International Journal of Information and Computer Security (7 papers in press)

Regular Issues

• Deep learning driven fusion of iris biometrics for optimised multimodal authentication informative security  
  by S.V. Sheela, K.R. Radhika 
  Abstract: Secure authentication methods have been made possible by the high level of maturity obtained by biometric-based technologies. Artificial neural networks forecast non-parametrically using interconnected artificial neurons, like the biological nervous system. For verification, iris, hand geometry, handwriting, fingerprint, speech, retina, face, and typing rhythm were studied. Iris recognition is most popular because it accurately identifies people. This study uses iris biometric authentication. The method simulates CASIA-Thousand-Iris utilising deep convolutional neural network (DCNN) architectures EfficientNetB0.1, CNN, DenseNet, and ConvNeXt. The experiment used our retinal recognition method to accurately identify numerous retinal samples. The suggested study introduced an MSAGFF module to EfficientNetB0.1 for iris biometrics. The attention mechanism uses channel spatial attention (CSA) to reduce redundant information and improve discriminative features for accurate recognition. Adaptive fusion strategy dynamically integrates recovered features from different receptive fields to increase model durability and decision-making. For secure Iris-based identification, EfficientNetB0.1s multimodal authentication is reliable. This end-to-end strategy boosts system performance. CNN (98.79%), DenseNet (92.11%), and ConvNeXt (66.66%) had worse accuracy than EfficientNetB0.1 (99.33%). CNN architectures in biometric systems are extended by deep learning-based iris recognition for safe authentication.
  Keywords: retina; convolutional neural network; CNN; informative security; multi-factor authentication; biometric identification; EfficientNetB0.1; DenseNet; ConvNeXt.
  DOI: 10.1504/IJICS.2026.10075572
   
  
• Optimised compact authentication scheme based on three factors for cloud-based electronic transactions  
  by Renuka Kondabala, Savadam Balaji, S. Sai Anuraghav 
  Abstract: Cloud services provide seamless data sharing, storage, and processing, enabling for the development of scalable applications and services capable of responding to real-time events. But as cloud technology becomes more common in daily life, it presents serious security risks, especially in relation to data breaches, illegal access, and complying with regulations standards. In order to increase the security of cloud-based electronic transactions, this study proposes a novel multi-factor authentication framework. In order to detect intrusions and reject malicious data before it is stored, the proposed approach incorporates an adaptive neuro-fuzzy inference system (ANFIS). Furthermore, sensitive user data, such as credentials and biometric information, is protected with homomorphic encryption (HE) for enhanced privacy. The framework for security is constructed with an optimised compact authentication (OCA) scheme that consists of three phases: setup, registration, and authentication. The system solves important vulnerabilities such session key leakage and provides procedures for user revocation and re-registration. The model performs significantly better than existing security methods when evaluated using throughput, latency and packet loss ratio. The Python platform is used to develop and test the complete system, proving its efficacy in boosting trust among users in cloud services and protecting electronic transactions conducted in the cloud.
  Keywords: cloud-based electronic transactions; ANFIS; optimised compact authentication; OCA; homomorphic encryption; HE.
  DOI: 10.1504/IJICS.2026.10076727
   
  
• Quantum-enhanced autonomous DNS over HTTPS security (CITADEL-DoH)  
  by Basharat Ali, Guihai Chen 
  Abstract: The security of DNS over HTTPS (DoH) faces escalating challenges driven by advances in quantum computing, adversarial artificial intelligence, and increasingly sophisticated attack strategies targeting encrypted communication channels. While existing protection mechanisms remain effective against known threats, they exhibit structural limitations when confronted with adaptive, multi-layered, and post-quantum adversarial models. To address these deficiencies, this study introduces CITADEL-DoH, a unified security architecture integrating quantum-resistant AI through hybrid LWE-CKKS encryption, decentralised trust enforcement via a proof-of-trust (PoT) protocol combined with federated byzantine agreement (FBA), and hardware-assisted integrity verification using FPGA-accelerated verifiable delay functions (VDFs). Novel contributions further include the application of topological data analysis (TDA) for encrypted anomaly detection and liquid time-constant (LTC) networks for adaptive traffic modelling. Experimental evaluation demonstrates 99.3% detection accuracy for domain generation algorithm (DGA) attacks and sub-millisecond query verification latency. Results confirm enhanced scalability, resilience to evasion, and robustness against emerging quantum-era threats.
  Keywords: network protocols; network security; DNS over HTTPS; enhancing network security; attack detections; AI and network security; cyber attacks in networks.
  DOI: 10.1504/IJICS.2026.10077135
   
  
• Adversarial threat evaluation of machine learning-based phishing detectors: a cybersecurity-focused black-box perspective  
  by Eiman Tamah Al-Shammari 
  Abstract: Phishing attacks often succeed through subtle URL modifications that evade detection. We evaluated five URL-based classifiers under a label-preserving black-box attacker with a fixed edit budget across three scenarios: unaltered data, edited phishing URLs, and mixed streams. Model reasoning was examined using permutation-based importance, TreeSHAP, and Kernel-LIME. On clean data, ensemble models achieved F1 scores of approximately 0.981, MLP reached 0.978, and linear models attained 0.971. Under maximum perturbation (k = 9), XGBoost and MLP maintained F1 above 0.975, while other models declined slightly. Precision remained stable; decreases reflected missed detections. A lightweight augmentation technique improved adversarial recall by 0.003 to 0.007 without reducing precision. Attribution analysis showed ensemble models shifted toward fragile features as edits increased, and growing SHAP-LIME divergence may indicate stability drift during deployment.
  Keywords: phishing detection; adversarial robustness; URL features; mixed‑set evaluation; SHAP; LIME; interpretability.
  DOI: 10.1504/IJICS.2026.10077273
   
  
• Intelligent cyber-attack detection in autonomous vehicles using residual network based deep learning model  
  by Masira M.S. Kulkarni, Prashant Dhotre, Mohd Shafi Pathan 
  Abstract: Research on autonomous vehicles (AVs) has made significant progress in enhancing cybersecurity through intrusion detection system (IDS). The major contribution of the proposed study lies in addressing the demerits of existing models that frequently suffer from limited feature significance, high dimensional data, and poor evaluation of critical attacks. The proposed technique involves a robust vulnerability assessment method that utilises optimal feature selection and efficient classification algorithms. The binary mutation-based coati optimisation algorithm (BMCOA) is employed for dimensionality reduction by selecting the optimal feature subset. Additionally, a depth wise separable residual network with a bidirectional long short-term memory (DSResNet-Bi-LSTM) is introduced for the classification and detection of cyber-attacks. The performance of the IDS is evaluated using two datasets, namely the car hacking and SWaT datasets. The results demonstrate that the DSResNet-Bi-LSTM model outperforms existing techniques with an accuracy of 99.48%, precision of 99.2%, recall of 98.59%, and an F1-score of 98.02%.
  Keywords: autonomous vehicles; intrusion detection systems; IDS; binary mutation-based coati optimisation algorithm; cyber-attacks and car hacking dataset.
  DOI: 10.1504/IJICS.2026.10077392
   
  
• Hybrid energy and opcode sequence-based detection of automation attacks in online social networks using SSFCM classification  
  by Anjali Rawat, Anand Rajavat 
  Abstract: Online social network automation attacks (OSNAA) increasingly employ automated tools to perform malicious activities such as bot-based interactions, email hijacking, and malware-driven manipulation. This study introduces an automated social network attack detection model (ASNADM) that integrates energy consumption footprint analysis (EComp-FP) with automated software opcode sequence analysis (ASOSA-OSM) to identify automation-based anomalies at the client side. The framework analyses behavioral energy traces from CPU and system activity alongside opcode n-gram representations derived from executable binaries using opcode frequency variance (OFV) and weighted term frequency (TF-W). These heterogeneous features are fused and classified using self-adaptive soft fuzzy C-means (SSFCM) clustering. Experiments conducted on the SPEMC-15K-E dataset demonstrate a detection accuracy of 99.93% with a 0.07% false-positive rate, outperforming DT, KNN, RF, and SVM models. Results confirm that abnormal energy patterns and low-entropy opcode sequences effectively reveal malicious automation in online social networks.
  Keywords: online social network automation attacks; OSNAA; energy consumption footprint analysis; EComp-FP; opcode sequence mining; self-adaptive soft fuzzy C-means; SSFCM; client-side anomaly detection.
  DOI: 10.1504/IJICS.2026.10077393
   
  
• DuoNet: a hybrid deep learning model for shilling attack detection in recommendation systems  
  by M. Sunitha, Naramula Venkatesh 
  Abstract: The study proposes a hybrid deep learning model, DuoNet, designed to detect and mitigate shilling attacks effectively. Data is collected from social media networks and e-commerce platforms, capturing user-item rating interactions. The pre-processing stage involves removing duplicate entries, imputing missing values using mean imputation and scaling the data with the min-max normalisation technique to ensure consistency. DuoNet integrates two advanced methodologies: T-Bi-LSTM for extracting temporal features and OCNN for capturing spatial features. The improved seagull optimisation algorithm (ISOA) optimises the CNNs hyperparameters, enhancing the models overall performance. The classification layer in the CNN combines temporal and spatial features to predict whether a user profile is genuine or represents a shilling attack. Experimental evaluations conducted on datasets from Amazon and Netflix demonstrate that DuoNet outperforms existing models, achieving higher accuracy, precision, F1-score, recall, and specificity.
  Keywords: shilling attacks; DuoNet; T-Bi-LSTM; OCNN; improved seagull optimisation algorithm; ISOA; temporal features and spatial features.
  DOI: 10.1504/IJICS.2026.10077481