The rise of Internet of Things devices in healthcare and sensor networks has greatly improved patient care, but it also brings major security and privacy risks. The existing works failed to detect multi-segment attacks or isolate threats in micro segmented areas. Therefore, this paper presents healthcare security and optimization with edge computing and micro segmentation in sensor networks, and optimized threat detection. Initially, Healthcare devices are registered and device ID is authenticated using UUID. Then the contextual information of the devices is extracted and micro segmentation is done by using CDBRSCAN. Then, for individual micro segments, threat detection and isolation is done by training the model with the dataset followed by feature extraction and from the extracted features the anomaly is detected using T3FENN. From the detected anomaly, behavior profiling takes place by using CDBRSCAN. After behavior profiling, threats are correlated and isolated by using CTCS. Based on the threat detected and its correlation, the access for the device is limited using ITR. At the same time, if there is no anomaly, privacy preservation of the information is done using KSEWPA, which ensures the privacy of sensitive data by using entropy-based weighting and probability measures. Finally, the privacy preserved information is optimized using PSEHO and stored in the cloud server. As per the exper-imental analysis, the proposed model attained 98.65% accuracy.
This work proposes a secure and reliable method to API authentication and intrusion detection in SDN-IoT networks. The framework leverages HMAC-based API authentication in conjunction with an LSTM-Lasso-based intrusion detection system optimized via Particle Swarm Optimization (PSO). HMAC provides data integrity by safeguarding the API against unauthorized access, while the LSTM-Lasso system enhances anomaly detection with a dramatic reduction in false alarms. The system design ensures high detection precision of 99.35
The process of securing IoT networks is large, particularly because the older Intrusion Detection Systems (IDS) are often unable to perform with accuracy, speed, and user privacy protection. This paper presents a hybrid Intrusion Detection System (IDS) which is an integration of powerful tools (Convolutional Neural Networks (CNN), Recurrent Neural Networks (RNN), Autoencoders), and Blockchain technology to develop a more efficient solution. We show that our system is better than existing techniques with F1 score of 99.71, 99.99
The integration of IoT devices has converted patient monitoring, diagnostics, and treatment. However, current methods often neglect integrity and security of these devices within healthcare settings. This paper introduces an Artificial Intelligence and Blockchain-based system to address these issues and enhance security in Healthcare IoT. Initially, HIoT devices are set up, and data is collected. The collected data is then pre-processed. Next, the data is secured using Dynamic Hierarchical ChaCha20 Key Derivative Function (DHCC20-KDF). The secured data is temporarily stored in a buffer to facilitate secure processing. Following this, anomaly detection is conducted by collecting the anomaly detection dataset. Then pre-processing, feature extraction, and dimensionality reduction are carried out. After that from the reduced dimensionality the anomaly is classified using Gradient Penalty with Curvature Regularized Deep Belief HSig Networks (GPCRDB-Hsig). If anamolies are detected, data transmission is blocked and blockchain records are updated. If no anomalies are found, hash value of data is generated and stored in blockchain. If any updates are available in the Healthcare IoT (HIoT) devices, the security of device during update is maintained using blockchain. For such updates, the process is followed by smart contract verification at each layer to ensure transparency and integrity. If the verification is successful then the update begins. If it is unsuccessful update is blocked and preventing the unauthorized changes. As per the experimental analysis, the proposed model attained 99.58
This paper presents a novel DL design for online shopping and financial fraud detection highlighting meticulous transaction analysis and risk assessment. The approach applies a comprehensive preprocessing pipeline consisting of min-max normalization, correlation-based filter and median imputation to obtain high quality data. Latent behavioural features are derived by an autoencoder, and anomaly sensitivity is enhanced by the Local Outlier Factor (LOF). For classification, a Bi LSTM and Transformer network hybrid model learns the sequential and global interdependence among transactions. This hybrid model is beneficial in identifying sophisticated schemes of fraud and assists in intelligent credit scoring. The final classification step correctly distinguishes legitimate and fraudulent transactions. Designed to be adaptive to new patterns of fraud the classical can be directly functional to real-world financial systems, internet shopping systems and internet payment gateways. Experimental results outperform traditional fraud detection techniques.
