Page 27 - AIH-1-4

Artificial Intelligence in Health Optimized clustering in medical app detection

Table 1. (Continued)
Names Details
referralMD Aims to standardize referral network communication between primary care physicians and specialists
RevUp A chronic pain management tool that enables patients to access, log, and monitor their health information for improved care
outcomes
Sherpaa Enables staff physicians to provide real-time medical advice to their clients’ employees
SmartConcierge Helps users understand health plans and benefit packages, schedule appointments, and provides 24/7 support from registered
nurses
Teladoc Resolves medical issues between patients and physicians via phone or video consults
Touch Surgery Uses graphic 3-D images to give surgeon trainees an idea of what they will see during the actual procedure
Twine Helps clinicians provide continuous care to chronically ill patients through their devices, anytime, anywhere
UpToDate Helps clinicians make the best evidence-based, point-of-care decisions from the most current medical information
YouPlus Provides daily lifestyle coaching to make healthy living easier with science-driven exercise programs, meal recommendations, etc.
ZocDoc Helps patients schedule an appointment that will get them in front of a physician within 24 h

criminals. Axelle Apvrille, a principal security researcher at help detect these apps. Increasing the quality and safety
Fortinet, highlighted the details of such malicious Android of medical apps through better standards and validation
apps during a presentation at the Virus Bulletin 2019 practices needs to be established to ensure the proper use
conference in London. Many free diabetes management and integration of these increasingly sophisticated tools
apps, while seemingly helpful, require users to download into medical practice. Understanding and quantifying the
additional apps that are loaded with adware to function risks associated with medical apps depends on two critical
as advertised. Meanwhile, another type of malicious app, factors: (i) The frequency of events that cause damage,
posing as a diabetes advisor, tracks almost all the user and (ii) the severity of the resulting damages. These risks
activities, including the GPS location of the device, the IP include potential harm such as damage to a doctor’s/
address, and the other installed apps on the device, putting hospital’s reputation, privacy issues, or clinical decision
the privacy of the user at total risk. All these apps also errors, in increasing order of severity. The decision to use
bombard users, including vulnerable patients who rely on any health-care app depends on whether the risk associated
them, with persistent pop-up advertisements. with the app is less in comparison to its benefits.
In addition, a concerning trend involves malicious 3.3. Digital health apps during COVID-19
apps claiming to predict users’ life expectancy within Along with the adverse effects of the pandemic, a concurrent
minutes based on health-related questions while secretly “infodemic” has emerged, characterized by widespread
transmitting these details to remote servers, raising misinformation circulating online about the coronavirus.
significant privacy concerns. Such stolen medical records Therefore, it is necessary to choose reliable apps. The Apple
often end up for sale on dark web forums, which in turn Corporation has adopted a cautious approach by cracking
results in financial gain for cybercriminals. The hackers down on potentially malicious software in its app store,
create malicious health-related apps as they serve as an allowing apps only from recognized institutions such as
easy way to steal data, install malware, or both, affecting governments, health organizations, or hospitals, excluding
numerous individuals. This trend is particularly concerning independent developers. Similarly, Google has also
as cybercriminals increasingly target individuals using implemented proactive measures; Google Play launched
health-care apps, a demographic that continues to grow. a section dedicated to COVID-19 with a curated list of
Many app developers lack formal medical training and certified apps. 13
do not involve clinicians in the development process.
Consequently, they may be unaware of patient safety issues 4. Methods
arising from inappropriate content or app functionality. 21-23
The proposed methodology relies on the integration of
Moreover, the exponential growth of medical apps machine learning techniques, specifically the ANN and
has made it impossible to thoroughly assess each one. the K-means clustering algorithm, to form an advanced
24
Despite this, evidence suggests that even a small number of clustering method for the detection of medical apps. The
medical apps can pose a risk to patient safety, underscoring ANN, inspired by the structure and functioning of the
the necessity of developing robust detection models to human brain, serves as a universal classifier in the proposed

Volume 1 Issue 4 (2024) 21 doi: 10.36922/aih.2585