Stochastic Variational Graph Networks for Uncertainty-Aware Biomedical Relation Extraction and Disease Network Modeling M. Narasimhulu, M Jahnavi, Sireesha Kakulla, V Rajasekhar, Kota Venkateswara Rao, P Naresh Proceedings of the 5th International Conference on Sentiment Analysis and Deep Learning Icsadl 2026, 2026 The discovery of biomedical knowledge depends on proper identification and modeling of complicated relationships between diseases, genes, proteins, and therapeutic agents. The traditional types of graph neural networks and relation extraction models may be useful in terms of reflected structural patterns, but they usually work on deterministic assumptions, and they do not measure uncertainty, which is a major drawback of high-stakes biomedical use where the degree of certainty in prediction is as valuable as the precision of prediction. We suggest a solution to this problem to include Stochastic Variational Graph Networks (SVGN) a new uncertainty-aware biomedical relation extraction and disease network modeling framework. The variational Bayesian inference and stochastic latent representation learning formed part of SVGN, which is used to predict probabilities of missing data and predicting edges to provide the inherent variability and noise that is found in biomedical data. The framework uses stochastic sampling and Monte Carlo inference to estimate relation probabilities, as well as uncertainty measures, including predictive entropy and variance, which can be used to make decisions that are risk-averse. Empirical data on benchmark datasets (Hetionet, CTD, BioRelEx) indicate that SVGN is more effective in F1-score, AUC-ROC and calibration and generates high-quality, reliable and interpretable disease networks, as well as, biologically significant networks. Moreover, the construction of uncertainty-aware graphs leads to better understanding of the disease mechanisms, prediction of drug repurposing, and detecting biomarkers. This piece of work is one step in the right direction towards reliable, interpretable, and confidence-scale AI in the field of biomedical knowledge discovery and translational medical software.
Sybil Attack Detection in VANET Using Machine Learning Approach Sireesha Kakulla, Srinivas Malladi Ingenierie Des Systemes D Information, 2022 VANET (Vehicular Ad-hoc Network) is a subclass of MANET in which many cars can connect with one another via node to node or equipment erected on the side of the road. However, due to the adaptability of centres and the unexpected trade in geography, there may be opportunities for attacks in VANET. One of the ostensible assaults is the Sybil attack, in which the attacker fabricates unequivocally unique equal personalities to undermine the value of VANET. Sybil creates fictitious identities inside the community as well in order to sabotage attempts to mediate conversations between community nodes. Sybil assaults have an impact on carrier transportation in relation to things like traffic congestion, road safety, and multimedia entertainment. VANETs therefore announce a security mechanism to protect you from Sybil attacks. In this regard, this work puts forth the SDTC method, which completely relies on machine learning techniques to prevent Sybil assaults in VANETs. In order to reduce identification time, increase detection accuracy, and enhance scalability, the SDTC (Sybil node detecting the use of Classification) mechanism uses a few vehicle-specific Extreme Learning Machine (ELM) features. The results suggest that SDTC is a suitable strategy to reduce Sybil assaults and sustain provider service in VANETs.
AN IMPLEMENTATION OF HYBRID APPROACH FOR SYBIL ATTACKS IN VEHICULAR AD-HOC NETWORKS (VANETS) Sireesha Kakulla, Srinivas Malladi Indian Journal of Computer Science and Engineering, 2022 The Vehicular Ad-hoc Network (VANET) is becoming increasingly important because of its high mobility and common link breakage architecture. There are several amusement services provided to all passengers via the VANET, and it is these services that ensure that the riding environment runs as smoothly as possible. Vehicle networks include a variety of routing protocols to help them communicate effectively, but these networks are vulnerable to a wide range of threats, including the introduction of rogue nodes. Many VANET systems face a serious security challenge today, as a misdirected conversation could result in catastrophic consequences for human lives, either immediately or in the future. In this Paper, the ns2 simulator can be used to construct the hybrid detection technique. P 2 DAP performs better than footprint as the number of large-capacity vehicles on the road rises. In contrast, when the number of cars increases, the footprint set of guidelines performs better. Encrypted facts, the authentication method and the car's trajectory can all be taken into consideration when creating a new Hybrid technique with set of rules.
Phishing and Sybil Enhanced Behavior Processing and Footprint Algorithms in Vehicular Ad Hoc Network Sireesha Kakulla, Srinivas Malladi International Journal of Safety and Security Engineering, 2022 Communication between vehicles is the core of VANETs, which are built on notions of mobile ad hoc networks (MANETs). Each car on the network is now seen as a mobile network node. All participating cars become wireless routers or nodes, depending on your viewpoint. VANET creates a huge network with a long-range by connecting all cars within range to a fixed unit. VANET assists with traffic regulation, communication between vehicles and the sharing of road data. There is a possibility that the VANET network could be compromised by identity and information concerns, resulting in data delays or theft. Attacks like Sybil and phishing are possible because of this network's weakness. Due to the two recent strikes, infrastructure and human lives may be in jeopardy. There are two novel algorithms developed by the authors to tackle Sybil and Phishing assaults on VANET networks: Phishing and Sybil Enhanced Behavior Processing and Footprints (P&SEBP&F). Originally known as the Phishing and Sybil Enhanced Behavior Processing and Footprint, P&STL&T was renamed P&STL&T. Phishing and Sybil were used as well as varied attacker-to-victim node ratios to test the effectiveness of the new tactics for assessing effectiveness. Compared to previous study and the work of the other authors cited, there were approximately 30% fewer attackers detected during the research.
A Survey of VANET Security Models and its Issues on Node Level Data Transmission K. Sireesha, Srinivas Malladi Proceedings of the 2nd International Conference on Artificial Intelligence and Smart Energy Icais 2022, 2022 Vehicular Ad hoc Network (VANET) plays a vital role in communication between moving vehicles on a network enabled and regulated by wireless network protocols. However, as the size of the network increases, the vulnerability of the node attacks and data attacks also increases exponentially. Most of the conventional models are depend on the network size, its topology structure and data size. In this paper, a study of different VANET security models and its issues are studied based on the network parameters and security metrics. Different types of VANET security models, network attacks and encryption models are discussed in this paper. Finally, the limitations of the conventional VANET security models are discussed on large VANETs.
Attenuation of co-channel interference in femtocell networks International Journal of Applied Engineering Research, 2015
