arXiv subjects

Mauro Conti

Publications and source records attributed to Mauro Conti.

At least 73 records · Page 4Linked to original sources

arXiv2025

DroidTTP: Mapping Android Applications with TTP for Cyber Threat Intelligence

The widespread adoption of Android devices for sensitive operations like banking and communication has made them prime targets for cyber threats, particularly Advanced Persistent Threats (APT) and sophisticated malware attacks. Traditional malware detection methods rely on binary classification, failing to provide insights into adversarial Tactics, Techniques, and Procedures (TTPs). Understanding malware behavior is crucial for enhancing cybersecurity defenses. To address this gap, we introduce DroidTTP, a framework mapping Android malware behaviors to TTPs based on the MITRE ATT&CK framework. Our curated dataset explicitly links MITRE TTPs to Android applications. We developed an automated solution leveraging the Problem Transformation Approach (PTA) and Large Language Models (LLMs) to map applications to both Tactics and Techniques. Additionally, we employed Retrieval-Augmented Generation (RAG) with prompt engineering and LLM fine-tuning for TTP predictions. Our structured pipeline includes dataset creation, hyperparameter tuning, data augmentation, feature selection, model development, and SHAP-based model interpretability. Among LLMs, Llama achieved the highest performance in Tactic classification with a Jaccard Similarity of 0.9583 and Hamming Loss of 0.0182, and in Technique classification with a Jaccard Similarity of 0.9348 and Hamming Loss of 0.0127. However, the Label Powerset XGBoost model outperformed LLMs, achieving a Jaccard Similarity of 0.9893 for Tactic classification and 0.9753 for Technique classification, with a Hamming Loss of 0.0054 and 0.0050, respectively. While XGBoost showed superior performance, the narrow margin highlights the potential of LLM-based approaches in TTP classification.

cs.CR↗

arXiv2025

Leaky Batteries: A Novel Set of Side-Channel Attacks on Electric Vehicles

Advancements in battery technology have accelerated the adoption of Electric Vehicles (EVs) due to their environmental benefits. However, their growing sophistication introduces security and privacy challenges. Often seen as mere operational data, battery consumption patterns can unintentionally reveal critical information exploitable for malicious purposes. These risks go beyond privacy, impacting vehicle security and regulatory compliance. Despite these concerns, current research has largely overlooked the broader implications of battery consumption data exposure. As EVs integrate further into smart transportation networks, addressing these gaps is crucial to ensure their safety, reliability, and resilience. In this work, we introduce a novel class of side-channel attacks that exploit EV battery data to extract sensitive user information. Leveraging only battery consumption patterns, we demonstrate a methodology to accurately identify the EV driver and their driving style, determine the number of occupants, and infer the vehicle's start and end locations when user habits are known. We utilize several machine learning models and feature extraction techniques to analyze EV power consumption patterns, validating our approach on simulated and real-world datasets collected from actual drivers. Our attacks achieve an average success rate of 95.4% across all attack objectives. Our findings highlight the privacy risks associated with EV battery data, emphasizing the need for stronger protections to safeguard user privacy and vehicle security.

cs.CR↗

arXiv2025

Through the Static: Demystifying Malware Visualization via Explainability

Security researchers grapple with the surge of malicious files, necessitating swift identification and classification of malware strains for effective protection. Visual classifiers and in particular Convolutional Neural Networks (CNNs) have emerged as vital tools for this task. However, issues of robustness and explainability, common in other high risk domain like medicine and autonomous vehicles, remain understudied in current literature. Although deep learning visualization classifiers presented in research obtain great results without the need for expert feature extraction, they have not been properly studied in terms of their replicability. Additionally, the literature is not clear on how these types of classifiers arrive to their answers. Our study addresses these gaps by replicating six CNN models and exploring their pitfalls. We employ Class Activation Maps (CAMs), like GradCAM and HiResCAM, to assess model explainability. We evaluate the CNNs' performance and interpretability on two standard datasets, MalImg and Big2015, and a newly created called VX-Zoo. We employ these different CAM techniques to gauge the explainability of each of the models. With these tools, we investigate the underlying factors contributing to different interpretations of inputs across the different models, empowering human researchers to discern patterns crucial for identifying distinct malware families and explain why CNN models arrive at their conclusions. Other then highlighting the patterns found in the interpretability study, we employ the extracted heatmpas to enhance Visual Transformers classifiers' performance and explanation quality. This approach yields substantial improvements in F1 score, ranging from 2% to 8%, across the datasets compared to benchmark values.

cs.CR↗

arXiv2025

The Road Less Traveled: Investigating Robustness and Explainability in CNN Malware Detection

Machine learning has become a key tool in cybersecurity, improving both attack strategies and defense mechanisms. Deep learning models, particularly Convolutional Neural Networks (CNNs), have demonstrated high accuracy in detecting malware images generated from binary data. However, the decision-making process of these black-box models remains difficult to interpret. This study addresses this challenge by integrating quantitative analysis with explainability tools such as Occlusion Maps, HiResCAM, and SHAP to better understand CNN behavior in malware classification. We further demonstrate that obfuscation techniques can reduce model accuracy by up to 50%, and propose a mitigation strategy to enhance robustness. Additionally, we analyze heatmaps from multiple tests and outline a methodology for identification of artifacts, aiding researchers in conducting detailed manual investigations. This work contributes to improving the interpretability and resilience of deep learning-based intrusion detection systems

cs.CR↗

arXiv2025

Differential Area Analysis for Ransomware: Attacks, Countermeasures, and Limitations

Crypto-ransomware attacks have been a growing threat over the last few years. The goal of every ransomware strain is encrypting user data, such that attackers can later demand users a ransom for unlocking their data. To maximise their earning chances, attackers equip their ransomware with strong encryption which produce files with high entropy values. Davies et al. proposed Differential Area Analysis (DAA), a technique that analyses files headers to differentiate compressed, regularly encrypted, and ransomware-encrypted files. In this paper, first we propose three different attacks to perform malicious header manipulation and bypass DAA detection. Then, we propose three countermeasures, namely 2-Fragments (2F), 3-Fragments (3F), and 4-Fragments (4F), which can be applied equally against each of the three attacks we propose. We conduct a number of experiments to analyse the ability of our countermeasures to detect ransomware-encrypted files, whether implementing our proposed attacks or not. Last, we test the robustness of our own countermeasures by analysing the performance, in terms of files per second analysed and resilience to extensive injection of low-entropy data. Our results show that our detection countermeasures are viable and deployable alternatives to DAA.

cs.CR↗

arXiv2024

Topology-Based Reconstruction Prevention for Decentralised Learning

Decentralised learning has recently gained traction as an alternative to federated learning in which both data and coordination are distributed. To preserve the confidentiality of users' data, decentralised learning relies on differential privacy, multi-party computation, or both. However, running multiple privacy-preserving summations in sequence may allow adversaries to perform reconstruction attacks. Current reconstruction countermeasures either cannot trivially be adapted to the distributed setting, or add excessive amounts of noise. In this work, we first show that passive honest-but-curious adversaries can infer other users' private data after several privacy-preserving summations. For example, in subgraphs with 18 users, we show that only three passive honest-but-curious adversaries succeed at reconstructing private data 11.0% of the time, requiring an average of 8.8 summations per adversary. The success rate depends only on the adversaries' direct neighbourhood, and is independent of the size of the full network. We consider weak adversaries that do not control the graph topology, cannot exploit the summation's inner workings, and do not have auxiliary knowledge; and show that these adversaries can still infer private data. We analyse how reconstruction relates to topology and propose the first topology-based decentralised defence against reconstruction attacks. We show that reconstruction requires a number of adversaries linear in the length of the network's shortest cycle. Consequently, exact attacks over privacy-preserving summations are impossible in acyclic networks. Our work is a stepping stone for a formal theory of topology-based decentralised reconstruction defences. Such a theory would generalise our countermeasure beyond summation, define confidentiality in terms of entropy, and describe the interactions with (topology-aware) differential privacy.

cs.CR↗

arXiv2024

IntellBot: Retrieval Augmented LLM Chatbot for Cyber Threat Knowledge Delivery

In the rapidly evolving landscape of cyber security, intelligent chatbots are gaining prominence. Artificial Intelligence, Machine Learning, and Natural Language Processing empower these chatbots to handle user inquiries and deliver threat intelligence. This helps cyber security knowledge readily available to both professionals and the public. Traditional rule-based chatbots often lack flexibility and struggle to adapt to user interactions. In contrast, Large Language Model-based chatbots offer contextually relevant information across multiple domains and adapt to evolving conversational contexts. In this work, we develop IntellBot, an advanced cyber security Chatbot built on top of cutting-edge technologies like Large Language Models and Langchain alongside a Retrieval-Augmented Generation model to deliver superior capabilities. This chatbot gathers information from diverse data sources to create a comprehensive knowledge base covering known vulnerabilities, recent cyber attacks, and emerging threats. It delivers tailored responses, serving as a primary hub for cyber security insights. By providing instant access to relevant information and resources, this IntellBot enhances threat intelligence, incident response, and overall security posture, saving time and empowering users with knowledge of cyber security best practices. Moreover, we analyzed the performance of our copilot using a two-stage evaluation strategy. We achieved BERT score above 0.8 by indirect approach and a cosine similarity score ranging from 0.8 to 1, which affirms the accuracy of our copilot. Additionally, we utilized RAGAS to evaluate the RAG model, and all evaluation metrics consistently produced scores above 0.77, highlighting the efficacy of our system.

cs.IR↗

arXiv2024

Towards a Self-rescuing System for UAVs Under GNSS Attack

There has been substantial growth in the UAV market along with an expansion in their applications. However, the successful execution of a UAV mission is very often dependent on the use of a GNSS. Unfortunately, the vulnerability of GNSS signals, due to their lack of encryption and authentication, poses a significant cybersecurity issue. This vulnerability makes various attacks, particularly the "GNSS spoofing attack," and "GNSS jamming attack" easily executable. Generally speaking, during this attack, the drone is manipulated into altering its path, usually resulting in an immediate forced landing or crash. As far as we know, we are the first to propose a lightweight-solution that enable a drone to autonomously rescue itself, assuming it is under GNSS attack and the GNSS is no longer available, and return safely to its initial takeoff position, thereby preventing any potential crashes. During the flight, wind plays a critical role as it can instantaneously alter the drone's position. To solve this problem, we have devised a highly effective 2-phases solution: (i) Forward Phase, for monitoring and recording the forward journey, and (ii) Backward Phase, that generates a backward route, based on the Forward Phase and wind presence. The final solution ensures strong performance in consistently returning the drone to the original position, even in wind situations, while maintaining a very fast computation time.

cs.CR↗

arXiv2024

Membership Privacy Evaluation in Deep Spiking Neural Networks

Artificial Neural Networks (ANNs), commonly mimicking neurons with non-linear functions to output floating-point numbers, consistently receive the same signals of a data point during its forward time. Unlike ANNs, Spiking Neural Networks (SNNs) get various input signals in the forward time of a data point and simulate neurons in a biologically plausible way, i.e., producing a spike (a binary value) if the accumulated membrane potential of a neuron is larger than a threshold. Even though ANNs have achieved remarkable success in multiple tasks, e.g., face recognition and object detection, SNNs have recently obtained attention due to their low power consumption, fast inference, and event-driven properties. While privacy threats against ANNs are widely explored, much less work has been done on SNNs. For instance, it is well-known that ANNs are vulnerable to the Membership Inference Attack (MIA), but whether the same applies to SNNs is not explored. In this paper, we evaluate the membership privacy of SNNs by considering eight MIAs, seven of which are inspired by MIAs against ANNs. Our evaluation results show that SNNs are more vulnerable (maximum 10% higher in terms of balanced attack accuracy) than ANNs when both are trained with neuromorphic datasets (with time dimension). On the other hand, when training ANNs or SNNs with static datasets (without time dimension), the vulnerability depends on the dataset used. If we convert ANNs trained with static datasets to SNNs, the accuracy of MIAs drops (maximum 11.5% with a reduction of 7.6% on the test accuracy of the target model). Next, we explore the impact factors of MIAs on SNNs by conducting a hyperparameter study. Finally, we show that the basic data augmentation method for static data and two recent data augmentation methods for neuromorphic data can considerably (maximum reduction of 25.7%) decrease MIAs' performance on SNNs.

cs.CR↗

arXiv2024

Subject Data Auditing via Source Inference Attack in Cross-Silo Federated Learning

Source Inference Attack (SIA) in Federated Learning (FL) aims to identify which client used a target data point for local model training. It allows the central server to audit clients' data usage. In cross-silo FL, a client (silo) collects data from multiple subjects (e.g., individuals, writers, or devices), posing a risk of subject information leakage. Subject Membership Inference Attack (SMIA) targets this scenario and attempts to infer whether any client utilizes data points from a target subject in cross-silo FL. However, existing results on SMIA are limited and based on strong assumptions on the attack scenario. Therefore, we propose a Subject-Level Source Inference Attack (SLSIA) by removing critical constraints that only one client can use a target data point in SIA and imprecise detection of clients utilizing target subject data in SMIA. The attacker, positioned on the server side, controls a target data source and aims to detect all clients using data points from the target subject. Our strategy leverages a binary attack classifier to predict whether the embeddings returned by a local model on test data from the target subject include unique patterns that indicate a client trains the model with data from that subject. To achieve this, the attacker locally pre-trains models using data derived from the target subject and then leverages them to build a training set for the binary attack classifier. Our SLSIA significantly outperforms previous methods on three datasets. Specifically, SLSIA achieves a maximum average accuracy of 0.88 over 50 target subjects. Analyzing embedding distribution and input feature distance shows that datasets with sparse subjects are more susceptible to our attack. Finally, we propose to defend our SLSIA using item-level and subject-level differential privacy mechanisms.

cs.CR↗

arXiv2024

Effectiveness of learning-based image codecs on fingerprint storage

The success of learning-based coding techniques and the development of learning-based image coding standards, such as JPEG-AI, point towards the adoption of such solutions in different fields, including the storage of biometric data, like fingerprints. However, the peculiar nature of learning-based compression artifacts poses several issues concerning their impact and effectiveness on extracting biometric features and landmarks, e.g., minutiae. This problem is utterly stressed by the fact that most models are trained on natural color images, whose characteristics are very different from usual biometric images, e.g, fingerprint or iris pictures. As a matter of fact, these issues are deemed to be accurately questioned and investigated, being such analysis still largely unexplored. This study represents the first investigation about the adaptability of learning-based image codecs in the storage of fingerprint images by measuring its impact on the extraction and characterization of minutiae. Experimental results show that at a fixed rate point, learned solutions considerably outperform previous fingerprint coding standards, like JPEG2000, both in terms of distortion and minutiae preservation. Indeed, experimental results prove that the peculiarities of learned compression artifacts do not prevent automatic fingerprint identification (since minutiae types and locations are not significantly altered), nor do compromise image quality for human visual inspection (as they gain in terms of BD rate and PSNR of 47.8% and +3.97dB respectively).

eess.IV↗

arXiv2024

EmoBack: Backdoor Attacks Against Speaker Identification Using Emotional Prosody

Speaker identification (SI) determines a speaker's identity based on their spoken utterances. Previous work indicates that SI deep neural networks (DNNs) are vulnerable to backdoor attacks. Backdoor attacks involve embedding hidden triggers in DNNs' training data, causing the DNN to produce incorrect output when these triggers are present during inference. This is the first work that explores SI DNNs' vulnerability to backdoor attacks using speakers' emotional prosody, resulting in dynamic, inconspicuous triggers. We conducted a parameter study using three different datasets and DNN architectures to determine the impact of emotions as backdoor triggers on the accuracy of SI systems. Additionally, we have explored the robustness of our attacks by applying defenses like pruning, STRIP-ViTA, and three popular preprocessing techniques: quantization, median filtering, and squeezing. Our findings show that the aforementioned models are prone to our attack, indicating that emotional triggers (sad and neutral prosody) can be effectively used to compromise the integrity of SI systems. However, the results of our pruning experiments suggest potential solutions for reinforcing the models against our attacks, decreasing the attack success rate up to 40%.

cs.CR↗

arXiv2024

Kronos: A Secure and Generic Sharding Blockchain Consensus with Optimized Overhead

Sharding enhances blockchain scalability by dividing the network into shards, each managing specific unspent transaction outputs or accounts. As an introduced new transaction type, cross-shard transactions pose a critical challenge to the security and efficiency of sharding blockchains. Currently, there is a lack of a generic sharding consensus pattern that achieves both security and low overhead. In this paper, we present Kronos, a secure sharding blockchain consensus achieving optimized overhead. In particular, we propose a new secure sharding consensus pattern, based on a buffer managed jointly by shard members. Valid transactions are transferred to the payee via the buffer, while invalid ones are rejected through happy or unhappy paths. Kronos is proved to achieve security with atomicity under malicious clients with optimal intra-shard overhead $kB$ ($k$ for involved shard number and $B$ for a Byzantine fault tolerance (BFT) cost). Besides, we propose secure cross-shard certification methods based on batch certification and reliable cross-shard transfer. The former combines hybrid trees or vector commitments, while the latter integrates erasure coding. Handling $b$ transactions, Kronos is proved to achieve reliability with low cross-shard overhead $O(n b λ)$ ($n$ for shard size and $λ$ for the security parameter). Notably, Kronos imposes no restrictions on BFT and does not rely on time assumptions, offering optional constructions in various modules. We implement Kronos using two prominent BFT protocols: asynchronous Speeding Dumbo and partial synchronous Hotstuff. Extensive experiments demonstrate Kronos scales the consensus nodes to thousands, achieving a substantial throughput of 320 ktx/sec with 2.0 sec latency. Compared with the past solutions, Kronos outperforms, achieving up to a 12* improvement in throughput and a 50% reduction in latency.

cs.CR↗

arXiv2024

The Impact of SBOM Generators on Vulnerability Assessment in Python: A Comparison and a Novel Approach

The Software Supply Chain (SSC) security is a critical concern for both users and developers. Recent incidents, like the SolarWinds Orion compromise, proved the widespread impact resulting from the distribution of compromised software. The reliance on open-source components, which constitute a significant portion of modern software, further exacerbates this risk. To enhance SSC security, the Software Bill of Materials (SBOM) has been promoted as a tool to increase transparency and verifiability in software composition. However, despite its promise, SBOMs are not without limitations. Current SBOM generation tools often suffer from inaccuracies in identifying components and dependencies, leading to the creation of erroneous or incomplete representations of the SSC. Despite existing studies exposing these limitations, their impact on the vulnerability detection capabilities of security tools is still unknown. In this paper, we perform the first security analysis on the vulnerability detection capabilities of tools receiving SBOMs as input. We comprehensively evaluate SBOM generation tools by providing their outputs to vulnerability identification software. Based on our results, we identify the root causes of these tools' ineffectiveness and propose PIP-sbom, a novel pip-inspired solution that addresses their shortcomings. PIP-sbom provides improved accuracy in component identification and dependency resolution. Compared to best-performing state-of-the-art tools, PIP-sbom increases the average precision and recall by 60%, and reduces by ten times the number of false positives.

cs.CR↗

arXiv2024

Can LLMs Understand Computer Networks? Towards a Virtual System Administrator

Recent advancements in Artificial Intelligence, and particularly Large Language Models (LLMs), offer promising prospects for aiding system administrators in managing the complexity of modern networks. However, despite this potential, a significant gap exists in the literature regarding the extent to which LLMs can understand computer networks. Without empirical evidence, system administrators might rely on these models without assurance of their efficacy in performing network-related tasks accurately. In this paper, we are the first to conduct an exhaustive study on LLMs' comprehension of computer networks. We formulate several research questions to determine whether LLMs can provide correct answers when supplied with a network topology and questions on it. To assess them, we developed a thorough framework for evaluating LLMs' capabilities in various network-related tasks. We evaluate our framework on multiple computer networks employing proprietary (e.g., GPT4) and open-source (e.g., Llama2) models. Our findings in general purpose LLMs using a zero-shot scenario demonstrate promising results, with the best model achieving an average accuracy of 79.3%. Proprietary LLMs achieve noteworthy results in small and medium networks, while challenges persist in comprehending complex network topologies, particularly for open-source models. Moreover, we provide insight into how prompt engineering can enhance the accuracy of some tasks.

cs.NI↗

arXiv2024

MoRSE: Bridging the Gap in Cybersecurity Expertise with Retrieval Augmented Generation

In this paper, we introduce MoRSE (Mixture of RAGs Security Experts), the first specialised AI chatbot for cybersecurity. MoRSE aims to provide comprehensive and complete knowledge about cybersecurity. MoRSE uses two RAG (Retrieval Augmented Generation) systems designed to retrieve and organize information from multidimensional cybersecurity contexts. MoRSE differs from traditional RAGs by using parallel retrievers that work together to retrieve semantically related information in different formats and structures. Unlike traditional Large Language Models (LLMs) that rely on Parametric Knowledge Bases, MoRSE retrieves relevant documents from Non-Parametric Knowledge Bases in response to user queries. Subsequently, MoRSE uses this information to generate accurate answers. In addition, MoRSE benefits from real-time updates to its knowledge bases, enabling continuous knowledge enrichment without retraining. We have evaluated the effectiveness of MoRSE against other state-of-the-art LLMs, evaluating the system on 600 cybersecurity specific questions. The experimental evaluation has shown that the improvement in terms of relevance and correctness of the answer is more than 10\% compared to known solutions such as GPT-4 and Mixtral 7x8.

cs.CR↗

arXiv2024

Your Car Tells Me Where You Drove: A Novel Path Inference Attack via CAN Bus and OBD-II Data

Despite its well-known security issues, the Controller Area Network (CAN) is still the main technology for in-vehicle communications. Attackers posing as diagnostic services or accessing the CAN bus can threaten the drivers' location privacy to know the exact location at a certain point in time or to infer the visited areas. This represents a serious threat to users' privacy, but also an advantage for police investigations to gather location-based evidence. In this paper, we present On Path Diagnostic - Intrusion \& Inference (OPD-II), a novel path inference attack leveraging a physical car model and a map matching algorithm to infer the path driven by a car based on CAN bus data. Differently from available attacks, our approach only requires the attacker to know the initial location and heading of the victim's car and is not limited by the availability of training data, road configurations, or the need to access other victim's devices (e.g., smartphones). We implement our attack on a set of four different cars and a total number of 41 tracks in different road and traffic scenarios. We achieve an average of 95% accuracy on reconstructing the coordinates of the recorded path by leveraging a dynamic map-matching algorithm that outperforms the 75% and 89% accuracy values of other proposals while removing their set of assumptions.

cs.CR↗

arXiv2024

Fingerprint Membership and Identity Inference Against Generative Adversarial Networks

Generative models are gaining significant attention as potential catalysts for a novel industrial revolution. Since automated sample generation can be useful to solve privacy and data scarcity issues that usually affect learned biometric models, such technologies became widely spread in this field. In this paper, we assess the vulnerabilities of generative machine learning models concerning identity protection by designing and testing an identity inference attack on fingerprint datasets created by means of a generative adversarial network. Experimental results show that the proposed solution proves to be effective under different configurations and easily extendable to other biometric measurements.

cs.CV↗