achieving an entity objective (e. g., “Minimize error rate of incorrect refunds”) that is threatened by a risk (e. g., “Internal fraud due to fraudulent behavior of employees”). A control objective is associated with a recommended course of action that should be taken to provide reasonable assurance that entity objectives will be met and, thus, corresponding risks of not achieving it are mitigated. The course of action can 11 involve policies, procedures, practices, or organizational structures as concrete measures or means of control implemented to ensure effectiveness of a control (Elder et al., 2010). An important means to prevent fraud is “segregation of duties” (Gelinas and Dull, 2010, p. 255). This means of control is intended to achieve the control objective and represents an abstraction over static means of control such as written policies or organizational structures and dynamic means of control such as activities and processes. Alternative denominators for the control means concept include “control activity” (IT Governance Institute, 2007) and “control plan” (Gelinas et al., 2004, p. 249). Both terms, however, carry a significant risk of misinterpretation: The term “activity” raises associations with dynamic abstractions neglecting static aspects while the term “plan” emphasizes a perspective other than the intended means-end association. Examples of general control means given in COSO publications include the authorization of transactions and adequate safeguarding of assets and records. The achievement of the desired outcome of a control objective is measured by an indicator (e. g., “Percentage of fraudulent refund transactions”) as is the severity of risk. Responsibilities, as in the RACI (responsible, accountable, consulted, informed)-conceptualization suggested, e. g., by the IT Governance Institute (2007), are typically defined for more than one organizational role (“executive”, “business process owner”, etc.) with respect to a control objective. It is important to note that the monitoring and auditing of an internal control system (e. g., performed as internal control assessment by an external auditor) are processes detached from the actual internal control system, in that the system itself becomes subject to the audit (COSO, 2009). These interpretations of domain terminology represent intermediate results of the conceptual reconstruction process which inform the subsequent language design. Further intermediate results of the reconstruction process including a semantic net are discussed in Strecker et al. (2010, 2011a). 6.2. Specification of the ControlML The language specification of the MEMO ControlML is composed of a meta model represented in the MEMO MML, a respective graphical notation, and explanations of the intended semantics of language concepts. The diagrammatic representation of the meta model is split into two figures to focus on different aspects of the language design. In the following sections, we focus on the key elements of the modeling language ControlML and discuss extensions to the earlier tentative language design (for a discussion of alternative designs and a more elaborate rationale of design decisions see Strecker et al., 2011a, pp. 17ff.). Fig. 4 presents the corresponding meta model excerpt of the ControlML with a focus on the language kernel; a focus on integration of ControlML with the MEMO family of modeling languages is illustrated in the meta model excerpt in Fig. 6 and discussed in Section 6.3. The initial design decision with respect to the targeted domain pertains to the specification of language concepts specifying an internal control type. It follows from the terminological analysis that the very conception of “internal control” cannot be conceptualized as a singular concept as such, but rather as 12 Control Involvement pertains to 0..* belongs to 0..* 0..* intendedState : String 0..* explanation : String areaFS : FinancialStatementArea 0..* 1..* Codification targets 0..* supports C1 governed by specification : String origin : PolicyProvider 1..1 recommendedAction : String 1..* assertion [0..*] : Assertion intendedEffect : {preventive,detective,corrective} isManual : Boolean 0..* realised by Control Category Organisational Unit has Control Means 1..1 Control Objective name : String description : String 0..* involvementType : String 0..* 0..* C1 Reference Object context Control Objective inv: self.ControlObjective->excludes(self) Figure 4: Meta model of the MEMO ControlML with a focus on the language kernel (revised from Strecker et al., 2011a, p. 20) an abstraction over various concepts which in turn constitute an internal control. An essential design decision has therefore been to represent an internal control by its control objective and by the means of achieving the control objective. Hence, two dedicated meta types, Control Objective and Control Means, are introduced (a similar kernel is proposed by Namiri and Stojanovic, 2007a; Karagiannis, 2008). The rationale for introducing those two meta types is to enable dedicated model-based audit analyses based on respective conceptual models of control objectives and control means. To account for essential requirements (see Section 4), further refinements are, however, necessary and are detailed below. Three constituents are central to the language specification (the meta model is shown in Fig. 4): (1) The Control Objective language concept and the ancillary concepts Codification and Control Category; (2) the Control Means modeling concept; (3) the Control Involvement meta type. Ad (1) Control Objective. The meta type Control Objective serves to describe the intentions connected with an internal control. It is described by a natural language description of the desired state of control by the attribute intendedState (“Prevent unauthorized refunds”) and by an explanation of the intended state, explanation (“The billing system receives shipping items from the shipping system [. . . ]”). Annotating the financial statement area, areaFS, links a control objective to financial reporting. The recursive supports association between respective types allows to represent the internal control system as a hierarchy or net of controls (see Fig. 5 for an example). Thus, it enables further analyses such as which IT controls relate to which business controls. The ancillary language concept Codification specifies a legal regulation the control objective is governed by and its originating policy provider. Feedback from practicing auditors advised keeping track of these codifications (e. g., a certain clause in an auditing standard) and of the originating policy provider (e. g., “PCAOB Auditing Standard No. 5”) as a contribution to justifying the existence and importance of an internal control. The meta type Control Category allows the creation of customized control taxonomies and the assignment of one or more categories to a control objective to support a flexible 13 categorization of internal controls, since in auditing practice there are different approaches to structuring control objectives (e. g., Dunn et al., 2005b, pp. 441ff.). Ad (2) Control Means. The Control Means language concept serves to describe the means to achieve reasonable assurance. It is determined by the recommended course of action provided in a natural language specification, recommendedAction. Such a specification might mention a written policy and corresponding procedures (e. g., “The Accounting Supervisor makes copies of the checks and sends the check copies along with the invoice hard copy supporting documentation to the Cash Application Department”). The current conceptualization as shown in Fig. 4 aims to offer flexibility while, at the same time, maintaining a structure for assessment purposes. The terminological analysis indicates that (1) multiple means exist that can be deployed to achieve a control objective; (2) the same means can be reused by several control objectives; and (3) a certain means can pertain to several structural and procedural elements and thus exhibit a “multidimensional” characteristic (e. g., a written policy connected with a process “Authorize credit”). Due to the chosen cardinality of the targets-relationship between the meta types Control Objective and Control Means, it is possible to assign a multitude of means to one control objective; similarly, a control mean can be used to support the achievement of different control objectives, thereby promoting the management and reuse of control means in an organization. Following suggestions elicited from the literature, further semantics is specified by assertion (what are assumptions underlying the function of a control means and what is its rationale?) (Carnaghan, 2006, p. 177) and by classifying control means into preventive, detective or corrective controls according to their effect in time relative to the occurrence of a risk, intendedEffect (Gelinas et al., 2004, p. 253). Another characteristic is captured by a functional differentiation between manual and automated control activities in isManual attribute. Ad (3) Control Involvement. The Control Involvement modeling concept serves to describe the relation between a control objective and an organizational unit. Since further types of involvement beyond RACI are conceivable and to be expected (e. g., a “supports” involvement), Control Involvement is not limited to these four types of involvement but allows organizations to define their own types of involvement—by means of an involvementType and a description. Figure 5 drafts a graphical notation of the ControlML and illustrates the internal control diagram type. The diagram is based on the refund returned goods process drawn on by Carnaghan (2006, p. 200) and is further enriched with exemplary controls reconstructed from the COBIT documentation (IT Governance Institute, 2007). Different from other notational mappings, multiple meta types, Control Objective, Control Means, Control Category, and Codification, map to a single graphical symbol (a rectangle with an “auditor’s eye” attached). The rationale for this design decision is to provide a visual overview of (selected) internal controls with this diagram type. Given respective tool support, the model user (e. g., an auditor) could then 14 Internal Control Diagram Control Objective: Segregation of duties Intended State: Prevent unauthorized refunds Recommended Action: Separate authorization process for refunds above 500 €, signed by some different actor Intended Effect: preventive is Manual: true Category: Business Control supports Recommended Action: Ensure different user access rights for refund transactions Intended Effect: preventive is Manual: false Control Means Control Means supports Control Objective: User Account Management Intended State: No user has privileges that are not required for his current tasks. Explanation: Address requesting, establishing, issuing, suspending, modifying, and closing user accounts and […] Category: IT Control, Codification: COBIT, DS5.4 Recommended Action: With each change in the organizational structure and in the staffing, check for necessary modifications in the user accounts. Intended Effect: Preventive is Manual: true Control Means supports Control Objective: Segregation of duties Intended State: Prevent unauthorized transactions Category: IT Control Control Objective: Network Security Intended State: No unauthorized information flows to and fromObjective: networks are possible. of IT Security Control Management Explanation: Use security techniqueswith and business related Intended State: IT security is consistent management procedures, e.g., firewalls, […] requirements. Control Objective: Identity Management Regulation: COBIT, Explanation: Manage security at the highest appropriate Intended State: Every userITDS5.10 can be uniquely identified. organisational level, management […] Explanation: Ensure thatsoallthe users (internal, external, and Regulation: DS5.1 on IT systems […] temporary) and COBIT, their activity Category: IT Control, Codification: COBIT, DS5.3 Recommended Action: After each software update it has to be checked, if the user accounts and their privileges are still valid Intended Effect: Preventive is Manual: true Meta Type Control Objective Meta Types Control Category & Codification Meta Type Control Means Figure 5: An internal control diagram showing internal controls and their interrelations as a visual system of internal controls (based on the refund returned goods process drawn on by Carnaghan, 2006, 200 and further enriched with exemplary controls reconstructed from the COBIT documentation). navigate from such a high-level overview to viewing specific control objectives and have their control means shown in another diagram type. Note that other graphical notations and diagram types are conceivable, for instance, a tabular representation or a diagram type focusing on control means rather than control objectives (showing which control objectives a control mean addresses). In particular, ControlML is designed to assign different notational elements to the same modeling concept (e. g., to address specific perspectives of particular groups of stakeholders) and to offer various diagram types for the same model (each with, e. g., different foci). Moreover, the formal foundation of the DSML allows for transforming the internal control model into different forms of representation (e. g., a spreadsheet or a check list) to accommodate the specific needs of groups of stakeholders. 6.3. Integration with the MEMO family of modeling languages with extensions for assessment of internal controls The ControlML is integrated with the MEMO family of modeling languages through reuse of common language concepts (see Section 5). Figure 6 presents another aspect of the meta model showing how key concepts of the ControlML integrate with concepts of other MEMO languages (marked by a colored rectangle attached to a meta type). The semantics of the Control Objective language concept is further refined by links to the entity objectives that are the target of the control objective (Goal) and to indicators providing measurements related to the control objective (Indicator). The Control Involvement modeling concept is refined to refer to organizational roles (Organisational Role), so that roles defined in an organizational structure model can be reused for 15 measures assures achievement of Control Involvement pertains to 0..* Indicator 0..* 0..* 0..1 1..* C3 C2 I affectedEntity 0..* affects monitors Control Means riskMeasure 0..* targets 0..* 0..* Business Process Risk mitigates 0..* 0..* 0..* 0..* <<surrogate>> Reference Object Organisational Unit Information System Software ... type : {manual, semi-autom., autom.} II II Explanation of ControlML-specific attribute types frequencyType : {per instance, on demand, event driven, <timeUnit>} scopeType : {total, partial, sample} Audit Process frequency : FrequencyType 0..1 scope : ScopeType isExternal : Boolean Symbols for concepts reused from other MEMO languages: GoalML C2 1..1 1..1 Control Objective measures Organisational Role has 0..* 0..* 1..* realised by Goal C3 context Control Objective inv: self.ControlMeans.ReferenceObject->excludes( self.affectedEntity ) OrgML ITML RiskML MetricML context Control Means inv: self.ReferenceObject.BusinessProcess->notEmpty() implies self.ReferenceObject.BusinessProcess->forAll( b : BusinessProcess | b->isOclTypeOf(AuditProcess).isEmpty() ) Figure 6: Meta model of the MEMO ControlML with a focus on integration with the MEMO family of modeling languages (revised from Strecker et al., 2011a, p. 20; Rectangles “I” and “II” mark extensions discussed in Section 6.3) internal controls modeling. The explicit association of risks (Risk) with control means enables further analyses for auditing purposes. For instance, identifying risks without controls and vice versa (as proposed by Spira and Page, 2002). Likewise, control means may be linked to indicators, Indicator, measuring the outcome of actions associated with a particular control means. Furthermore, the organizational context of an internal control is introduced by associating the Control Means modeling concepts with those organizational entities (e. g., Business Process, Organisational Unit, Information System) that may participate in implementing (or realizing) an internal control—thereby shaping the semantics of control activities in terms of describing the organizational context the activity is executed in. For this purpose, the surrogate concept Reference Object represents all modeling constructs relevant to internal control assessment—with the intention of providing the necessary flexibility in modeling the means of internal control (see Section 6.1). The present language specification allows, for example, to represent the realization of a control means to achieve segregation of duties by referencing the two relevant business processes (“Prepare credit” and “Authorize credit”) as well as the two organizational roles involved (“Sales Account Manager” and “A/R Manager”). Extending the earlier language design, the affects relation is introduced between Control Objective and Reference Object. Besides representing the various organizational elements that realize a control means, 16 the meta type Reference Object is also used to represent the organizational entities that are affected by an internal control (see Rectangle “I” in Fig. 6). For example, a business processes type affected by the “segregation of duties” control or a server type affected by a “ensure correct user access rights” control. Note that affected and realizing reference objects may coincide. Although this seems confusing at first, the differentiation is necessary, since it enables analyses such as “which organizational elements are subject to internal control?” (affects-association), “does an organizational element contribute to the realization of an internal control?” (realised by-association) or “which organizational elements are affected by a control objective but are missing a respective control means?” (comparison of affects- and realised by-association). An example for this differentiation into affected and realizing reference objects is provided in Figure 7 in Section 7: The control objective “Segregration of duties” (at the center of the diagram) affects the “Refund returned goods” and, at the same time, is realized by the process “Authorize credit”. A further extension pertains to the support for the modeling of auditing and monitoring processes. Audit processes differ from business processes in that they are specifically designed to run outside of a firm’s regular operations with the intention to control a particular audit object such as a business process, a record of transactions etc. From a modeling perspective, however, audit processes exhibit characteristics of business processes in that a certain regularity can be assumed as well as a temporal order of tasks producing events. Hence, auditing and monitoring processes can, in principle, be represented using a business process modeling language such as BPMN or, in context of enterprise modeling, the MEMO OrgML. We, therefore, propose reusing and extending—that is, specializing—the Business Process concept specified in the MEMO OrgML to an audit process subtype (see Rectangle “II” in Fig. 6). The semantics of the Audit Process modeling concepts are further detailed by the particular frequency of process instantiation (i. e., at which intervals is the process triggered?), the scope of its analysis (e. g., are all, only a part or a specific sample of the instances of the audit object inspected?) and whether it is an external audit process. Specializing the generic Business Process meta type leads to a particular advantage: Current diagram types, modeling techniques, and implementations in modeling tools—optionally even the graphical notation—can be reused for modeling audit processes. An example is shown in Fig. 7 where the audit process “Audit refund transactions and detect irregularities” is represented using the regular business process notation symbol defined by the MEMO OrgML. This reuse of familiar modeling concepts is assumed to foster the acceptance of audit process modeling. The two OCL-constraints in Fig. 6 ensure that reference objects cannot realize a control means and, at the same time, be affected by a control objective this control means targets (C2); and that audit processes cannot realize a control mean (C3). 17 7. Language application: Applying MEMO ControlML to support the assessment of internal controls We now apply the MEMO ControlML to an application scenario to demonstrate how to utilize language concepts for the assessment of internal controls. The scenario is based on and inspired by a refund returned goods process drawn on by Carnaghan (2006, p. 200) for which internal controls are described in natural language. For example, “[t]he sales account manager must authorize returns by completing a ‘return merchandise authorization’ (RMA) paper form that is sent to customer service” and “[t]he information system restricts the ability to create, change, or delete sales order return and credit requests to authorized personnel” (quoted in Carnaghan, 2006). Note how these sentences refer to relevant organizational roles (“sales account manager”, “authorized personnel”), responsibilities (“must authorize”), resources such as documents (“RMA”) and IT artifacts (“information system”). Figure 7 reconstructs the description of the scenario in Carnaghan (2006, pp. 191ff.) using the MEMO ControlML and further languages from the MEMO family of modeling languages: It shows a goal model (top left) that represents (an excerpt of) a hierarchy of the enterprise’s strategic goals and subsequent business objectives; a business process model for “refunding returned goods” at three levels of detail (i. e., an aggregated process and its decompositions; bottom left); a model of the corresponding organizational structure (including organizational roles and committees; top right) and a model of IT assets used in the process (showing an information system abstraction of an ERP system; bottom right). Relationships between concepts in different models are explicitly modeled by associations (e. g., the ERP system used in the business process “Authorize credit”) or by shared concepts (e. g., the organizational role “A/R manager” in both the process “Authorize credit” and in the organizational structure model). Further abstractions provided by MEMO such as corresponding object models are omitted for the sake of clarity. The enterprise model shown supports dedicated analyses for the assessment of internal control systems based on the conceptual models of internal controls: The control objective “Prevent unauthorized refunds” recommends a segregation of duties in the aggregated process “Refund returned goods”. The internal control semantics is further specified by the IT control “Prevent unauthorized transactions”, by the risk “Internal fraud” it aims to mitigate, by the audit activity “Audit refund transactions and detect irregularity”, and by the indicator “Percentage of fraudulent refund transactions” to measure achievement of the control objective. Figure 7 also demonstrates how the control objectives can be associated with concepts that represent the organizational action system they are embedded in. First, they can be associated with the entity objectives they are aimed at assuring the achievement of (i. e., the goal “Minimize error rate of incorrect refunds” in the goal model). Second, control objectives can be linked to static and dynamic abstractions representing means of control. For example, the above mentioned control objective affects the business process “Refund returned goods”, whereas the actual realization of the recommended action “segregation of duties” 18 19 Legend Goods returned ec ts Control Objective Aggregated Business Process uses res a chie v em ent o f Organizational Role Committee Goods refunded Goods refunded refers to assu uses Authorize credit < A/R Manager > Goal Prepare credit < Sales Account Manager > Link to prior sale < Inventory clerk > Refund returned goods < A/R department > Credit customer account Goal: Minimize error rate of incorrect refunds (max. 1.5%) Business Process Goods linked to sale Authorize return of goods < Sales Account Manager > aff Goal: Process 99% of all refunds in under 1 Week affects Goal System Business Process by d ali ze re ? ‐ Fraudulent refund transactions [%] IT me as ur es # measures Software comprises ? # IT Risk Indicator Control Objective: Segregation of duties Intended State: Prevent unauthorized transactions […] Category: IT Control IT hardware resource Risk: Internal fraud Visibility: high Uncertainty: medium monitors to refers to responsible accountable informed Audit refund transactions and detect irregularity refers Control Objective: Optimize IT infrastructure Recommended Action: Establish an IT architecture board to guide architecture […] Category: IT Control, Regulation: COBIT, PO 3.5 Internal Control Diagram Control Objective: Segregation of duties Intended State: Prevent unauthorized refunds […] Category: Business Control Frequency of meetings [#/month] IT # s ate mitig Goal: Increase in customer satisfaction by 5% ERP Software ... Accounting Application Server ERP System A/R Manager Financial Officer Database Server IT Architecture Board IT directs Executive Board requires ... Sales Backup Sales Account Manager directs Organizational Structure IT Landscape Figure 7: Application scenario: Return refund goods process drawn on by Carnaghan (2006, 200) as a reconstruction using the MEMO ControlML and MEMO family of modeling languages (revised from Strecker et al., 2011a). is depicted at the most detailed level of the process model. Third, the IT control refers to an IT asset in the IT landscape model (“ERP system”) that realizes the segregation at the information system level by authorization and system access policies. Linking the two control objectives also demonstrates how associations between controls aid in analyzing internal control systems. Associating a control objective with a corresponding risk (“Internal fraud”) provides an implicit rationale for the existence of a control objective possibly indicated by a performance measure (“Fraudulent refund transactions in %”). Finally, the integration with an organizational structure model emphasizes different types of involvement of organizational roles. For instance, the two roles participating in the “Credit customer account” process—“Sales Account Manager” and “A/R Manager”—are linked to the control objective specifying their type of involvement (i. e., “accountable” and “responsible”, respectively), whereas a “Financial Officer” is one regularly informed but not explicitly modeled as part of the business process. Such integrated models of the enterprise and its internal control system promise to provide intuitive access to internal control matters and a comprehensible conceptual foundation for differentiated analyses during the assessment of internal control systems. By associating internal controls with further abstractions (e. g., of organizational goals or IT landscapes) and by providing dedicated diagram types for specific types of analyses—deliberately omitting confounding details—such an approach facilitates internal control-related communication and collaboration between groups of stakeholders with different professional backgrounds. Additionally, such enterprise models can serve as foundation for auditing purposes in that they support the comparison of internal controls in place with the likes of prescribed internal controls as defined by, for instance, COBIT (IT Governance Institute, 2007). ControlML can also be applied to reconstruct prescribed internal controls as reference models, that is, reusable conceptual models of internal controls associated with the promise to be applicable to a large class of prospective use contexts, and to improve the productivity of language users. Figure 8 illustrates how ControlML is applied to reconstruct controls defined by the COBIT 4.1 framework: Internal controls defined by the framework (e. g., COBIT PO 4.11) are reconstructed using the ControlML once (upper right quadrant) and stored in a repository (i. e., a model database) accessible by prospective language users (bottom right quadrant). Language users select those reference models from the repository they consider relevant for their own modeling purposes and import the reference models into their modeling tool (upper left quadrant). Given the inevitably imprecise specification of internal controls in respective frameworks (being applicable to a wide range of use contexts demands sacrifices to their level of detail), corresponding reference models of internal controls will remain underspecified in certain regards. For instance, the “Management of IT Security” control objective defined by COBIT demands involvement of “the highest appropriate organizational management” (IT Governance Institute, 2007). Since organizational structures vary to a great extent in organizations, it is not feasible to preconceptualize reference models with precise organizational roles (conceivable roles include the CIO or any other IT executive) with respect to control 20 Relevant Control Objectives for the Enterprise realized by <CIO/ IT executive> realized by Control Objective: Management of IT Security Intended State: IT security is consistent with business requirements. Explanation: Manage IT security at the highest appropriate organizational level, so the management […] Regulation: COBIT, DS5.1 Control Objective: IT Strategy Committee Control Objective: Data and System Ownership Control Objective: Supervision Control Objective: Technology Standards selection of relevant Control Objectives Control Objective: Network Security Intended State: No unauthorized information flows to and from networks are possible. Explanation: Use security techniques and related management procedures, e.g., firewalls, […] Regulation: COBIT, DS5.10 Control Objective: IT Architecture Board Intended State: The IT architecture conforms to […] Explanation: Establish an IT architecture board to […] Regulation: COBIT, PO3.5 Control Objective: IT Steering Committee Intended State: An IT steering committee is established Explanation: Establish an IT steering commitee […] Regulation: COBIT, PO4.3 Control Objective: Segregation of Duties Intended State: Prevent unauthorized financial operations. Explanation: Implement a division of roles and […] Regulation: COBIT, PO4.11 IT Architecture Board realized by IT Steering Committee < OrgUnit > realized by Approve Financial Transaction Enriched Enterprise Model loaded from Repository for Reference Models adaption & integration realized by Reference Models for Control Objectives realized by <Information System> ... ASU SIC Internal Controls Cobit 4.1: IT General Controls Control Objective: Identity Management Intended State: Every user can be uniquely identified. Explanation: Ensure that all users (internal, external, and temporary) and their activity on IT systems […] Regulation: COBIT, DS5.3 Figure 8: Reconstructing COBIT 4.1 control objectives as reference models using MEMO ControlML involvement. Hence, the reference models will, in most cases, have to be adapted to the specific organizational use context in a modeling project for internal controls (lower left quadrant). Yet, the overall modeling effort is reduced and, at the same time, compliance with auditing standards and guidelines is fostered due to the reuse of proven artifacts. 8. Conclusion The present work directs the discussion on supporting the assessment of internal control systems through conceptual models to include further abstractions common to enterprise modeling that go beyond business process modeling (i. e., modeling constructs representing internal controls and the corresponding organizational action system). In this context, a DSML for the modeling of internal controls is specified and applied to demonstrate its practical use. The DSML addresses five essential domain-specific requirements: It accounts for the organizational context important to internal control interpretation (R1); explicit and qualified relationships among internal controls (R2); and the multiplicity of control means (R3). The language aids in the perception of a traceable rationale of an internal control by way of explicit (assertion) and implicit (association with risk types) modeling concepts (R4). Further research is required with respect to constructing tailored visual representations meaningful to groups of prospective users (i. e., diagram types and notation elements) (R5). The language’s meta model-based specification and corresponding graphical notation provides abstract 21 and concrete syntax and formal semantics of dedicated modeling constructs for key domain-specific concepts. The demonstration indicates that the ControlML allows for modeling of coherent and consistent conceptual models of internal controls that promise to facilitate interpretation and assessment of internal control systems. In particular, the design artifact contributes to existing work in that it adds essential concepts to internal controls modeling and semantics to key modeling concepts. However, the present conceptualizations in ControlML assume that the precise semantics of concepts promote comprehension by language users. Further research has to study the effects of the presented conceptualizations on prospective users, that is, whether the proposed concepts do indeed lower communication barriers and increase transparency of internal control matters. The number of concepts and the diversity of relations in ControlML suggests that the language itself may initially increase complexity in practice, before complexity is reduced after mastering the language. At present, it is therefore not feasible to predict the effects of MEMO ControlML on the complexity-reducing effects of using a DSML in the assessment of internal control systems and the complexity-increasing effects of the language itself. With regard to auditing practice, enterprise models created with ControlML provide an instrument for audit reviews and function as audit evidence (i. e., as a structured documentation of a firm’s internal control system) to facilitate interpretation and assessment of controls by stakeholders. The feedback received from practicing auditors on the use scenario described in Section 7 indicates at least partial confirmation of this working hypothesis (as does Cendrowski et al., 2007, p. 208). In this respect, the present work addresses the suggested lack of research on internal control assessment methods (Mock et al., 2009, p. 66), and opens paths for future research on such methods: A method based on ControlML requires a corresponding process model to guide the application of language concepts to the assessment of the internal control system. While the application scenario in Section 7 illustrates the principle use of essential language concepts, a corresponding process model should provide more detailed guidelines on how to use the language for specific model-based analyses and assessments. The development of such a process model is on our research agenda. References Bailey, Jr., A. D., Han, K. S., Stansifer, R. D., Whinston, A. B., 2000. The Intelligent Internal Accounting Control Model using a Logic Programming Approach. In: Vasarhelyi, M. A., O’Leary, D. (Eds.), Artificial Intelligence in Accounting and Auditing: Creating value with AI. Vol. 5 of Rutgers Series in Accounting Information Systems. Markus Wiener Publishers, Princeton, NJ, pp. 66–93. Carnaghan, C., 2006. Business process modeling approaches in the context of process level audit risk assessment: An analysis and comparison. International Journal of Accounting Information Systems 7 (2), 170–204. Cendrowski, H., Martin, J. P., Petro, L. W. (Eds.), 2007. The Handbook of Fraud Deterrence. John Wiley and Sons, Hoboken, New Jersey. Chen, K. T., Lee, R. M., 2003. Knowledge-based evaluation of internal accounting control systems — a pattern recognition approach. In: Proceedings of the American Accounting Association Conference. Honolulu, HI. 22 COSO, Sep. 1992. Internal Control — Integrated Framework. The Committee of Sponsoring Organizations of the Treadway Commission. COSO, Sep. 2004. Enterprise Risk Management – Integrated Framework (Executive Summary). The Committee of Sponsoring Organizations of the Treadway Commission. COSO, 2009. Guidance on Monitoring Internal Control Systems: Introduction. The Committee of Sponsoring Organizations of the Treadway Commission. COSO, Sep. 2010. What is internal control? http://www.coso.org/resources.htm, The Committee of Sponsoring Organizations of the Treadway Commission. Crawford, M., Stein, W., 2002. Auditing Risk Management: Fine in Theory but who can do it in Practice? International Journal of Auditing 6 (2), 119–131. Damianides, M., 2005. Sarbanes-Oxley and IT Governance: New Guidance on IT Control and Compliance. Inf. Sys. Manage. 22 (1), 77–85. Davis, C., Schiller, M., Wheeler, K., 2007. IT Auditing : Using Controls to Protect Information Assets. McGraw-Hill. Dunn, C. L., 2006. Business Process Modeling Approaches in the Context of Process Level Audit Risk Assessment: An Analysis and Comparison : Discussion Comments. International Journal of Accounting Information Systems 7 (2), 205–207. Dunn, C. L., Cherrington, J. O., Hollander, A. S., 2005a. Enterprise Information Systems: A Pattern-Based Approach, 3rd Edition. McGraw-Hill Irwin, Boston, MA. Dunn, C. L., Gerard, G. J., Grabski, S. V., 2005b. Critical evaluation of conceptual data models. International Journal of Accounting Information Systems 6 (2), 83–106. Elder, R. J., Beasley, M. S., Arenes, A. A., 2010. Auditing and Assurance Services: An Integrated Approach, 13th Edition. Pearson, Boston et al. Ferstl, O. K., Sinz, E. J., 1998. SOM Modeling of Business Systems. In: Bernus, P., Mertins, K., Schmidt, G. (Eds.), Handbook on Architectures of Information Systems. Springer, Berlin, pp. 339–358. Frank, U., 1994. Multiperspektivische Unternehmensmodellierung: Theoretischer Hintergrund und Entwurf einer objektorientierten Entwicklungsumgebung. Oldenbourg, München. Frank, U., 1998. Essential Research Strategies in the Information Systems Discipline—Reflections on Formalisation, Contingency and the Social Construction of Reality. The Systemist 20, 98–113. Frank, U., 2002. Multi-Perspective Enterprise Modeling (MEMO): Conceptual framework and modeling languages. In: Proceedings of the 35th Annual Hawaii International Conference on System Sciences (HICSS). Honululu, pp. 72–82. Frank, U., 2006a. Evaluation of Reference Models. In: Fettke, P., Loos, P. (Eds.), Reference Modeling for Business Systems Analysis. Idea Group, Hershey, PA, pp. 118–140. Frank, U., 2006b. Towards a Pluralistic Conception of Research Methods in Information Systems Research. ICB Research Report 7, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Germany. Frank, U., 2010. Outline of a Method for Designing Modelling Languages. ICB Research Report 42, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Essen, Germany. URL http://www.icb.uni-due.de/fileadmin/ICB/research/research_reports/ICB-Report-No42.pdf Frank, U., 2011a. MEMO Organisation Modelling Language: Focus on Business Processes. ICB-Research Report 49, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Essen, Germany. URL http://www.icb.uni-due.de/fileadmin/ICB/research/research\_reports/ICB-Report-No49.pdf Frank, U., 2011b. MEMO Organisation Modelling Language: Focus on Organisational Structure. ICB-Research Report 48, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Essen, Germany. URL http://www.icb.uni-due.de/fileadmin/ICB/research/research\_reports/ICB-Report-No48.pdf Frank, U., 2011c. MEMO Organisation Modelling Language (OrgML) – Requirements and Core Diagram Types. ICB-Research 23 Report 47, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Essen, Germany. URL http://www.icb.uni-due.de/fileadmin/ICB/research/research\_reports/ICB-Report-No47.pdf Frank, U., 2011d. The MEMO Meta Modelling Language (MML) and Language Architecture. 2nd Edition. ICB-Research Report 43, Institute for Computer Science and Business Information Systems (ICB), University of Duisburg-Essen, Essen, Germany. URL http://www.icb.uni-due.de/fileadmin/ICB/research/research_reports/ICB-Report_No43.pdf Frank, U., 2012. Multi-Perspective Enterprise Modeling: Foundational Concepts, Prospects and Future Research Challenges. Software and Systems Modeling, available online first, DOI 10.1007/s10270-012-0273-9. Frank, U., 2013. Domain-Specific Modeling Languages - Requirements Analysis and Design Guidelines. In: Reinhartz-Berger, I., Sturm, A., Clark, T., Wand, Y., Cohen, S., Bettin, J. (Eds.), Domain Engineering: Product Lines, Conceptual Models, and Languages. Springer. Frank, U., Heise, D., Kattenstroth, H., Ferguson, D., Hadar, E., Waschke, M., 2009. ITML: A Domain-Specific Modeling Language for Supporting Business Driven IT Management. In: Rossi, M., Gray, J., Sprinkle, J., Tolvanen, J.-P. (Eds.), Proceedings of the 9th OOPSLA workshop on domain-specific modeling (DSM). Geerts, G. L., 2011. A design science research methodology and its application to accounting information systems research. International Journal of Accounting Information Systems 12, 142–151. Gelinas, U. J., Dull, R. B., 2010. Accounting Information Systems, 8th Edition. South-Western Cengage Learning, Mason, OH. Gelinas, U. J., Sutton, S. G., Fedorowicz, J., 2004. Business processes and information technology. South-Western Thomson Learning, Mason, Ohio. Governatori, G., Hoffmann, J., Sadiq, S. W., Weber, I., 2008. Detecting Regulatory Compliance for Business Process Models through Semantic Annotations. In: Ardagna, D., Mecella, M., Yang, J. (Eds.), Business Process Management Workshops. Vol. 17 of Lecture Notes in Business Information Processing. Springer, pp. 5–17. Governatori, G., Milosevic, Z., Sadiq, S., October 2006. Compliance checking between business process and business contracts. In: Proceedings of the 10th IEEE International Enterprise Distributed Object Computing Conference (EDOC’06), Hong Kong, China, Oct 16, 2006. IEEE Computer Society, Los Alamitos, CA, USA, pp. 16–20. IT Governance Institute (Ed.), 2007. CobiT 4.1: Framework, Control Objectives, Management Guidelines, Maturity Models. IT Governance Institute of the Information Systems Audit and Control Association, Rolling Meadows. Karagiannis, D., 2008. A Business process Based Modelling Extension for Regulatory Compliance. In: Bichler, M. et al. (Ed.), Multikonferenz Wirtschaftsinformatik. pp. 1159–1173. Karagiannis, D., Mylopoulos, J., Schwab, M., 2007. Business Process-Based Regulation Compliance: The Case of the SarbanesOxley Act. In: Requirements Engineering : 15th IEEE International Requirements Engineering Conference, RE 2007, October 15-19th, 2007, New Delhi, India. IEEE, pp. 315–321. Köhling, C., 2012. Entwurf einer konzeptuellen Modellierungsmethode zur Unterstützung rationaler Zielplanungsprozesse in Unternehmen. Doctoral dissertation, University of Duisburg-Essen, Institute for Computer Science and Business Information Systems (ICB), Essen, Germany. Lu, R., Sadiq, S. W., Governatori, G., 2007. Compliance Aware Business Process Design. In: ter Hofstede, A. H. M., Benatallah, B., Paik, H.-Y. (Eds.), Business Process Management Workshops. Vol. 4928 of Lecture Notes in Computer Science. Springer, pp. 120–131. Lu, R., Sadiq, S. W., Governatori, G., 2008. Measurement of Compliance Distance in Business Processes. Inf. Sys. Manag. 25 (4), 344–355. Maijoor, S., 2000. The Internal Control Explosion. International Journal of Auditing 4 (1), 101–109. Mock, T. J., Sun, L., Srivastava, R. P., Vasarhelyi, M., 2009. An evidential reasoning approach to Sarbanes-Oxley mandated 24 internal control risk assessment. International Journal of Accounting Information Systems 10 (2), 65–78. Moeller, R. R., 2008. Sarbanes-Oxley Internal Controls : Effective Auditing with AS5, CobiT and ITIL. John Wiley & Sons, Hoboken, NJ. Namiri, K., Stojanovic, N., 2007a. A formal approach for internal controls compliance in business processes. In: 8th Workshop on Business Process Modeling, Development, and Support, BPMDS 2007 in conjunction with CAiSE 2007. Trondheim, Norway. Namiri, K., Stojanovic, N., 2007b. Pattern-based design and validation of business process compliance. In: Proceedings of the 2007 OTM Confederated International Conference on On the move to meaningful internet systems: CoopIS, DOA, ODBASE, GADA, and IS-Volume Part I. Springer, pp. 59–76. Ortner, E., 2008. Language-critical Enterprise and Software Engineering. In: Proceedings of the Fourteenth Americas Conference of Information Systems, AMCIS 2008, Toronto, ON, Canada, Aug 14–17, 2008. Peffers, K., Tuunanen, T., Rothenberger, M. A., Chatterjee, S., 2007. A design science research methodology for information systems research. Journal of Management Information Systems 24 (3), 45–77. Pitthan, J., Philipp, M., 1997. Einsatz von Petri-Netzen fuer die Aufnahme, Dokumentation und Analyse Interner Kontrollsysteme im Rahmen der Jahresabschlusspruefung. In: Stucky, W., Winand, U. (Eds.), Petri-Netze zur Modellierung verteilter DV-Systeme. No. 350. Universitaet Karlsruhe, Karlsruhe, Germany, pp. 87–104. Rikhardsson, P., Best, P., Green, P., Rosemann, M., 2006. Business Process Risk Management, Compliance, and Internal Control: A Research Agenda. Tech. rep., Department of Business Studies, Management Accounting Research Group, Aarhus School of Business, Aarhus, Denmark. Rikhardsson, P., Rohde, C., Rom, A., 2005. Exploring Enterprise Systems and Management Control in the Information Society: Developing a Conceptual Framework. Management Accounting Research Group Working Papers M-2005-05, University of Aarhus, Aarhus School of Business, Department of Business Studies, Aarhus, Denmark. Sadiq, S. W., Governatori, G., Namiri, K., 2007. Modeling Control Objectives for Business Process Compliance. In: Alonso, G., Dadam, P., Rosemann, M. (Eds.), BPM 2007. Springer, Berlin, pp. 149–164. Scheer, A.-W., 1992. Architecture of Integrated Information Systems : Foundations of Enterprise Modelling. Springer, Berlin. Spira, L. F., Page, M., 2002. Risk Management — The reinvention of internal control and the changing role of internal audit. Accounting, Auditing & Accountability Journal 16 (4), 640–661. Strecker, S., Frank, U., Heise, D., Kattenstroth, H., 2012. MetricM: A modeling method in support of the reflective design and use of performance measurement systems. Information Systems and e-Business Management (ISeB) 10 (2), 241–276. Strecker, S., Heise, D., Frank, U., 2010. Toward modeling constructs for audit risk assessment: Reflections on internal controls modeling. In: Esswein, W., Turowski, K., Juhrisch, M. (Eds.), Modellierung betrieblicher Informationssysteme (MobIS 2010). GI, Bonn, pp. 131–148. Strecker, S., Heise, D., Frank, U., 2011a. Prolegomena of a modelling method in support of audit risk assessment: Outline of a domain-specific modelling language for internal controls and internal control systems. Enterprise Modelling and Information Systems Architectures: An International Journal 6 (3), 5–24. Strecker, S., Heise, D., Frank, U., 2011b. RiskM: A multi-perspective modeling method for IT risk assessment. Information Systems Frontiers (ISF). Special Issue on Governance, Risk and Compliance Applications in Information Systems 13 (4), 595–611. Sutton, S. G., Hampton, C., 2003. Risk assessment in an extended enterprise environment: Redefining the audit model. International Journal of Accounting Information Systems 4 (1), 57–73. van der Aalst, W., van Hee, K., van der Werft, J. M., Kumar, A., Verdonk, M., 2011. Conceptual model for online auditing. Decision Support Systems 50, 636–647. Verschuren, P., Hartog, R., 2005. Evaluation in Design-Oriented Research. Qual Quant 39 (6), 733–762. 25 Wand, Y., Monarchi, D. E., Parsons, J., Woo, C. C., 1995. Theoretical foundations for conceptual modelling in information systems development. Dec. Supp. Syst. 15 (4), 285–304. Wand, Y., Weber, R., 2002. Research commentary: information systems and conceptual modeling — A research agenda. Inf. Syst. Res. 13 (4), 363–376. Weigand, H., Elsas, P., 2012. Model-based auditing using REA. International Journal of Accounting Information Systems 13, 287–310. 26