Sentinel

Cosson, A.; Sikder, Amit Kumar; Babun, Leonardo; Celik, Z. Berkay; McDaniel, Patrick; Uluagac, A. Selcuk

doi:10.1145/3450268.3453533

Cited by 10 publications

(2 citation statements)

References 36 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Smart home security and privacy. In the field of smart home privacy and security, substantial prior research emphasizes limiting malicious activities, which focus on device access control and authentication for single-user scenarios [49]- [56]. He et al [45] give a detailed smart home user study that depicts users' concerns about fine-grained access control in multi-user smart environments.…”

Section: Related Workmentioning

confidence: 99%

A Closer Look at Access Control in Multi-User Voice Systems

Shafei,

Tan

2024

IEEE Access

View full text Add to dashboard Cite

Voice-controlled systems have revolutionized user interactions, making technology more accessible and intuitive across various settings. In multi-user environments, such as households, voice assistants like Amazon Alexa are favored as they enable seamless interaction with devices and services. However, the convenience these systems offer comes with challenges, especially concerning privacy and security. In environments where multiple users interact with the same voice assistant, the need for sophisticated access control mechanisms becomes apparent to prevent unauthorized access to sensitive information. This study assesses the effectiveness of voice access control mechanisms within these multi-user contexts, shedding light on the inherent privacy risks associated with shared voice-controlled systems. First, we show that the current access control is vulnerable in handling users' private data. Second, we develop a framework for automated testing to explore the access control weaknesses to determine whether the accessible data is of consequence, as not all information may be equally sensitive or vital to users. Third, we identify two flaws within the access control mechanisms offered by the voice system, and we show that existing access controls are susceptible to unauthorized access. Finally, our study shows that operations on the system are protected, whereas other operations that are not protected still reveal user's private information. These findings underscore the need for enhanced privacy safeguards and improved access control systems in multi-user environments. We offer recommendations to mitigate risks associated with unauthorized access, shedding light on securing the user's private data on the voice assistant.

show abstract

Section: Related Workmentioning

confidence: 99%

A Closer Look at Access Control in Multi-User Voice Systems

Shafei,

Tan

2024

IEEE Access

View full text Add to dashboard Cite

show abstract

“…Using these techniques, Sentinel was able to achieve 96 % accuracy in detecting threats in an IoT network, as well as keeping power usage to a minimum while not affecting device performance, proving its e cacy. [7] Advantages:…”

Section: Disadvantagesmentioning

confidence: 99%

Detection of Unauthorized IoT devices using Machine Learning

Sonawane

2023

Preprint

View full text Add to dashboard Cite

As technology advances, so will the risk, as new types of vulnerabilities emerge with each new release of software. Such is the case with IoT devices and new upcoming technologies such as 5G, which will soon take over the world as each user may have two or three devices monitoring their health and safety. Allowing vulnerabilities in such devices can be dangerous and lead to fatal accidents. It is not possible to find a solution for every vulnerability, nor is it possible to continuously monitor the condition of a device. So why not let machines monitor and analyze such flaws? We can determine whether a device or its network is compromised by employing machine learning techniques. These techniques can perform device verification at high speeds while also taking action against unauthorized devices. We used parameters like a node's battery power consumption, what data it collects, how much power the CPU consumes, how many packets are exchanged and received, and so on to determine whether the node is authorized or not. This model is limited to a single type of IoT device; to add multiple devices, the model must be trained separately for each one. The method is based on the idea that if a device has unusual power consumption and packet exchange parameters, it will identify those devices as unauthorized. We were able to achieve an accuracy of over 97 percent for identifying authorized while training this model, which means that in this model we basically check whether the device is authorized by calculating a threshold value for that device. If the threshold is less than the maximum allowable threshold, the device is authorized; otherwise, it is not.

show abstract

Design of a quality of service‐aware fault‐proof secure Q‐learning‐based Internet of Things kernel patch with multipath and multichannel capabilities

Shankhpal

Hanumantha

2022

Int J Communication

View full text Add to dashboard Cite

Summary In recent times, the expeditious growth of Internet of Things (IoT) offers applications to ease day‐to‐day activities with minimum human effort. Once the IoT application installed, the connected devices perform their tasks without human intervention. Hence, the need of performance optimization and security enhancement is vital to minimize end‐to‐end communication delay, improve kernel‐level security, mitigate faults adaptively, and have suitable backup options in case of node failure. This paper proposes a quality of service (QoS)‐aware fault‐proof secure Q‐learning‐based IoT (QIoT) kernel‐level protocol that integrates multipath aggregation and fuzzy authentication for security, with multichannel communication for improved QoS. Especially, this protocol integrates a source‐level clustering mechanism based on Q‐learning that aims at reducing route search delay. In order to provide fault tolerance, the kernel is equipped with real‐time fault‐tolerance mechanism that is activated in case of node‐level faults. Due to integration of Q‐learning, computational overheads are reduced by over 15% when compared with Zephyr, AliOS, and RTX kernels. This reduction in computational overheads facilitates light‐weight behavior of the kernel, due to which other QoS parameters like energy consumption, throughput, and routing overhead are reduced. The proposed QIoT kernel‐level protocol is compared with standard kernel modules, and performance evaluation showcases an improvement in authentication security by 8%, end‐to‐end delay by 5%, energy efficiency by 25%, and fault mitigation by 18%, thereby assisting the use of the proposed kernel for real‐time deployments.

show abstract

Sentinel

Cited by 10 publications

References 36 publications

A Closer Look at Access Control in Multi-User Voice Systems

A Closer Look at Access Control in Multi-User Voice Systems

Detection of Unauthorized IoT devices using Machine Learning

Design of a quality of service‐aware fault‐proof secure Q‐learning‐based Internet of Things kernel patch with multipath and multichannel capabilities

Contact Info

Product

Resources

About