On August 28, 2026, Google, Microsoft, Anthropic, OpenAI, and 100 other companies signed an open letter warning of a large-scale AI attack. AI has created systemic risks in the digital world, but the physical world has no defense mechanism. This paper defines the Physical Ledger—a physical world namespace rooted in the Cui coordinate. The Physical Ledger DNS is not a copy of the domain name system; it is an object-addressing protocol for the physical world: every object (shelf position, robot, door, vehicle, starship) is assigned a unique Cui coordinate address. This paper presents a draft protocol for the Physical Ledger DNS, a catalog of 108 problems, the genesis valuation of $100,000,000, and a reward distribution scheme. It proposes the §13 security mechanism (Proof-of-Problem): a distributed firewall for the Physical Ledger DNS, powered by the 108 problems. The more solvers participate, the thicker the firewall. AI can attack digital protocols, but it cannot solve problems—because solving requires understanding the coordinate origin itself. The genesis valuation of the Cui-attribute Shell is defined as US$100,000,000, anchored at 2026-08-27. The appendix includes the Cui-coordinate naming rights and the passphrase lock (recognition of 1/7/8 for entry).
Nelli Yaswanth Kumar, Dr. Singothu Jhansi Rani, Setti Sarika
The rapid proliferation of Internet of Things (IoT) devices under sixth-generation (6G) networks introduces a highly dynamic, decentralized environment in which static, perimeter-based security models are no longer adequate. This paper proposes AZTM-v3 an adaptive Zero Trust framework that couples behavior-driven trust management with a Random Forest classifier to identify and isolate malicious nodes in real time. The framework is evaluated on an NS-3 simulation of a 150-node 6G IoT network subjected to Sybil, Denial-of-Service (DoS), spoofing, replay and ON-OFF attacks. Unlike prior trust-management proposals that report only qualitative or partial outcomes this work quantifies performance across five dimensions i.e detection accuracy, F1-score, false-positive rate, end-to-end latency and consensus-convergence time and benchmarks AZTM-v3 against PKI-based, centralized-trust and static-blockchain baselines. AZTM-v3 attains a 98.1% overall detection accuracy with a 1.6% false-positive rate at 150 nodes and sustains 95.4% accuracy at 200 nodes outperforming the PKI baseline by 12–18 percentage points across all tested loads. These results indicate that combining tiered trust evaluation with machine learning based classification yields a measurably more scalable and resilient security layer for 6G-enabled IoT deployments than existing static or purely cryptographic approaches.
Wisam Makki Alwash, Weam Husham Aljabbari, Muhammed Ali Aydın, Hasan H. Balık
Internet of Vehicles (IoV) intrusion detection systems (IDSs) require collaborative learning that preserves raw-data locality while producing independently checkable post-run evidence. This paper presents FL-BC-IDS, an evidence-native, privacy-aware hierarchical federated IDS in which vehicles train Differentially Private XGBoost models, roadside units perform deterministic admission and tree-bagging aggregation, and the GLOBAL stage forms an equal-weight ensemble over validated RSU models. Signed reports, privacy records, SHA-256/Poseidon commitments, scoped Groth16 proofs, reconstructable public inputs, and digest-pinned blockchain receipts provide a unified verification path. Across 10 seed-controlled runs, the mean ± SD accuracy/F1 values were 0.998021±0.000246/0.983597±0.002053 on CSE-CIC-IDS2018 and 0.999867±0.000152/0.999495±0.000579 on CICIoV2024. With thresholds fixed exclusively from development data, the strict held-out-attack macro recall was 0.8031 and 0.9090 on CSE-CIC-IDS2018 and CICIoV2024, respectively, indicating residual attack-specific generalization limitations; supervised rolling-origin temporal refresh on CSE-CIC-IDS2018 achieved 0.984788 pooled seen-attack recall at a 0.005700 test FPR. A controlled 20-vehicle, eight-round heterogeneity and participation stress test retained 0.998151 accuracy and 0.984782 F1-score. Verification rejected invalid or context-mismatched artifacts and independently checked model–anchor consistency, RSU aggregation replay, commitments, and public inputs. The reported DP budgets are conditional learner-stage bounds for learner-input record instances, not end-to-end guarantees for original pre-preprocessing records.
With the rapid evolution of the Decentralized Finance (DeFi) ecosystem, stablecoins have emerged as a critical infrastructure bridging the cryptocurrency market with traditional financial paradigms. However, stablecoin systems rely heavily on smart contracts to execute automated operations. The immutable nature of these systems post-deployment means that the exploitation of security vulnerabilities can lead to irreversible, massive economic losses and potentially trigger systemic financial risks. Current research on stablecoin smart contract security faces challenges such as a lack of domain-specific targeting and the obsolescence of static defense models. To address this, this paper systematically analyzes common attack vectors in stablecoin environments and proposes a practical, real-time dynamic defense architecture. By analyzing 12 real-world security incidents, we elucidate the underlying mechanisms of high-risk patterns such as reentrancy attacks, oracle manipulation, and composite flash loan attacks. Concurrently, we construct a real-time anomaly detection model utilizing multi-dimensional on-chain temporal features and the Bi-LSTM algorithm. Experimental results demonstrate that this model achieves a classification accuracy of 96.61\%, with an average recall rate of 97.70\% for malicious attack samples, and a single inference latency ranging from 1.5 to 2.8 milliseconds.
In the age of fast industrial digitalization, securing the heterogeneous and high-volume data produced by the Industrial IoT systems is a basic need. The chapter is dedicated to the application of machine learning and deep learning methods in the process of securing multimodal data within the context of Industrial Internet of Things (IIoT). It includes a detailed discussion of multimodal sources of data and the corresponding cyber threat environment, and then it introduces the machine learning (ML)-based and deep learning (DL)-based anomaly detection and intrusion prevention techniques. The chapter reviews the secure architectural designs, which combine edge, fog, and cloud intelligence and privacy-sensitive and trust management schemes like federated learning and blockchain. The practical applicability of such approaches is pointed out by the real-life industrial applications and case studies. The main implementation issues and the performance evaluation metrics are examined to ensure a successful implementation. The chapter ends by highlighting the future directions and new trends, focusing on adaptive, explainable, and resilient intelligent security solutions in next-generation IoT systems of the industrial world.
Recent intrusion detection systems (IDS) increasingly rely on machine learning (ML) and deep learning techniques to detect sophisticated cyberattacks. However, many existing frameworks still suffer from limited explainability, black-box decision-making, and the absence of secure trust verification mechanisms for intrusion records. To address these challenges, this paper proposes TAE-IDS, a Trust-Aware Explainable Intrusion Detection Framework that integrates attention-based meta-ensemble learning, SHapley Additive exPlanations (SHAP)-driven explainability, and blockchain-inspired tamper-evident validation within a unified cybersecurity architecture. The proposed framework employs heterogeneous base classifiers, namely Logistic Regression (LR), Extra Trees (ET), and XGBoost (XGB), to capture diverse network traffic characteristics. Uncertainty-aware meta-features, including logits, confidence scores, and entropy representations, are extracted from the base learners and processed by an adaptive Bidirectional Long Short-Term Memory (BiLSTM) attention-based meta-classifier for contextual intrusion reasoning and adaptive ensemble aggregation. To enhance transparency and analyst trust, SHAP-based explainability is incorporated to provide both global and local interpretations of intrusion predictions. Furthermore, a blockchain-inspired tamper-evident validation mechanism based on SHA-256 cryptographic hashing is integrated to enable tamper-proof intrusion logging, immutable auditing, and secure forensic verification of IDS outputs. The proposed framework was evaluated on the UNSW-NB15 and CICIDS2017 benchmark datasets under both binary and multiclass intrusion detection settings. Experimental results demonstrate that TAE-IDS achieves strong intrusion detection performance, interpretable intrusion reasoning, and effective blockchain-assisted tamper-evident validation on the evaluated benchmark datasets. The integration of explainable artificial intelligence (XAI) and blockchain-assisted validation enhances transparency, forensic traceability, and the integrity of intrusion records while providing a foundation for future validation in operational network environments.
Web security has become a critical domain as modern applications increasingly rely on dynamic user-generated content, making them highly vulnerable to Cross-Site Scripting (XSS) attacks. Traditional detection systems struggle to cope with evolving payload patterns, limited generalisation across institutions, and strict privacy restrictions that prevent sharing of sensitive request logs. To address these challenges, this work proposes a Zero-Knowledge Federated Sequence Learning (ZK-FSL) framework that enables collaborative XSS detection without exposing raw data or intermediate gradients. The model integrates attention-based deep sequence learning with zero-knowledge proof validation, ensuring both strong predictive capability and verifiable trust among participating clients. Experimental evaluation demonstrates that ZK-FSL achieves superior performance compared to centralised and federated baselines, reaching 96.3% accuracy , 96.7% precision , 95.9% recall , 96.3% F1-score , and an AUC of 0.98 . These results confirm that the proposed framework effectively enhances privacy-preserving threat detection while maintaining high robustness against diverse and sophisticated XSS attack patterns.
Network slicing and resource provisioning in 6G focus on creating multiple customized virtual networks over a shared infrastructure. However, these approaches also introduce challenges, like increased architectural complexity, higher implementation costs, security vulnerabilities between slices in resource optimization across highly dynamic and heterogeneous network environments. In this work, Exponentially Tactical Unit Algorithm (ETUA) is devised for network slicing in 6G. Initially, blockchain-enabled 6G network is simulated, and the set of features, like user device type, delay rate and packet loss rate are collected from various devices. Moreover, network slicing is done by ETUA that integrates Exponentially Weighted Moving Average (EWMA) and Tactical Unit Algorithm (TUA). Finally, resource allocation is performed using Attention High-order Deep Network (AHoNet) by considering the parameters that includes bit error probability, sum rate and trust. The efficacy of ETUA is examined by bit error probability, utility and latency with 0.012, 0.950 and 0.509 Sec.
Over the past few years, Wireless Sensor Networks (WSNs) have been increasingly deployed for numerous sensing and monitoring purposes in environmental monitoring, industrial automation, health monitoring, military surveillance, smart agriculture and disaster management among others. The inherent limitations in terms of processing power, memory, communication bandwidth and energy of sensor nodes make WSNs highly susceptible to malware attacks. A wide variety of malware such as sensor network worms, Trojans, viruses, botnets and ransomware can easily propagate in a network through inter node communication. Such malware can cause serious damage to communication, compromise sensitive data, consume energy of the infected nodes thereby reducing the lifetime of network among others. In the last decade, numerous approaches have been proposed for the detection of malware infecting sensor nodes. These approaches range from traditional signature-based detection and behavior-based detection to more advanced approaches such as machine learning (ML)-based, deep learning (DL) -based, blockchain-based, trust management-based and federated learning-based detection. Most of the existing approaches for malware detection in WSNs have been designed to work on WSNs and have not been tested on real scenarios. Most of the approaches have their own strengths and weaknesses and the most suitable approach for a given application depends on various factors. In this paper, we present a comprehensive review of approaches for the detection of malware infecting sensor nodes in WSNs. We present a taxonomy of reviewed approaches for detection of malware. We also present a discussion on approaches for modeling malware propagation in a WSN as well as review on various categories of malware that have been designed to attack sensor nodes in WSNs along with detection frameworks for different categories of malware. We also present a comparative study of approaches used for the detection of malware in WSNs on the basis of various parameters such as detection accuracy, computational complexity, energy efficiency, scalability, detection latency and deployability. The review and taxonomy presented in this paper will be highly beneficial for researchers and practitioners designing approaches and systems for the detection of malware in WSNs. Various open research challenges in this area have also been discussed in this paper including detection of zero-day malware, designing of intelligent models to be light enough to be deployed on sensor nodes, use of explainable artificial intelligence for detection of malware in WSNs, designing approaches for privacy-preserving collaborative learning in WSNs and designing adaptive security approaches for WSNs.
Multi-cloud adoption has widened the enterprise attack surface to a degree that perimeter-based defence can no longer address. Traffic is now flowing continuously across AWS, Azure, and GCP, and the majority of deployed Zero Trust Architecture (ZTA) systems are still using static rule tables, with no ability to provide an audit trail of the reasoning behind decisions, and with logs stored in datastores that can be modified by an insider without detection. This paper proposes ZT-ChainGuard, a framework that overcomes these three limitations in one architecture that integrates an ensemble machine learning trust-scoring engine, ZTA policy enforcement and a blockchain-based audit trail. The trust-scoring engine is a two-layer stacking ensemble, with XGBoost and Random Forest as base learners, and Logistic Regression as a meta-learner, and it returns a continuous trust score, P(Attack | flow), for each network flow, which is then used to trigger the ZT policy decision at a threshold of 0.5. The explanation of each decision is provided by SHAP values at both the global and per-flow level, and each decision is stored as an immutable, SHA-256 hash-chained block. On CICIDS2017 (2.83 million flows, 14 attack classes) the framework achieves 99.90% accuracy, 99.71% F1-score, and 99.99% ROC-AUC; on ToN-IoT (2.23 million IoT records, 9 attack types) it achieves 99.81% accuracy, 99.88% F1-score, and 100% ROC-AUC. The latency of inferences is 0.006ms per sample, and the overhead of auditing the blockchain is 0.019ms per block. This performance is not just a quirk of a particular split, as it is shown to be stable across the three folds of three-fold cross validation.
Cloud computing infrastructures are facing serious risks from Distributed-Denial-of Service attacks, which include historically high attack volumes and ineffectiveness of conventional defensive strategies. The usefulness of blockchain based security measures in detecting and preventing DDoS attacks on cloud computing infrastructure is covered in this paper. Analyzed a hybrid approach that integrated distributed ledger technology with smart contracts for the identification of attack patterns across 50 enterprise cloud environments over a period of 18 months. Our results show a reduction of false positives by as much as 87% for blockchain-based validation compared to the conventional approach and a 94% success rate in the detection of advanced DDoS variants. The response time remained, on average, around 2.3 seconds during the high-volume attack in comparison to traditional centralized solutions. The results in the paper gives an idea that an immutable and distributed consensus characteristic based on blockchain provides robust defenses against modern DDoS threats. This work represents a growing body of evidence supporting the validity of incorporating blockchain into the architecture of the next generation of cloud-based security.
Cybersecurity is one of the most pressing concerns with regard to autonomous systems' ever-increasing adoption across multiple sectors, including transportation, health care, and smart city developments. The aim of this chapter is to focus on the various methods of securing autonomous systems through artificial intelligence (AI)-powered intrusion detection systems (IDS) and privacy-preserving mechanisms. For example, this chapter will explore the use of machine learning for anomaly detection as well as secure federated learning and blockchain technology to enhance the integrity of data in autonomous systems. Furthermore, it will provide an overview of the use of adversarially attacking AI models and provide recommendations for reducing cyber risk. Utilizing AI-based security frameworks, autonomous systems can identify threats and respond to them almost instantly, while ensuring user privacy. Finally, this chapter addresses the regulatory hurdles surrounding autonomous technology, as well as potential areas for future research related to security in autonomous systems.
The rising use of the Internet of Things (IoT) has changed the communication and automation landscape in various industries. However, the growing number of interconnected and vulnerable IoT devices has created several cybersecurity challenges, and the conventional intrusion detection system is not designed to handle the dynamicity of sophisticated cyber-attacks and secure information management. This study presents a blockchain-based security framework for intrusion detection in an IoT environment that uses a Gated Recurrent Unit (GRU) to achieve high-level detection accuracy and blockchain technology to guarantee information security. Edge-IIoTset benchmark data containing about 2.2 million traffic instances and 61 traffic features were collected, preprocessed, and split into training, validation, and testing datasets at a ratio of 70:15:15 for model development and evaluation. The GRU network was trained to capture sequential patterns in network traffic with high accuracy, while the blockchain layer was leveraged to ensure secure detection record storage and information sharing. The model attained 99.12% accuracy, 99.08% precision, 98.97% recall, 99.02% F1-score, and 0.9987 ROC-AUC. Additionally, the blockchain layer achieved an average of 850 transactions per second with a 2.3-second block confirmation time, while the framework recorded an average of 3.2 millisecond traffic detection time. Thus, the proposed framework was efficient and effective in detecting and responding to cyber-attacks in an IoT network.
Cloud-based academic environments such as Learning Management Systems (LMS), Open Journal Systems (OJS), institutional repositories, and web applications face increasing cybersecurity challenges due to heterogeneous users, distributed services, and extensive exposure to public networks. Existing security approaches remain fragmented, where machine learning focuses on threat detection, Zero Trust Architecture (ZTA) emphasizes access control, and blockchain is primarily used for secure logging. The lack of integration among these components limits the ability of security systems to adapt dynamically to evolving cyber threats. This study proposes an Adaptive Cybersecurity Framework (ACF) that integrates unsupervised machine learning-based anomaly detection, a risk-based Zero Trust Policy Engine, and blockchain-based immutable audit logging within a continuous adaptive feedback loop. The framework was evaluated using 450,000 anonymized HTTP and Web Application Firewall (WAF) events collected from a multi-domain academic cloud environment consisting of LMS, OJS, repositories, and supporting web applications. The analysis revealed structured and repetitive attack behaviors dominated by automated endpoint probing and cross-domain propagation patterns, indicating ecosystem-level security threats. The proposed risk assessment mechanism demonstrated effective alignment between anomaly detection and policy-based decision making. Experimental results achieved an AUROC of 0.7296 for risk-based threat detection while maintaining an average decision latency of approximately 11 ms, indicating suitability for real-time deployment. Blockchain integration further provided verifiable, tamper-resistant audit trails for mitigation actions and policy enforcement activities. This study contributes an ecosystem-aware adaptive cybersecurity paradigm that bridges threat detection, policy enforcement, and auditability through a unified security architecture for Academic Cloud Environments.
The multiplication of digital infrastructures and cloud services and of cyber-physical systems connected with each other has resulted in the significant growth of cybersecurity challenges. Traditional security mechanisms are usually not qualified to process large quantities of security data and identify complex cyber threats in real-time. The combination of artificial intelligence (AI), parallel computing, and blockchain technologies show a promising way forward on creating secure, scalable and resilient cybersecurity frameworks for distributed digital infrastructures. This Chapter puts forth a conceptual framework that incorporates parallel computing architectures and block-chain enabled security mechanisms in order to increase the efficiency and trustworthiness of AI-enabled Cybersecurity systems. Parallel computing helps high-performance processing of big amounts of network traffic and security logs, so that AI models can perform better analysis of picker threat patterns and find out anomalies more efficiently.
G. N. Girish, Ashutosh Sahoo, Ajay Bhat, Akshay SP · 7 authors
Incentive programs are central to user acquisition in decentralized finance, but many reward systems rely on raw volume, transaction count, and wallet count, making them vulnerable to bots and sybil operations. We present ZAPs, a reward attribution framework that combines economic contribution scoring with adversarial robustness. A composite activity score uses protocol-specific percentile normalization to limit whale dominance while preserving differentiation among users. A two-layer weighting mechanism combines protocol share within sector and sector share within the ecosystem, which reduces the profitability of farming small protocols. We show that the maximum reward obtainable from any protocol is bounded by that protocol's global volume share. ZAPs also introduces a four-layer defense stack consisting of transaction-level integrity checks, a parallel anomaly ensemble, post-distribution behavioral memory, and graph-based sybil clustering. The anomaly ensemble combines a one-class reconstruction model with an isolation forest and applies graduated rather than binary penalties. On 1,073 labeled malicious wallets covering 124,638 transactions, the ensemble achieves 0.923 +/- 0.013 ROC-AUC, compared with 0.891 +/- 0.016 for the reconstruction model alone, when the isolation forest is trained on benign wallets. Training it on the pooled population reverses its polarity and removes the ensemble gain. Controlled simulations reduce adversarial reward capture by 30-90 percent while legitimate-user scenarios change by 1-8 percent. Live campaigns recorded a 56 percent reduction in sybil allocation, a 49 percent increase in quality-wallet participation, and a 50 percent reduction in sell pressure.
Decentralized Finance (DeFi) has suffered over $5 billion in cumulative losses from security incidents, yet the academic community lacks a large-scale, multi-source-verified dataset to systematically characterize these threats. We present DEFIHACK-824, a curated dataset of 823 DeFi security incidents spanning 2017 to 2026, cross-validated against three independent intelligence sources (Rekt News, SlowMist, and CertiK). Each record is annotated with attack category, confidence level (Gossip/Classified/Ground Truth), and estimated financial loss. We classify incidents into 14 attack categories and conduct statistical analyses: (1) flash-loan-enabled price manipulation and reentrancy together account for 51.5% of all attacks; (2) a chi-squared test rejects the null hypothesis of uniform category distribution at p < 0.0001 (chi-squared = 1,273.2, df = 13); (3) despite widespread deployment of automated detection tools, the annual attack count has not monotonically decreased. We further propose a six-layer DeFi threat model and quantify the effectiveness of four defense classes. The dataset, threat model, and 50 categorized Solidity vulnerability patterns are released under the MIT license.
Context The exponential evolution and widespread integration of Artificial Intelligence (AI) and Machine Learning (ML) systems have fundamentally transformed industries, establishing AI as a central component in decision-making processes, task automation, and the optimization of complex operational pipelines. From healthcare diagnostics to financial forecasting and increasingly across critical cybersecurity infrastructure such as intrusion detection systems and malware classifiers, AI models are being deployed in environments where the correctness and authenticity of their outputs carry direct operational and safety consequences. Nevertheless, as the deployment of AI systems becomes widespread, the conditions under which these models are trained have evolved in a direction where the security landscape of them radically changes. The traaditional assumption of a centralized, fully controlled training environment, where a single trusted entity acquires data, trains the model, and deploys it, no longer reflects the reality of modern machine learning practice. The frequent use of remote sensing, federated learning and/or outsourced machine learning has introduced architectures where the entity that acquires the data, the entity that trains the model and the entity that ultimately relies on the model's output are three distinct and mutually distrusting parties. In a remote sensing scenario, sensors owned by a data provider transmit raw measurements to a training node that may be geographically or administratively distant. In a federated learning scenario, multiple decentralized devices train local models on their private data and submit the results to a central aggregator. In an outsourced learning scenario, a resource-constrained model sponsor delegates the training computation entirely to a third-party cloud provider. In all three cases, the common factor is the same: the model sponsor, the entity that is ultimately responsible for and dependent on the trained model, that does not control the data acquisition process, does not observe the training execution and has no native mechanism to verify that the model they receive is the result of the computation they requested, performed on the data they provided. This separation of control is the main focus addressed by this dissertation. It is not merely a theoretical concern: the literature has documented a wide range of attacks that exploit precisely this gap. When a malicious trainer substitutes data, alters labels, ignores some dataset's subsets or modifies model parameters, the resulting model may appear functionally correct on standard evaluation metrics while being systematically compromised for specific classes of input, an attack vector particularly dangerous in cybersecurity applications where a model that has been quietly trained to misclassify a specific type of malicious traffic provides no observable anomaly until the attack it was designed to hide occurs. Problem and Motivation The main motivation of this dissertation can be addressed as follows. Given a sensor, that produces a set of data points in a given time frame, or a dataset owned by a data provider and a model computed by a model trainer from that data, the model sponsor wants to ensure that the trained model is the result of executing a known training process over the complete and authenticated dataset $D_t$. That is, all data points in $D_t$ and only those data points were used as the training set. No modifications were made to those points or their labels and the obtained model is indeed the result obtained from the execution of the agreed training algorithm. This guarantee cannot be provided by standard Machine Learning procedures, like accuracy, precision or F1-score. A malicious trainer can submit a model that passes all the standard evaluation metrics on benign inputs while maintaining a targeted misclassification on a specific attack pattern. The only way to close this gap is to make the training process itself verifiable by requiring the trainer to produce and submit a cryptographic proof that is mathematically impossible to forge without having correctly executed the agreed computation on the authenticated data. This verification challenge comes together with a second problem, the \emph{model integrity gap} that exists between a trained model and its deployed representation. Even if the training process was all validated, the model must subsequently be transpiled and deployed into a certain non-ML format. In the context of this dissertation, this gap is particularly sensitive, the Python model trained by the data scientist must be translated into a ZoKrates arithmetic circuit for zero-knowledge proof generation, a process that involves converting continuous floating-point decision boundaries into discrete integer arithmetic. If this translation introduces a small inversion in a comparison operator or a shifted threshold values, the deployed circuit will produce systematically different predictions from the intended model and standard testing may not surface the discrepancy. The literature has proposed cryptographic solutions to the verifiable training but has largely left the second problem unaddressed. The foundational work by Keshavarzkalhori et al. demonstrated that it is possible to construct a pipeline combining hash chains, digital signatures and zero-knowledge proofs to verify that a simulated Naive Bayes classifier was trained on authenticated sensor data. Their implementation, built on the ZoKrates toolset, provided a proof-of-concept that the building blocks exist for end-to-end training verification. However, scaling this approach from a simple probabilistic classifier to a more complex, non-linear ensemble model, in this specific case, a Random Forest, introduces severe architectural bottlenecks that their work explicitly identified as open problems: the computational overhead of bitwise hashing inside arithmetic circuits, the floating-point to integer translation problem and the absence of any mechanism to verify that the transpilation of the model into the circuit was performed faithfully. This dissertation directly addresses these open problems. It proposes, implements and evaluates an end-to-end verifiable machine learning architecture for Random Forest classifiers that provides mathematical guarantees over three distinct integrity boundaries: the origin of the training data, the correctness of the training computation and the fidelity of the model's translation into a verifiable circuit. The framework is evaluated on both a simulated sensor dataset used by Keshavarzkalhori et al. and the CICIDS2017 network intrusion detection benchmark, the real-world cybersecurity dataset used by the most directly comparable prior work, demonstrating that the proposed integrity guarantees are achievable at practical computational cost for cybersecurity-relevant workloads. Research Questions The main objective of this thesis was to build a framework capable of protecting the overall AI Models from data and model poisoning attacks. In alignment with the goal, four research questions were set: Research Question 01: What state-of-the-art mechanisms exist to verify the integrity of AI models across the training pipeline? Research Question 02: What threats exist against AI models integrity? Research Question 03: What computational overhead do integrity verification mechanisms introduce across the AI modeling pipeline and how does this overhead scale with model complexity?
Inundating networks with traffic to cripple service availability defines a DDoS attack. Traditional defences, like firewalls and centralized scrubbing centers, can suffer from single points of failure during large-scale attacks. Enter new blockchain technology, with Ethereum probably leading the way in decentralized solutions. Since Ethereum smart contracts enable DDoS detection and enforcement of validation rules, events can be managed automatically, applying rewards or penalties without a central authority. In this way it leverages the positive aspects of crypto-economic mechanisms and reputation systems; giving people an incentive to honestly participate while making abusing the system unprofitable, creating a trustless, transparent, resilient decentralized defence against cyber threats. The architecture of a blockchain-validated system that gates access to services includes a back-end gateway responsible for verifying transactions on the blockchain before processing user queries. The off-chain detection algorithm identifies unexpected traffic spikes that exceed a predefined threshold. Attackers were deterred by cost, the system itself stayed up and running, and the smart contract worked autonomously. The suggested approach maintained 96% access success for authorized users, successfully blocked 92% of DDoS traffic, and guaranteed 98% uptime during simulated attacks. All validated access attempts were 100% immutably recorded on-chain, and attackers had to pay 300% more employs transparent on-chain rules and Ethereum smart contracts to manage access. Because every access attempt is permanently documented on the blockchain, it is difficult to alter logs or stop denial-of-service attacks without detection.
Thomas Bakaysa, Ahmet Kurt, Abdul-Salem Beibitkhan, J E Hernández Leon · 9 authors
Bitcoin's Lightning Network (LN) can be exploited as a covert, low-cost command-and-control (C&C) channel for botnets, as demonstrated by the LNBot and D-LNBot designs. However, both remain proof-of-concept prototypes evaluated only through simulation, leaving key questions about real-world topology formation, propagation complexity, and resilience to takedowns unanswered. We present LNTest, the first reusable testbed for LN-based botnets, built from Core Lightning nodes containerized with Docker over a shared Bitcoin Core regtest chain. LNTest supports three overlay topology modes (a deterministic chain, autonomous peer discovery, and user-supplied graphs), enabling controlled experiments across different botnet structures. Using LNTest, we report three main findings. First, D-LNBot's autonomous formation protocol does not produce the uniform chain from its design; instead, it creates a clustered chain in which cliques are linked by bridge nodes whose removal fragments the network. Second, command propagation scales linearly with botnet size ($Θ(n)$), not the $O(m \log n)$ previously claimed, and gains nothing from higher neighbor connectivity. Third, the overlay topology determines the effectiveness of takedown strategies: uniform-degree chains resist targeted removal but fragment under random failure, scale-free topologies show the opposite pattern, and the autonomous clustered chain is fragile under both, making it the most vulnerable of the three. LNTest is released as open source, with a script that reproduces all our experiments, to support reproducible research on LN-based botnet defenses.
Muhammad Umar Janjua, Akshaya Mani, Uğur Şen, Daniel Kaiser
Privacy and anonymity of validators, especially regarding IP address linkability, are essential to protect the Ethereum network from various attacks. Network-level attacks, such as DoS, can interrupt validators and affect the overall security of the Ethereum network. Correlating the IP addresses of validators with their identities, along with knowledge about their action slots can be exploited by attackers to cause network delays, MEV exploitation, and finality risks. Therefore, ensuring the unlinkability of a validator's IP and identity is crucial for maintaining the network's trust and resilience. In this techreport, we first provide a review of the existing network and consensus layer techniques that have been proposed for maintaining validator privacy in the Ethereum blockchain. Secondly, we evaluate a Tor-based protocol named Tor push that helps unlink validator identities (IDs) from their nodes' IP addresses, thereby making it difficult to determine any end-to-end correlation between validator IDs and IP addresses of validators' beacon nodes. To evaluate the effectiveness of Tor push, we present a working, deployed proof-of-concept (PoC) implementation in the Nimbus Ethereum client. Our PoC deployment pushes attestations, aggregations, and block proposals over Tor to the Goerli testnet. Furthermore, we also analyse the security and latency of Tor push. Our experimental results suggest that Tor can be incorporated into the existing Ethereum network with a tolerable latency overhead of 613.82 ms on average and without compromising the overall network performance while enhancing the location privacy of validators in the Ethereum network.