2021 36th IEEE/ACM International Conference on Automated Software Engineering Workshops (ASEW) 2021 36th IEEE/ACM International Conference on Automated Software Engineering Workshops (ASEW) | 978-1-6654-3583-3/21/$31.00 ©2021 IEEE | DOI: 10.1109/ASEW52652.2021.00039 ACSIMA: A Cyber Security Index for Mobile Health Apps Hamza Sellak1 , Mohan Baruwal Chhetri1 , Marthie Grobler1 , Kristen Moore1 , 2 1 CSIRO’s Data61, Australia Security Cooperative Research Centre, Australia Email: {hamza.sellak, mohan.baruwalchhetri, marthie.grobler, kristen.moore}@data61.csiro.au 2 Cyber Abstract—Mobile health (mHealth) apps can make healthcare more accessible and affordable by enabling users to perform a range of self-service activities to manage their own health and wellbeing. However, the mHealth app market is largely unregulated, potentially exposing app users to substantial cyber security risks. In this exploratory study, we present the Australian Cyber Security Index for mHealth Apps (ACSIMA) as a curated cyber security checklist that can guide the assessment of existing mHealth apps as well as the design/development of new apps. In contrast to existing mHealth app assessment frameworks, ACSIMA (1) focuses exclusively on cyber security, (2) takes a multi-stakeholder approach to cyber security assessment, and (3) is specific to the Australian digital health context. ACSIMA is aimed at raising the level of cyber security acuity among different stakeholder groups, including app users and app developers, and can be considered a first step towards the provision of reliable and trustworthy digital health services. We validate ACSIMA’s usability and practicability through an online survey with the different stakeholder groups. The study finds that different stakeholder groups exhibit different levels of familiarity and importance towards the ACSIMA checklist, validating the need for raising the awareness of cyber security concerns in mHealth apps across all stakeholder groups. Index Terms—mHealth apps, security, privacy, checklist I. I NTRODUCTION Advances in smartphone technology and the digitisation of healthcare is enabling more people to use mobile health (mHealth) apps to actively self-manage their health and wellbeing. There are two broad categories of mHealth apps: (a) medical apps used for the purpose of prevention, diagnosis, and treatment of diseases, and (b) non-medical apps related to lifestyle, fitness and wellbeing. In Australia, the Therapeutic Goods Administration (TGA) is the regulatory authority for therapeutic goods, including medical devices, and covers Software as a Medical Device (SaMD), i.e., software that by itself functions as a medical device [1]. Such SaMDs are assessed for compliance against a list of essential principles [2], and audited for conformity to the International Organization for Standardization (ISO) medical devices regulation requirements (ISO-13485). There are specific guidelines to help consumers understand and manage their risks and responsibilities when using the SaMDs [3]. However, the majority of mHealth apps are lifestyle and fitness apps that are considered “simply sources of information, or tools to manage a healthy lifestyle” and therefore do not fall under the TGA’s remit [1]. The market for 978-1-6654-3583-3/21/$31.00 ©2021 IEEE DOI 10.1109/ASEW52652.2021.00039 such non-SaMD mHealth apps is therefore largely unregulated. While published apps are required to meet the quality criteria of the app stores through which they are offered, compliance to TGA regulations is not mandatory. Consequently, many mHealth apps are not subjected to a thorough cyber security assessment and validation process, and can potentially expose app users to substantial security and privacy risks, as highlighted by a recent study of 18,000 popular free apps available on Google Play Store. This study found that ∼50% of the apps were missing a privacy policy even though ∼70% of these were processing personally identifiable information [4]. Researchers have attributed this to apptimism [5], i.e., an implicit trust in smartphones and everything that runs on them. Our preliminary research shows that even though multiple government agencies in Australia are focusing their attention on the privacy and security of health data [6]–[8], there is currently no regulation in place to ensure that mHealth apps adhere to security and privacy principles to keep their users safe from cyber insecurities. As outlined in Section II-B, there are several frameworks that support the design and assessment of mHealth apps; however, they either do not sufficiently address security and privacy, or only target particular stakeholder groups. To address this gap in the intersection between digital health and cyber security practices, our exploratory research study puts forward the Australian Cyber Security Index for mHealth Apps (ACSIMA) as a comprehensive checklist that can help Australian stakeholders, specifically app users, app developers, healthcare providers, and policy makers, to use, design, and develop mHealth apps with more confidence. The index was compiled after an extensive review of published literature on digital health, security and privacy, as well as existing Australian and international standards, regulations and guidelines to create a criteria list that is customised for different stakeholder groups. Following this, an online survey was conducted to answer the following research questions: RQ1: How familiar are the different stakeholder groups with the cyber security criteria included in ACSIMA?, and RQ2: What is their perceived importance of the cyber security criteria included in ACSIMA? The main finding from the survey is that different stakeholder groups show different levels of familiarity and perceived importance towards the ACSIMA checklist. These perceptions are likely shaped by their personal and professional experiences. However, the familiarity is low 145 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. across all stake holder groups, validating the overall need for raising the awareness of cyber security concerns in mHealth apps across all stakeholder groups. The paper is structured as follows. Section II presents an overview of related work in the context of mHealth apps’ standards, regulations and assessment frameworks. Section III describes the methodology followed in compiling and refining ACSIMA’s criteria list. Section IV details the results from our multi-stakeholder study, whilst Section V provides a discussion on our findings. Section VI concludes the paper with an outlook on future work. II. R ELATED W ORK In this section, we briefly overview the key standards and regulations applicable to mHealth apps, and prominent mHealth assessment frameworks proposed in the literature. A. Standards and Regulations mHealth apps that cater to health and lifestyle are largely unregulated in major digital health markets such as the United States (US), Europe and Australia. In the US, the Food and Drug Administration does not regulate health and wellness apps that are not intended for medical use despite an increasing push for it to play a much stronger role [9]. In Australia, the TGA does not regulate health and lifestyle apps and software that do not fit their definition of a medical device [1]. There are some progress in Europe with the European Commission facilitating the Privacy Code of Conduct [10] to provide app developers with practical guidance on data protection principles when developing mHealth apps. However, it is yet to achieve formal approval of European data authorities. Similarly, the Office of the Australian Information Commissioner (OAIC) promotes the Australian Privacy Principles (APPs) [11] to govern the standards, rights, and obligations around the handling of personal information and are therefore very important in the context of mHealth apps. It also provides a guide for app developers operating in the Australian market to comply with Australian privacy law and best practices [12]. Also, the Australian Commission on Safety and Quality in Healthcare has recently released the National Safety and Quality Digital Mental Health Standards [13] in partnership with key stakeholders, including mental health service users, consumers, carers, families, and technical experts. The standards apply to digital mental health services and cover technical governance including privacy, security, and transparency of digital systems, but is yet to extend to cover all mHealth apps. Similarly, the Australian Digital Health Agency, which plays a key role in the adoption and use of digital health services and technologies, provides only basic guidance to users on how to stay private and secure online but none regarding the use of mHealth apps. Notwithstanding the existence of different government entities and their ongoing efforts looking at different aspects of digital health, the mHealth app market is still largely unregulated. Therefore, there is a strong need for tools and frameworks to help users navigate the mHealth app market. B. mHealth Assessment Frameworks Several cyber security assessment frameworks and tools have been proposed to cater for the needs of different stakeholders. The Interactive Mobile App Review Toolkit (IMART) [14] is a framework for reviewing mHealth apps and is aimed at assisting clinicians select healthcare apps. However, it offers little consideration for security and privacy. Similarly, the American Psychiatric Association (APA) App Evaluation Framework [15], which is targeted at patients with psychiatric conditions and clinicians treating them, provides cursory treatment of cyber security concerns. The Technology Evaluation and Assessment Criteria for Health Apps (TEACHApps) [16] is a structured 4-step process for evaluating mHealth apps built on top of the APA framework and is targeted at hospitals and health care facilities. The GeoHealth Privacy Impact Assessment (PIA) framework [17], which identifies and evaluates the potential privacy risks of mHealth data collection systems, is targeted at community health workers. The Mobile App Rating Scale (MARS) [18], [19] is a tool for classifying and assessing the quality of mHealth apps; it is targeted at researchers, developers and health professionals. The Mobile App Development and Assessment Guide (MAG) [20] provides a set of criteria that can help health care providers, developers, patients, and other stakeholders to guide the development and assessment of mHealth apps. THESIS [21] is a rating tool to objectively measure the risks and benefits of mHealth apps. This tool can be used by policy makers, app developers, patients and healthcare providers. Finally, the Organisation for the Review of Care and Health Applications - 24 Questions (ORCHA-24) framework [22] consists of 24 app assessment criteria that are derived from published standards on mHealth apps, data privacy and user experience and presented as ‘yes’ or ‘no’ questions without the requirement for scaling or value judgements. There is a lot of diversity in existing assessment frameworks; some are targeted at specific stakeholders (e.g., [14]– [16]) while others cater to multiple stakeholders (e.g., [17], [18], [20]–[22]). However, most of them fail to sufficiently address cyber security concerns (e.g., [18]), while some provide basic coverage only (e.g., [15], [21], [22]). C. Distinction from Prior Work All stakeholders have a shared responsibility in addressing the cyber security issues associated with medical devices, and by extension, mHealth apps. Yet, a key limitation of existing mHealth app assessment frameworks is that they fail to acknowledge that different stakeholders may have different cyber security concerns, and may be differently impacted by them. ACSIMA addresses this shortcoming by providing customised cyber security checklists for the identified key stakeholder groups. Furthermore, ACSIMA complements and covers existing Australian standards and regulations. III. M ETHODOLOGY The main objective of this study was to generate a curated cyber security checklist that can be used as an index to guide 146 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. TABLE I: Source Documents for Index Generation Source Document Literature Baumel et al. [25] Benjumea at al. [26] Bandaronek et al. [27] Camacho et al. [16] Huckvale et al. [28] Huckvale et al. [29] Hutton et al. [30] Karekla et al. [31] Llorens-Vernet and Miró [20] O’Loughlin et al. [32] Scott et al. [33] Other ADHA [34] ADHA [35] ANDHealth [24] APP [11] IMDRF [23] NSQDMH [13] OAIC [12] TGA [1] in the process. The initial criteria list was refined through multiple iterations resulting in a final list of 50 criteria. Type Systematic review Scoping review App review Primary research App review App review App review Primary research Systematic review App Review Literature review B. Index Customisation Following the finalisation of the criteria list, a customised version was generated for each stakeholder group by considering which criteria would be more relevant and applicable to each group. Thereafter, the Delphi method was used to scale down the customised lists to a more manageable number with the research team voting for the inclusion/exclusion of each criterion for each stakeholder group. The following specifications were considered for each group: Guide Guide Report Law Guide Standard Guide Regulation • • the evaluation of existing mHealth apps and the design and development of new ones. This section provides an overview of how the checklist was generated and validated for usability and applicability. • A. Index Generation The cyber security checklist was generated in two phases using two main sources of information. In the first phase, an initial criteria list was drafted by searching through published academic literature, using Web of Science and Google Scholar, for studies that focused on identifying standards for mHealth apps. Starting with an initial set of articles, snowballing was used to identify further relevant articles. Key terms used in the search include mobile health app, mHealth app, privacy, security, cyber security, evaluation and assessment framework. In the second phase, different Australian and International cyber security regulations and other online guidelines were examined to complement the criteria list. This includes the regulations provided by TGA [1], the guidelines provided by ADHA, the NSQDMH standards from the Australian Commission on Safety and Quality in Healthcare [13], the guidelines from the International Medical Device Regulators Forum (IMDRF) [23], and research reports by Australia’s National Digital Health Initiative (ANDHealth) [24]. The initial list compiled after analysing the sources listed in Table I had 144 criteria. It covers a broad spectrum of cyber security practices, guidelines and recommendations, and is applicable to a range of stakeholder groups. These criteria were separated into two classes (CLs): CL1: Design and Development criteria, which may be considered during the design and development phases of mHealth apps, and CL2: Evaluation criteria, which may be considered during the evaluation of existing mHealth apps. A thorough review of this list was conducted by the research team to remove any duplicate criteria and to merge similar criteria. 26 redundant criteria were removed, and a further 86 criteria were merged • • • Policy makers (PM) are responsible for formulating policies in health and/or cyber security. The customised list should focus on criteria that belong to both CL1 and CL2, balancing both technical1 and non-technical criteria. Health Professionals (HP) provide health care treatment and advice based on formal training and experience. The customised list should focus on criteria that belong to CL2 and balance both technical/non-technical criteria. Researchers (R) conduct research in the digital health and/or cyber security domains. The customised list should focus on criteria that belong to both CL1 and CL2, and put more emphasis on technical criteria. Software Engineers/App Developers (SE/AD) design and develop software and digital (health) solutions. The customised list should focus on criteria that belong to CL1, and put more emphasis on technical criteria. Consumers (C) refer to those who are currently using or have previously used an mHealth app. The customised list should focus on criteria that belong to CL2, and should avoid technical criteria. Others (O) that do not fall under any of the previous stakeholder groups. The customised list should focus on criteria that belong to CL2, and avoid technical criteria. C. Index Validation An exploratory user study was conducted to validate the perceived usability and applicability of the developed cyber security index. An online survey was conducted2 targeting the identified stakeholder groups and using the corresponding customised criteria list. Table II provides a summary of the different sections of the survey and the corresponding questions for each stakeholder group. Participants were specifically asked to comment on their familiarity and perceived importance of each criterion. The multi-stakeholder feedback was incorporated to ensure the relevance and usability of the index within the targeted stakeholder groups. 1 Technical criteria describe processes related to cyber security and/or software engineering such as encryption, security controls, and risk management. 2 Following approval by the organisational ethics review board, the online survey was conducted through SurveyMonkey3 during December 2020 and January 2021. 147 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. TABLE II: Survey Section and Question Overview Stakeholder Group PM HP R SE/AD C O TABLE III: Sociodemographic Characteristics of the Stakeholder Groups Section [Question(s)] Sociodemographics [Q1-Q3] Expertise [Q4] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q11] Sociodemographics [Q1-Q3] Expertise [Q5] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q12] Sociodemographics [Q1-Q3] Expertise [Q6] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q13] Sociodemographics [Q1-Q3] Expertise [Q7] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q14] Sociodemographics [Q1-Q3] Expertise [Q8] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q15] Sociodemographics [Q1-Q3] Cyber security risks awareness [Q9] Cyber security & mHealth apps [Q10] Customised ACSIMA [Q16] PM 5 (4%) HP 16 (12%) PM 5 (4%) HP 16 (12%) 18-25 26 (20%) 26-35 40 (30%) Stakeholder Group R SE/AD C O 42 (32%) 34 (25%) 17 (13%) 18 (14%) Gender R SE/AD C O 42 (32%) 34 (25%) 17 (13%) 18 (14%) Age Group 36-45 46-55 56-65 >65 PNTA 39 (29%) 8 (6%) 10 (8%) 6 (5%) 3 (2%) TABLE IV: Expertise of the Stakeholder Groups PM Have you ever been part of a team that prepared a policy related to (digital) health and/or cyber security? No 2 (50%) Yes, I am/was a team leader 2 (50%) Yes, I am/was a team member 1 (25%) HP What is your exact role as a health professional? Medical specialist 1 (6%) Allied health professional 11 (69%) Other 5 (31%) R What are your research focus areas and expertise? Health research 12 (30%) Digital health 14 (35%) Cyber security 21 (53%) Computer science (in general) 14 (35%) Other 2 (5%) SE/AD How do you describe your expertise as a software engineer/app developer? I am/was involved in the development of at least one 7 (22%) mHealth app (or similar digital health services) I am/was involved in the development of non-health23 (72%) related mobile apps Other 6 (19%) C Have you ever used any technology medium to help you achieve your health and/or wellbeing goals? No 1 (6%) Yes, I used an activity tracker watch 14 (82%) Yes, I used a medical wearable device 1 (6%) Yes, I joined an online health program or group about 2 (12%) a health issue Yes, I entered information on a website about my 13 (76%) physical activity, nutrition, sleep and/or wellbeing Yes, I searched for information about a disease, health 13 (76%) problem or a doctor Yes, I used a smart home environment for healthcare 2 (12%) and assistance D. Limitations The purpose of this research study was to develop an index for mHealth apps that focuses exclusively on cyber security, considers multiple stakeholders and is mainly applicable to the Australian digital health context. An important consideration in this research is that individuals’ perceptions of security and privacy are influenced by their own personal and professional experiences, their technological innovations and their role within the digital domain. All of these aspects can therefore play a role in influencing the participants’ responses to our survey. We found that there was not enough representation across all the different stakeholder groups. In particular, the policy makers group was under represented. We are therefore unable to define conclusive findings on the perceived usability and applicability of the index for this specific group. IV. R ESULTS This section presents the results from our exploratory survey. We report on the results of each survey section using statistical metrics such as count (N ), percentage (%), mean (µ), standard deviation (σ), and coefficient of variation. We also summarize the key findings. A. Sociodemographics All participants were required to complete a set of sociodemographic questions (Q1-Q3 in Table II) pertaining to gender, age and self-classification. Depending upon the participants’ response to self-classification, logic branching was applied to direct them to subsequent questions related to their selected stakeholder group. Table III provides the summary statistics of the 132 participants’ responses. B. Expertise All participants except those in the other group were required to respond to the expertise section of the survey (Q4-Q8 in Table II). The key findings include: (1) most of the stakeholder groups demonstrated a connection with digital health and/or mHealth apps; (2) despite their small number, most policy makers indicated that they have been involved in the making of at least one policy related to (digital) health and/or cyber security; (3) one role (allied health 148 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. 100 60 40 20 0 No, not at all Don t know/can t say Yes, definitely Yes, to some extent 80 percentage 80 percentage 100 Poor Fair Good Very good 60 40 20 PM HP R SE/AD Participant group C 0 O Fig. 1: Distribution of Cyber Security Risk Awareness. PM HP R SE/AD C Participant group O Fig. 2: Distribution of the Assumption that Cyber Security is Built into mHealth apps. professional) was dominant among health professionals with zero participation from general practitioners, dentists or oral health professionals, complementary medicine practitioners, and other medical specialities; (4) there was relatively equal participation from researchers in cyber security, digital health and health research with many specialising in two or more research domains; and (4) participants from the consumer stakeholder group indicated that all of them are tech-savvy and had medium interactions with different technology mediums that help to achieve health and/or wellbeing goals. Table IV provides the summary statistics of the responses of the 132 participants. C. Cyber Security Risk Awareness All participants were asked a question related to cyber security risk awareness (Q9 in Table II). Around 95% of the participants (N = 125) answered this question with 40% (N = 50) rating their current awareness of cyber security risks as “Good”, 34% (N=43) as “Fair”, 16% as “Very good”, and 10% (N=12) as “Poor”. None of the participants rated their current awareness of cyber security risks as “Very poor”. Figure 1 presents the overall distribution of ratings for each stakeholder group. At a group level, ratings from the software engineers/app developers and researchers were slightly higher than the rest of the groups. The overall rating provided by health professionals can be considered closer to “Poor”. The majority of consumers indicated that they have overall knowledgeable awareness of cyber security risks. D. Cyber Security and mHealth Apps All participants were asked if they thought that cyber security is built into all mHealth apps (Q10 in Table II). 95% of participants answered the question with 41% (N = 51) indicating that they think cyber security is built into all available mHealth apps to some extent, 38% (N = 48) disagreed with this assumption, 16% (N = 16) indicated that they do not know or cannot say, and only 5% (N = 6) strongly supported the assumption (i.e., answered with “Yes, definitely”). Figure 2 presents the distribution of responses of each stakeholder group to the assumption that cyber security is built into all available mHealth apps. Key conclusions from the responses to Q10 are: (1) there was consensus that cyber security is not naturally built into all mHealth apps; (2) health professionals were more supportive of the assumption, whereas researchers expressed less acceptance of the assumption in comparison to the rest of the stakeholder groups; (3) almost half of the software engineers/app developers indicated that they “don’t know/can’t say” which is justified by the fact that only 22% (N = 7) are/were involved in the development of at least one mHealth app or similar digital health services (see Section IV-B); and (4) consumers and other stakeholders had relatively similar and equal distributions over both “Yes” and “No” answers. E. Customised ACSIMA The customised ACSIMA section was presented to all stakeholder groups, respectively as Q11-16 (see Table II). Only 47% of the participants completed this section of the survey. Since we did not have enough representation from the policy makers group, we excluded them in this section. We compiled a separate criteria list for each stakeholder group and asked the participants to consider: (a) how familiar they are with each criterion using a 0–5 scale (0: Not familiar at all – 5: Extremely familiar) and (b) how important each criterion is to them using a 0–5 scale (0: Not important at all – 5: Extremely important). Table A1 presents the comprehensive criteria list. Table V lists the top 10 criteria according to each stakeholder group, ranked in terms of familiarity and perceived importance. This table gives an indication of how relevant stakeholders consider the criteria to be for their use of the apps. Table VI shows the overall mean, standard deviation, and coefficient of variation given by each stakeholder group to their criteria list. If two criteria have the same mean, the one with the lowest standard deviation4 is considered better. The last 4 The standard deviation describes how much an individual measurement should be expected to deviate on average from the mean. A smaller standard deviation means greater consistency, predictability and quality. 149 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. TABLE V: Top 10 Criteria based on Familiarity and Perceived Importance (for each Stakeholder Group) HPF C8 C23 C14 C6 C15 C25 C16 C1 C4 C24 HPI C4 C17 C6 C23 C14 C15 C25 C19 C27 C26 RF C23 15 C9 C8 C34 C32 C16 C33 C12 C6 RI C15 C23 C29 C4 C6 C34 C32 C9 C13 C39 SE/ADF C33 C8 C12 C17 C41 C32 C42 C43 C15 C16 SE/ADI C9 C32 C33 C12 C13 C16 C4 C15 C6 C17 CF C16 C24 C25 C13 C47 C9 C12 C8 C11 C23 CI C13 C16 C9 C23 C24 C17 C25 C15 C21 C8 OF C23 C25 C47 C12 C8 C9 C16 C6 C10 C24 OI C23 C8 C9 C25 C4 C21 C24 C6 C17 C13 TABLE VI: Mean, Standard Deviation, and Coefficient of Variation of Responses (for each Stakeholder Group) Familiarity µF HP 2.935 R 3.232 SE/AD 3.106 C 2.995 O 3.108 σF 1.164 1.452 1.675 1.511 1.278 Importance CVF 0.396 0.449 0.539 0.504 0.411 µI 4.447 4.087 3.841 3.695 3.869 σI 0.806 1.035 1.221 1.430 1.065 |µF − µI | |σF − σI | CVI 0.181 0.253 0.317 0.387 0.275 1.511 0.854 0.735 0.700 0.769 0.807 0.416 0.454 0.080 0.212 two columns (|µF − µI | and |σF − σI |) show the discrepancy between overall expressed familiarity and perceived importance. The best and worst results of each column are highlighted in bold font and underlined, respectively. Researchers gave the highest overall rating regarding familiarity with criteria (µF = 3.232), health professionals gave the lowest overall rating (µF = 2.935) and showed less variation and greater consistency (σF = 1.164), whereas software engineers/app developers have the highest variation and less consistency (σF = 1.675) in comparison to the rest of the stakeholder groups. In terms of perceived importance, health professionals provided the highest overall rating (µI = 4.447) with less variation and greater consistency (σI = 0.806), whereas consumers provided the lowest overall rating (µI = 3.695) with higher variation and less consistency (σI = 1.430). These results are confirmed by each stakeholder group CV5 values for both familiarity (CVF ) and perceived importance (CVI ). Consumers followed by software engineers/app developers (|µF − µI | = 0.70 and |µF − µI | = 0.735, respectively) demonstrated less discrepancy, whereas health professionals demonstrated the highest discrepancy (|µF − µI | = 1.511). These results are confirmed by each stakeholder group’s |σF − σI | value from the last column (i.e., the smaller the value the greater the consistency between the familiarity and perceived importance). The key findings to consider from this section are: (1) across all stakeholder groups, participants rated the criteria that were of a non-technical or legal/regulatory nature higher than the more technical criteria; (2) criteria that focus on privacy, such as C4, C6, C9 and C23 (see Table A1) were perceived as being more important by all stakeholder groups – they appeared in the top five of at least three of the 5 The coefficient of variation shows the extent of variability in relation to the population mean. A smaller coefficient of variation means less variability. customised lists; (3) the inconsistency between familiarity and perceived importance in some stakeholder groups is likely due to their lack of cyber security awareness – e.g., the majority of researchers stated that they specialised in cyber security and expressed higher familiarity with the criteria, whereas health professionals expressed less familiarity and higher perceived importance which aligns with their overall “Poor” to “Fair” cyber security risks awareness and their strong support of the assumption that cyber security is built into all available mHealth apps; and finally, (4) the less familiarity participants have with a criterion, the higher their perceived importance of that criterion – e.g., health professionals. V. D ISCUSSION In this study, we developed ACSIMA, a cyber security index for mHealth apps, by compiling a comprehensive set of criteria and then iteratively refining it to generate a curated list. We created customised versions of the list for each of the target stakeholder groups and performed an online survey to explore the following two questions: • RQ1: How familiar are the different stakeholder groups with the key cyber security criteria included in ACSIMA? Some stakeholder groups are more familiar and aware of the security and privacy criteria included in ACSIMA. For example, the cyber security researchers and software engineers/app developers had a higher mean familiarity rating compared to the consumers and health professionals. However, the mean familiarity ratings for all stakeholder groups were low indicating that there is an overall need for raising the awareness of cyber security concerns in mHealth apps across all stakeholder groups. • RQ2: What is their perceived importance of the key cyber security criteria included in ACSIMA? Even though the different stakeholder groups indicated different levels of familiarity with the cyber security concerns in mHealth apps, they all considered the criteria included in their customised checklists to be important. As expressed by one of the participants, “there were many criteria that I was not formally aware of, but which I thought were important once brought to my attention”. This again highlights the need for increased awareness of cyber security concerns across all stakeholder groups. ACSIMA aims to help designers and developers to deliver mHealth apps that are reliable and supported by secure and privacy-aware technology; and educate end-users about the potential cyber security risks of mHealth apps. The development of ACSIMA is a precursor to transforming the state-of-theart of digital health and wellbeing interventions. The wider implementation of the index has the potential to increase cyber security awareness among all stakeholders, leading to increase stakeholder engagement, which in turn would enable enhanced security, privacy and transparency in mHealth apps. VI. C ONCLUSION Mobile health (mHealth) apps have the potential to make healthcare more accessible and affordable for everyone, but 150 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. the current market is largely unregulated. Most apps available in major app stores are not subjected to a thorough cyber security assessment and validation process, leaving the end-users vulnerable to cyber insecurities. We developed the Australian Cyber Security Index for Mobile Health Apps (ACSIMA) as a first step towards the transition to trustworthy mHealth apps. In contrast to existing assessment frameworks, ACSIMA (1) exclusively focuses on cyber security, (2) provides customised checklists for different stakeholders groups in the digital health domain, and (3) is specific to the Australian context. Although ACSIMA is currently in a prototype stage, still to be considered by representative stakeholder groups in a large collaborative setting, it has the potential to not only educate about digital health services, their benefits and potential cyber security risks, but also to build trust and increase engagement and adherence to personal health and wellbeing management services. ACKNOWLEDGMENT This research was financially supported by CSIRO’s Precision Health Future Science Platform. R EFERENCES [1] Therapeutic Goods Administration, “Regulation of software based medical devices,” https://tinyurl.com/tag-regulation-2020, 2020, [Online; Accessed 3 September 2021]. [2] ——, “How the TGA regulates software-based medical devices,” https://tinyurl.com/tga-samd2021, 2021, [Online; Accessed 3 September2021]. [3] ——, “Medical device cyber security information for users,” https://tinyurl.com/tga-md-cs2019, 2019, [Online; Accessed 3 September 2021]. [4] Zimmeck, Wang et al., “Automated analysis of privacy requirements for mobile apps,” in NDSS, 2017. [5] Wyatt, “How can clinicians, specialty societies and others evaluate and improve the quality of apps for patient use?” BMC medicine, vol. 16, no. 1, pp. 1–10, 2018. [6] Australian Digital Health Agency, “Safe, seamless and secure: Evolving health and care to meet the needs of modern Australia,” https://tinyurl.com/adha-plan-2017, 2020, [Online; Accessed 3 September 2021]. [7] Office of the Australian Information Commissioner, “Guide to securing personal information,” https://tinyurl.com/oaic-2018, 2018, [Online; Accessed 3 September 2021]. [8] ——, “Guide to health privacy,” https://tinyurl.com/oaic-2019, 2019, [Online; Accessed 3 September 2021]. [9] Kasperbauer and Wright, “Expanded FDA regulation of health and wellness apps,” Bioethics, vol. 34, no. 3, pp. 235–241, 2020. [10] European Commission, “Privacy Code of Conduct on mobile health apps,” https://tinyurl.com/privacy-code-conduct, 2020, [Online; Accessed 10 February 2021]. [11] Office of the Australian Information Commissioner, “Australian Privacy Principles,” https://tinyurl.com/oaic-app, 2014, [Online; Accessed 3 September 2021]. [12] ——, “Mobile privacy: A better practice guide for mobile app developers,” https://tinyurl.com/oaic-guide2014, 2014, [Online; Accessed 3 September 2021]. [13] Australian Commission on Safety and Quality in Health Care, “National safety and quality digital mental health standards,” https://tinyurl.com/safetyquality2020, 2020, [Online; Accessed 3 September 2021]. [14] Maheu, Nicolucci et al., “The interactive mobile app review toolkit (IMART): A clinical practice-oriented system,” Journal of Technology in Behavioral Science, vol. 1, no. 1, pp. 3–15, 2016. [15] Torous, Chan et al., “A hierarchical framework for evaluation and informed decision making regarding smartphone apps for clinical care,” Psychiatric Services, vol. 69, no. 5, pp. 498–500, 2018. [16] Camacho, Hoffman et al., “Technology evaluation and assessment criteria for health apps (TEACH-Apps): Pilot study,” Journal of Medical Internet Research, vol. 22, no. 8, p. e18346, 2020. [17] Iwaya, Fischer-Hübner, Åhlfeldt, and Martucci, “Mobile health systems for community-based primary care: Identifying controls and mitigating privacy threats,” JMIR mHealth and uHealth, vol. 7, no. 3, p. e11642, 2019. [18] Stoyanov, Hides et al., “Mobile app rating scale: A new tool for assessing the quality of health mobile apps,” JMIR mHealth and uHealth, vol. 3, no. 1, p. e27, 2015. [19] Terhorst, Philippi et al., “Validation of the mobile application rating scale (MARS),” Plos one, vol. 15, no. 11, p. e0241480, 2020. [20] Llorens-Vernet and Miró, “Standards for mobile health–related apps: Systematic review and development of a guide,” JMIR mHealth and uHealth, vol. 8, no. 3, p. e13057, 2020. [21] Levine, Co et al., “Design and testing of a mobile health application rating tool,” NPJ Digital Medicine, vol. 3, no. 1, pp. 1–7, 2020. [22] Leigh, Ouyang, and Mimnagh, “Effective? Engaging? Secure? Applying the ORCHA-24 framework to evaluate apps for chronic insomnia disorder,” Evidence-based mental health, vol. 20, no. 4, pp. e20–e20, 2017. [23] International Medical Device Regulators Forum (IMDRF), “Principles and practices for medical device cybersecurity,” https://tinyurl.com/imdrf-2019, 2020, [Online; Accessed 3 September 2021]. [24] Australia’s National Digital Health Initiative, “Digital Health: The sleeping giant of Australia’s health technology industry,” https://tinyurl.com/andhealth-2020, 2020, [Online; Accessed 16 February 2021]. [25] Baumel, Faber, Mathur, Kane, and Muench, “Enlight: A comprehensive quality and therapeutic potential evaluation tool for mobile and webbased eHealth interventions,” Journal of Medical Internet Research, vol. 19, no. 3, p. e82, Mar 2017. [26] Benjumea, Ropero, Rivera-Romero, Dorronzoro-Zubiete, and Carrasco, “Privacy assessment in mobile health apps: Scoping review,” JMIR mHealth and uHealth, vol. 8, no. 7, p. e18868, 2020. [27] Bondaronek, Alkhaldi, Slee, Hamilton, and Murray, “Quality of publicly available physical activity apps: Review and content analysis,” JMIR mHealth and uHealth, vol. 6, no. 3, p. e53, Mar 2018. [28] Huckvale, Prieto, Tilney, Benghozi, and Car, “Unaddressed privacy risks in accredited health and wellness apps: A cross-sectional systematic assessment,” BMC Medicine, vol. 13, no. 1, Sep 2015. [29] Huckvale, Torous, and Larsen, “Assessment of the data sharing and privacy practices of smartphone apps for depression and smoking cessation,” JAMA Network Open, vol. 2, no. 4, p. e192542, Apr 2019. [30] Hutton, Price et al., “Assessing the privacy of mHealth apps for selftracking: Heuristic evaluation approach,” JMIR mHealth and uHealth, vol. 6, no. 10, p. e185, Oct 2018. [31] Karekla, Kasinopoulos et al., “Best practices and recommendations for digital interventions to improve engagement and adherence in chronic illness sufferers,” European Psychologist, vol. 24, no. 1, pp. 49–67, Jan 2019. [32] O'Loughlin, Neary, Adkins, and Schueller, “Reviewing the data security and privacy policies of mobile apps for depression,” Internet Interventions, vol. 15, pp. 110–115, Mar 2019. [33] Scott, Richards, and Adhikari, “A review and comparative analysis of security risks and safety measures of mobile health apps,” Australasian Journal of Information Systems, vol. 19, Nov 2015. [34] Australian Government Department of Health, Australian Digital Health Agency (ADHA), “Cloud services: Considerations for healthcare organisation,” https://tinyurl.com/dh-cloud-2020, 2020, [Online; Accessed 3 September 2021]. [35] ——, “Selecting secure IT products and services: Questions to ask your IT vendors. A companion for the information security guide for small healthcare businesses,” https://tinyurl.com/adha-202, 2020, [Online; Accessed 3 September 2021]. 151 Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. TABLE A1: ACSIMA Criteria List Stakeholder Group # Criterion Authorized licensed use limited to: Nazarbayev University. Downloaded on April 10,2023 at 05:25:27 UTC from IEEE Xplore. Restrictions apply. PM C1. The app is developed by a trusted developer. 3 C2. The app has an Australian sponsor to be accountable for its quality, safety or performance in case the app is not Australian. 3 C3. The app declares whether it has been subject to a regulatory scrutiny or not. 3 C4. The app takes reasonable steps to ensure compliance with Australian applicable legislation, including: (1) Privacy Act 1988; (2) State and Territory Privacy Legislation; and (3) My health record 3 legislation. C5. The app has the ability to meet digital health regulation in major markets overseas. 3 C6. The app has a complete and detailed privacy policy. 3 C7. The app explains the security systems used, including documented security risks and mitigations. 3 C8. The app has been tested by potential end-users before being made available to the public. 3 C9. The app declares the data to be collected including personal identifiable information (PII) and purpose(s) for collection. 3 C10. The app ensures the right of easy access to collected information. 3 C11. The app gives detailed information about any kind of agreements with third parties. 3 C12. The app requires users to always give their explicit consent before any action is taken. 3 C13. The app instantly notifies users of any cyber security risks when they become known. 3 C14. The app guarantees secure collection, storage and exchange of information. 3 C15. The app declares who owns the data that is collected. 3 C16. The app declares whether it allows the users: (1) to opt out of data collection or not; (2) to delete previously collected data or not; and (3) to share the data collected with other entities or not. 3 C17. The app developer has a data breach response plan that includes documented processes for taking appropriate action if a suspected or actual data breach occurs. 3 C18. The app guarantees that in case it lacks sufficient fundamental or inherent protective measures, and updates are not feasible, risk-mitigating alternatives will be applied as compensating controls. 3 C19. The app developer guarantees that any vulnerability will be corrected and disclosed in all jurisdictions in which the app operates. 3 C20. The app developer always coordinates with the regulatory authority to determine if a software change requires approval prior to release. 3 C21. The app provides contact information for end-users to report cyber security incidents and events, or at least the users can still submit through regular customer support channels. 3 C22. The app developer guarantees that cyber security incidents which impact user safety and privacy will be reported to applicable regulatory agencies as required by regulation. When a criminal activity is 5 identified through the course of investigation, local and applicable law enforcement agencies will be notified. C23. The app uses the data collected only for the purpose that is explicitly declared. 5 C24. The app gives users control over their data. 5 C25. The app guarantees the availability of security updates when needed, and states if an update modifies or affects how it will function. 5 C26. The app developer provides basic cyber security training through the app to create security awareness and introduce cyber hygiene practices among all users. 5 C27. The app developer guarantees that healthcare providers have access to any information needed to implement a recommendation, and to ensure the protection of their patients. 5 C28. The app declares potential software-related risks and states the risk classification (if known). If the app is classified as high-risk, it has to additionally declare whether it was subject to third-party 5 oversight or not. C29. At the verification and validation stage in the design and development process of the app, various types of security testing were employed to provide assurance that the code is free of significant 5 known vulnerabilities and that security controls have been effectively implemented. C30. The app guarantees periodic security testing that can include: static code analysis, dynamic analysis, robustness testing, vulnerability scanning and software composition analysis. 5 C31. The app identifies cyber security vulnerabilities, detects possible threats, and applies security measures to minimise the risks. 5 C32. The app implements an encryption mechanism to secure collection, storage and exchange of information. 5 C33. The app implements reliable authorisation and authentication mechanisms. 5 C34. The app secures data transfer to prevent unauthorised access, modification, or replay. 5 C35. The app declares if confidentiality risk control measures are required to protect message control/sequencing fields in communication protocols or to prevent the compromise of cryptographic keying 5 materials. C36. If the app is open source, the main developers declare how it will be updated and controlled. 5 C37. The app implements mechanisms to detect, resist, respond and recover from cyber security attacks in order to maintain its essential performance. 5 C38. Documentation is available that shows the app architecture, and where the data will be stored (e.g., user information is stored within Australia, to meet privacy requirements). 5 C39. The app is designed and developed under a secure development framework to minimize the number of legacy components in the future, and at a minimum, it meets a security baseline and includes 5 mechanisms for updates and patches. C40. The app integrates a cyber security risk management process that: (1) identifies any cyber security vulnerability; (2) estimates and evaluates the associated risks; (3) controls those risks to an acceptable 5 level; (4) assesses and monitors the effectiveness of the risk controls; and (5) communicates risks via coordinated disclosure to key stakeholders. C41. The app declares whether it stores data on the device, on the cloud or on both. 5 C42. The app explicitly states terms and conditions of the used cloud service, and makes sure that relevant security measures are present. 5 C43. Every time an update of a third-party component is published, the changes are inspected, and the risk is revaluated. 5 C44. As cyber security threats will continuously evolve, the app proactively monitors, identifies and addresses vulnerabilities and exploits as part of its cyber security management plan across the total 5 product life cycle. C45. The app implements a formalized process for gathering information from vulnerability finders, developing mitigation and remediation strategies, and disclosing the existence of vulnerabilities and 5 mitigation or remediation approaches to stakeholders. C46. The app developer documents and summarises their activities related to cyber security including test reports that describe all tests performed to verify the security of the app and the effectiveness of 5 any security controls. C47. The app developer uses plain language, at an appropriate reading level for the intended user, to communicate actionable information regarding cyber security vulnerabilities and threats. 5 C48. The app allows anonymised sharing of information to improve security and user safety. The app explicitly declares that the shared information is not to be used to gain a commercial advantage. 5 C49. The app applies appropriate measures to protect minor users in accordance with applicable legislations. 5 C50. The app developer guarantees to continuously inform consumers, healthcare providers and regulators about any problem in released software versions and how to obtain and install changes. 5 152 N HP R SE/AD C O 3 3 5 3 5 5 5 3 5 5 5 3 3 3 5 3 3 3 5 3 4/6 4/6 1/6 6/6 5 3 5 3 5 3 5 5 5 3 3 3 3 5 3 5 3 5 3 3 3 3 3 3 5 3 3 5 3 3 3 5 5 5 3 3 3 3 3 3 3 3 3 3 3 5 3 3 3 5 3 5 3 3 5 3 3 3 3 3 3 3 3 5 3 3 3 5 5 5 3 5 5 3 3 3 3 3 3 3 3 5 3 3 3 5 5 5 3 5 3/6 6/6 5/6 6/6 5/6 6/6 4/6 5/6 5/6 2/6 6/6 6/6 6/6 1/6 3/6 1/6 6/6 2/6 3 3 3 3 3 5 3 5 3 3 5 3 5 5 3 5 5 3 3 3 3 3 5 3 3 3 3 3 5 3 4/6 3/6 5/6 4/6 1/6 4/6 5 3 5 5 5 1/6 5 5 5 5 5 5 3 3 3 3 3 3 3 3 3 3 5 5 5 5 5 5 5 5 5 5 5 5 5 5 2/6 2/6 2/6 2/6 1/6 1/6 5 5 5 5 3 3 3 3 3 5 5 3 5 5 5 5 5 5 5 5 2/6 1/6 1/6 2/6 5 5 3 5 5 1/6 5 5 5 5 5 5 5 5 3 3 3 3 5 5 5 5 5 5 5 5 1/6 1/6 1/6 1/6 5 5 3 5 5 1/6 5 5 3 5 5 1/6 5 5 5 5 5 5 5 5 3 3 3 3 3 5 3 3 3 5 3 3 3/6 1/6 3/6 3/6