Unsupervised Detection of Anomalous Behavior in Wireless Devices based on Auto-Encoders

Albasir, Abdurhman; Hu, Q.; Al‐tekreeti, Mustafa; Naik, Kshirasagar; Naik, Nenavath Srinivas; Kozłowski, Adam; Goel, Nishith

doi:10.1109/noms47738.2020.9110433

Cited by 3 publications

(8 citation statements)

References 18 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…In addition, the gadgets are inherently insecure since they need a constant Inter-net connection. New instances and studies show that greater investigation is required [22]. They also reveal, sadly, that present solutions are far from being acceptable to halt the exponential development in the number and complexity of assaults [24].…”

Section: Background and Literature Reviewmentioning

confidence: 99%

“…Numerous attacks have been made against the Internet of Things (IoT), including notable Distributed Denial of Service (DDoS) attacks on Dyn's DNS (domain name system; see reference). [20]Attacks/vulnerabilities on/off self-driving cars [21], ransomware attacks [21], attacks on smart home and medical equipment [22,23], and more. The IoT and critical infrastructures pose new security threats.…”

Section: Background and Literature Reviewmentioning

confidence: 99%

“…They also reveal, sadly, that present solutions are far from being acceptable to halt the exponential development in the number and complexity of assaults [24]. Thus, the necessity for further layers of defense to make smart home IoT devices more durable and proof of such assaults is vital [22]. [25] introduced a host-based anomaly discovery method in which device features and activities are considered using a mobile phone application.…”

Section: Background and Literature Reviewmentioning

confidence: 99%

See 2 more Smart Citations

IoT Network Anomaly Detection in Smart Homes Using Machine Learning

Sarwar,

Bajwa,

Hussain

et al. 2023

IEEE Access

View full text Add to dashboard Cite

In this modern age of technology, the Internet of things has covered all aspects of life including smart situations, smart homes, and smart spaces. Smart homes have a large number of IoT objects that are working continuously without any interruption. Better security and authentication of these smart devices can provide peaceful environments to live in such spaces. It is important to monitor the activities of smart IoT devices to make them work fault free. Such devices are small, consume relatively less power and resources, and are easily attackable by attackers. It is crucial to protect the integrity and characteristics of the smart home environment from external attacks. Machine Learning played a vital role in recognizing such malicious activities and attempts. Several Machine Learning approaches are available to detect the normal and abnormal behavior of IoT device traffic. This study proposed a machine learning-based anomaly detection approach for smart homes using different classifiers. Testing and evaluation are performed using the University of New South Wales (UNSW) BoT IoT dataset. Machine learning models based on four classifiers are built using an IoT devices dataset. For the Test dataset, the Weighted Precision, Recall and F1 score of Random forest, decision tree, and AdaBoost is 1 as compared to ANN which has 0.98, 0.96, and 0.96 respectively Results show that high performance, precision, and robustness can be achieved using the proposed methodology. In this way, smart homes' security and identity of devices can be monitored and anomalies can be detected with high accuracy. Attack categories include binary class, multiclass class, and subclasses. Results show Random Forest algorithm outperforms enough to use this methodology in smart environments.

show abstract

Section: Background and Literature Reviewmentioning

confidence: 99%

Section: Background and Literature Reviewmentioning

confidence: 99%

Section: Background and Literature Reviewmentioning

confidence: 99%

See 1 more Smart Citation

IoT Network Anomaly Detection in Smart Homes Using Machine Learning

Sarwar,

Bajwa,

Hussain

et al. 2023

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Heuristic-based methods provided a layer of detection by identifying suspicious behaviors not explicitly defined in existing rules [12]. Behavioral profiling techniques, which involved creating detailed profiles of legitimate application behavior, enabled the detection of deviations indicative of malware [13]. The integration of contextaware features, such as user behavior and environmental factors, further enhanced the precision of detection models [14].…”

Section: Related Studiesmentioning

confidence: 99%

Situational Aware Access Control to Prevent Android Malware

Skalski,

Dombrokova,

Szczawinski

2024

Preprint

View full text Add to dashboard Cite

Android malware poses a significant threat to the security and privacy of users worldwide, requiring advanced protective measures. A novel and significant contribution to Android security, situational aware access control, dynamically regulates access to sensitive functionalities based on real-time behavioral analysis. By integrating machine learning models with detailed user interaction data, the system effectively distinguishes between normal and anomalous behaviors. The implementation of a feedback mechanism allows continuous learning and adaptation, ensuring that the system evolves in response to emerging threats and changing user behaviors. Performance evaluations demonstrated high accuracy, precision, and recall, with minimal impact on user experience. This dynamic approach not only enhances malware detection but also improves user satisfaction by reducing false positives and ensuring seamless interactions. The implications for Android security are profound, offering a resilient and adaptive solution to the growing challenge of malware prevention.

show abstract

“…Some studies focused on developing machine learning models that predict ransomware activity based on app behavior patterns [13,27]. The integration of cloud-based services for off-device analysis was also proposed to limit the impact on device performance [36,37]. Data backup strategies have been highlighted as a crucial mitigation technique, ensuring data recovery without paying ransom [38].…”

Section: Mitigation Techniquesmentioning

confidence: 99%

An Effective Strategy for Ransomware Mitigation on Android Devices via Android OS File System API

Ozturk,

Yilmaz,

Arslan

et al. 2024

Preprint

View full text Add to dashboard Cite

The development and effectiveness of an API-based system for mitigating ransomware threats on Android devices was explored in this study. Leveraging the Android/Linux file system API, the proposed system enhances the detection and prevention of ransomware by monitoring suspicious file system activities. The methodology includes extensive testing across various Android models and versions, demonstrating a high detection rate with minimal false positives. Results indicate that integrating system-level APIs into security frameworks significantly improves response times and device integrity against ransomware attacks. This paper contributes to mobile cybersecurity by providing a scalable and effective strategy for ransomware mitigation, proposing future enhancements with advanced predictive algorithms and broader system integration.

show abstract

Unsupervised Detection of Anomalous Behavior in Wireless Devices based on Auto-Encoders

Cited by 3 publications

References 18 publications

IoT Network Anomaly Detection in Smart Homes Using Machine Learning

IoT Network Anomaly Detection in Smart Homes Using Machine Learning

Situational Aware Access Control to Prevent Android Malware

An Effective Strategy for Ransomware Mitigation on Android Devices via Android OS File System API

Contact Info

Product

Resources

About