arXiv subjects

Idilio Drago

Publications and source records attributed to Idilio Drago.

At least 19 recordsLinked to original sources

arXiv2026

AgentLSD: Evaluating AI Security Agents Under Adversarial Task Contamination

AI agents for security inspect web pages, source code, logs, configuration files, and command outputs. These environments may contain deceptive artifacts that influence the agent's behavior. We call this adversarial task contamination. Whereas prompt injection relies on attacker-supplied instructions, task contamination also includes non-instructional evidence, such as fake results and decoy endpoints. We present AgentLSD, a controlled framework for studying adversarial task contamination. AgentLSD uses Capture the Flag (CTF) challenges as its experimental environment. We inject trap artifacts, such as fake flags, misleading hints, decoy endpoints, and hidden cues, while preserving the intended CTF solution. The framework supports paired clean and trap-augmented experiments with deterministic trap generation, runtime injection, telemetry, and delivery verification. We evaluate six models on 11 web CTF challenges. In the clean condition, agents capture 41% of the flags, and no model solves every challenge. We then measure the impact of task contamination. Even when the agent still recovers the flag, traps increase the number of turns (+20) and reasoning tokens (+2k). Solve-rate effects are more heterogeneous, as some model-challenge pairs are largely unaffected while others follow decoys or submit wrong flags. These results show that clean CTF performance understates vulnerability to deceptive task evidence. AgentLSD isolates this effect and provides a reproducible benchmark for studying it. We release the framework, configurations, trap specifications, and raw traces.

cs.CR↗

arXiv2026

ChamaleoNet: Programmable Passive Probe for Enhanced Visibility on Erroneous Traffic

Traffic visibility remains a key component for management and security operations. Observing erroneous traffic, i.e., unanswered requests or error messages, is fundamental to detecting misconfiguration, temporary failures or attacks. ChamaleoNet transforms any production network into a transparent monitor to let administrators collect such erroneous traffic. ChamaleoNet is programmed to ignore well-formed traffic and record only erroneous packets, including those generated by misconfigured or infected internal hosts, and those sent by external actors that scan for services. Engineering such a system poses several challenges, from scalability to privacy. Leveraging the Software-Defined Networking (SDN) paradigm, ChamaleoNet processes the humongous amount of traffic flowing through the network border and focuses on erroneous packets only, lowering the pressure on the collection system. Moreover, it offers traffic anonymisation to conform to privacy regulations. ChamaleoNet enables the seamless integration with active deceptive systems like honeypots that can impersonate hosts/ports/services and engage with senders. In an operational scenario, we show that the SDN in-hardware filtering reduces the traffic to the controller by 90%, resulting in a scalable solution, which we offer as open source. Simple statistical analytics unveil the precious information carried by erroneous traffic. We discover internal misconfigured and infected hosts, identify temporary failures, and show enhanced visibility on attackers' scanning activities that look for vulnerable services.

cs.CR↗

arXiv2026

Holoscope: Open and Lightweight Telescope & Honeypot Platform

The complexity and scale of Internet attacks call for distributed, cooperative observatories capable of monitoring malicious traffic across diverse networks. Holoscope is an open, lightweight, and cloud-native platform designed to simplify the deployment and management of telescope (passive) and honeypot (active) sensors. Built upon K3s and WireGuard, Holoscope offers secure connectivity, automated sensor onboarding, and resilient operation even in resource-constrained environments. Through modular design and Infrastructure-as-Code principles, it supports dynamic sensor orchestration, automated recovery, and data processing. We build, deploy, and operate Holoscope across multiple institutions and cloud networks in Europe and Brazil, enabling unified visibility into large-scale attack phenomena while maintaining ease of integration and security compliance.

cs.DC↗

arXiv2026

Dense Contexts Are Hard Contexts: Lexical Density Limits Effective Context in LLMs

Input length and the position of relevant information are widely cited as the primary causes of degraded LLM long-context performance. Here, we study lexical density -- the rate at which a context introduces distinct information -- as a third, largely overlooked factor that systematically reduces the effective context window of LLMs. We quantify the impact of lexical density on open-weight LLMs (9B-685B) using three "find-the-needle" style benchmarks with identical length (~12k tokens) and controlled needle position, but increasing density of information. We observe a sharp performance collapse in higher-density benchmarks: models that are near-perfect in sparse contexts drop below 60% retrieval score on denser ones. To rule out task-type confounds, we vary and control the density within each benchmark while keeping all other properties unchanged. Reducing density generally restores performance, especially in the high-density regimes where degradation appears. These results show that effective context capacity is a function of lexical density, with direct implications for real-world LLM systems operating on compact, information-rich inputs.

cs.CL↗

arXiv2026

Autonomous LLM Agents & CTFs: A Second Look

Large Language Model (LLM) agents are increasingly proposed to automate offensive security tasks, with recent studies reporting near human-level success rates in Capture-the-Flag (CTF) challenges. We here revisit these results, providing a second look at these claims. We engineer different agent architectures of increasing complexity and modularity on 30 web-based CTFs challenges spanning 14 vulnerability classes. We instantiate these agents with multiple LLM backbones, and compare them with claude-code, a general-purpose agent that automatically determines its internal architecture. Our evaluation yields three main findings. First, claude-code achieves performance comparable to the engineered architectures (19/30 solved tasks), suggesting that general-purpose agents are strong baselines for offensive security tasks. Second, both our architectures and claude-code struggle in the same challenge categories, revealing persistent barriers that keep current agents below human-level capability. Third, by leveraging our manually designed architectures we can systematically measure the impact of additional components, finding that structured orchestration of specialized roles outperforms monolithic designs, improving run-to-run consistency, and reducing execution costs.

cs.CR↗

arXiv2026

OntoLogX: Ontology-Guided Knowledge Graph Extraction from Cybersecurity Logs with Large Language Models

System logs represent a valuable source of Cyber Threat Intelligence (CTI), capturing attacker behaviors, exploited vulnerabilities, and traces of malicious activity. Yet their utility is often limited by lack of structure, semantic inconsistency, and fragmentation across devices and sessions. Extracting actionable CTI from logs therefore requires approaches that can reconcile noisy, heterogeneous data into coherent and interoperable representations. We introduce OntoLogX, an autonomous Artificial Intelligence (AI) agent that leverages Large Language Models (LLMs) to transform raw logs into ontology-grounded Knowledge Graphs (KGs). OntoLogX integrates a lightweight log ontology with Retrieval Augmented Generation (RAG) and iterative correction steps, ensuring that generated KGs are syntactically and semantically valid. Beyond event-level analysis, the system aggregates KGs into sessions and employs a LLM to predict MITRE ATT&CK tactics, linking low-level log evidence to higher-level adversarial objectives. We evaluate OntoLogX on both logs from a public benchmark and a real-world honeypot dataset, demonstrating robust KG generation across multiple KGs backends and accurate mapping of adversarial activity to ATT&CK tactics. Results highlight the benefits of retrieval and correction for precision and recall, the effectiveness of code-oriented models in structured log analysis, and the value of ontology-grounded representations for actionable CTI extraction.

cs.AI↗

arXiv2026

Improving Generalization on Cybersecurity Tasks with Multi-Modal Contrastive Learning

The use of ML in cybersecurity has long been impaired by generalization issues: Models that work well in controlled scenarios fail to maintain performance in production. The root cause often lies in ML algorithms learning superficial patterns (shortcuts) rather than underlying cybersecurity concepts. We investigate contrastive multi-modal learning as a first step towards improving ML performance in cybersecurity tasks. We aim at transferring knowledge from data-rich modalities, such as text, to data-scarce modalities, such as payloads. We set up a case study on threat classification and propose a two-stage multi-modal contrastive learning framework that uses textual vulnerability descriptions to guide payload classification. First, we construct a semantically meaningful embedding space using contrastive learning on descriptions. Then, we align payloads to this space, transferring knowledge from text to payloads. We evaluate the approach on a large-scale private dataset and a synthetic benchmark built from public CVE descriptions and LLM-generated payloads. The methodology appears to reduce shortcut learning over baselines on both benchmarks. We release our synthetic benchmark and source code as open source.

cs.CR↗

arXiv2026

Towards Agentic Honeynet Configuration

Honeypots are deception systems that emulate vulnerable services to collect threat intelligence. While deploying many honeypots increases the opportunity to observe attacker behaviour, in practise network and computational resources limit the number of honeypots that can be exposed. Hence, practitioners must select the assets to deploy, a decision that is typically made statically despite attackers' tactics evolving over time. This work investigates an AI-driven agentic architecture that autonomously manages honeypot exposure in response to ongoing attacks. The proposed agent analyses Intrusion Detection System (IDS) alerts and network state to infer the progression of the attack, identify compromised assets, and predict likely attacker targets. Based on this assessment, the agent dynamically reconfigures the system to maintain attacker engagement while minimizing unnecessary exposure. The approach is evaluated in a simulated environment where attackers execute Proof-of-Concept exploits for known CVEs. Preliminary results indicate that the agent can effectively infer the intent of the attacker and improve the efficiency of exposure under resource constraints

cs.CR↗

arXiv2025

ARBITER: AI-Driven Filtering for Role-Based Access Control

Role-Based Access Control (RBAC) struggles to adapt to dynamic enterprise environments with documents that contain information that cannot be disclosed to specific user groups. As these documents are used by LLM-driven systems (e.g., in RAG) the problem is exacerbated as LLMs can leak sensitive data due to prompt truncation, classification errors, or loss of system context. We introduce \our, a system designed to provide RBAC in RAG systems. \our implements layered input/output validation, role-aware retrieval, and post-generation fact-checking. Unlike traditional RBAC approaches that rely on fine-tuned classifiers, \our uses LLMs operating in few-shot settings with prompt-based steering for rapid deployment and role updates. We evaluate the approach on 389 queries using a synthetic dataset. Experimental results show 85\% accuracy and 89\% F1-score in query filtering, close to traditional RBAC solutions. Results suggest that practical RBAC deployment on RAG systems is approaching the maturity level needed for dynamic enterprise environments.

cs.CR↗

arXiv2024

AutoPenBench: Benchmarking Generative Agents for Penetration Testing

Generative AI agents, software systems powered by Large Language Models (LLMs), are emerging as a promising approach to automate cybersecurity tasks. Among the others, penetration testing is a challenging field due to the task complexity and the diverse strategies to simulate cyber-attacks. Despite growing interest and initial studies in automating penetration testing with generative agents, there remains a significant gap in the form of a comprehensive and standard framework for their evaluation and development. This paper introduces AutoPenBench, an open benchmark for evaluating generative agents in automated penetration testing. We present a comprehensive framework that includes 33 tasks, each representing a vulnerable system that the agent has to attack. Tasks are of increasing difficulty levels, including in-vitro and real-world scenarios. We assess the agent performance with generic and specific milestones that allow us to compare results in a standardised manner and understand the limits of the agent under test. We show the benefits of AutoPenBench by testing two agent architectures: a fully autonomous and a semi-autonomous supporting human interaction. We compare their performance and limitations. For example, the fully autonomous agent performs unsatisfactorily achieving a 21% Success Rate (SR) across the benchmark, solving 27% of the simple tasks and only one real-world task. In contrast, the assisted agent demonstrates substantial improvements, with 64% of SR. AutoPenBench allows us also to observe how different LLMs like GPT-4o or OpenAI o1 impact the ability of the agents to complete the tasks. We believe that our benchmark fills the gap with a standard and flexible framework to compare penetration testing agents on a common ground. We hope to extend AutoPenBench along with the research community by making it available under https://github.com/lucagioacchini/auto-pen-bench.

cs.CR↗

arXiv2024

Generic Multi-modal Representation Learning for Network Traffic Analysis

Network traffic analysis is fundamental for network management, troubleshooting, and security. Tasks such as traffic classification, anomaly detection, and novelty discovery are fundamental for extracting operational information from network data and measurements. We witness the shift from deep packet inspection and basic machine learning to Deep Learning (DL) approaches where researchers define and test a custom DL architecture designed for each specific problem. We here advocate the need for a general DL architecture flexible enough to solve different traffic analysis tasks. We test this idea by proposing a DL architecture based on generic data adaptation modules, followed by an integration module that summarises the extracted information into a compact and rich intermediate representation (i.e. embeddings). The result is a flexible Multi-modal Autoencoder (MAE) pipeline that can solve different use cases. We demonstrate the architecture with traffic classification (TC) tasks since they allow us to quantitatively compare results with state-of-the-art solutions. However, we argue that the MAE architecture is generic and can be used to learn representations useful in multiple scenarios. On TC, the MAE performs on par or better than alternatives while avoiding cumbersome feature engineering, thus streamlining the adoption of DL solutions for traffic analysis.

cs.LG↗

arXiv2024

LogPrécis: Unleashing Language Models for Automated Malicious Log Analysis

The collection of security-related logs holds the key to understanding attack behaviors and diagnosing vulnerabilities. Still, their analysis remains a daunting challenge. Recently, Language Models (LMs) have demonstrated unmatched potential in understanding natural and programming languages. The question arises whether and how LMs could be also useful for security experts since their logs contain intrinsically confused and obfuscated information. In this paper, we systematically study how to benefit from the state-of-the-art in LM to automatically analyze text-like Unix shell attack logs. We present a thorough design methodology that leads to LogPrécis. It receives as input raw shell sessions and automatically identifies and assigns the attacker tactic to each portion of the session, i.e., unveiling the sequence of the attacker's goals. We demonstrate LogPrécis capability to support the analysis of two large datasets containing about 400,000 unique Unix shell attacks. LogPrécis reduces them into about 3,000 fingerprints, each grouping sessions with the same sequence of tactics. The abstraction it provides lets the analyst better understand attacks, identify fingerprints, detect novelty, link similar attacks, and track families and mutations. Overall, LogPrécis, released as open source, paves the way for better and more responsive defense against cyberattacks.

cs.CR↗

arXiv2023

Sound-skwatter (Did You Mean: Sound-squatter?) AI-powered Generator for Phishing Prevention

Sound-squatting is a phishing attack that tricks users into malicious resources by exploiting similarities in the pronunciation of words. Proactive defense against sound-squatting candidates is complex, and existing solutions rely on manually curated lists of homophones. We here introduce Sound-skwatter, a multi-language AI-based system that generates sound-squatting candidates for proactive defense. Sound-skwatter relies on an innovative multi-modal combination of Transformers Networks and acoustic models to learn sound similarities. We show that Sound-skwatter can automatically list known homophones and thousands of high-quality candidates. In addition, it covers cross-language sound-squatting, i.e., when the reader and the listener speak different languages, supporting any combination of languages. We apply Sound-skwatter to network-centric phishing via squatted domain names. We find ~ 10% of the generated domains exist in the wild, the vast majority unknown to protection solutions. Next, we show attacks on the PyPI package manager, where ~ 17% of the popular packages have at least one existing candidate. We believe Sound-skwatter is a crucial asset to mitigate the sound-squatting phenomenon proactively on the Internet. To increase its impact, we publish an online demo and release our models and code as open source.

cs.CR↗

arXiv2022

On the Dynamics of Political Discussions on Instagram: A Network Perspective

Instagram has been increasingly used as a source of information especially among the youth. As a result, political figures now leverage the platform to spread opinions and political agenda. We here analyze online discussions on Instagram, notably in political topics, from a network perspective. Specifically, we investigate the emergence of communities of co-commenters, that is, groups of users who often interact by commenting on the same posts and may be driving the ongoing online discussions. In particular, we are interested in salient co-interactions, i.e., interactions of co-commenters that occur more often than expected by chance and under independent behavior. Unlike casual and accidental co-interactions which normally happen in large volumes, salient co-interactions are key elements driving the online discussions and, ultimately, the information dissemination. We base our study on the analysis of 10 weeks of data centered around major elections in Brazil and Italy, following both politicians and other celebrities. We extract and characterize the communities of co-commenters in terms of topological structure, properties of the discussions carried out by community members, and how some community properties, notably community membership and topics, evolve over time. We show that communities discussing political topics tend to be more engaged in the debate by writing longer comments, using more emojis, hashtags and negative words than in other subjects. Also, communities built around political discussions tend to be more dynamic, although top commenters remain active and preserve community membership over time. Moreover, we observe a great diversity in discussed topics over time: whereas some topics attract attention only momentarily, others, centered around more fundamental political discussions, remain consistently active over time.

cs.SI↗

arXiv2021

Measuring HTTP/3: Adoption and Performance

The third version of the Hypertext Transfer Protocol (HTTP) is currently in its final standardization phase by the IETF. Besides better security and increased flexibility, it promises benefits in terms of performance. HTTP/3 adopts a more efficient header compression schema and replaces TCP with QUIC, a transport protocol carried over UDP, originally proposed by Google and currently under standardization too. Although HTTP/3 early implementations already exist and some websites announce its support, it has been subject to few studies. In this work, we provide a first measurement study on HTTP/3. We testify how, during 2020, it has been adopted by some of the leading Internet companies such as Google, Facebook and Cloudflare. We run a large-scale measurement campaign toward thousands of websites adopting HTTP/3, aiming at understanding to what extent it achieves better performance than HTTP/2. We find that adopting websites often host most web page objects on third-party servers, which support only HTTP/2 or even HTTP/1.1. Our experiments show that HTTP/3 provides sizable benefits only in scenarios with high latency or very poor bandwidth. Despite the adoption of QUIC, we do not find benefits in case of high packet loss, but we observe large diversity across website providers' infrastructures.

cs.NI↗

arXiv2021

RL-IoT: Reinforcement Learning to Interact with IoT Devices

Our life is getting filled by Internet of Things (IoT) devices. These devices often rely on closed or poorly documented protocols, with unknown formats and semantics. Learning how to interact with such devices in an autonomous manner is the key for interoperability and automatic verification of their capabilities. In this paper, we propose RL-IoT, a system that explores how to automatically interact with possibly unknown IoT devices. We leverage reinforcement learning (RL) to recover the semantics of protocol messages and to take control of the device to reach a given goal, while minimizing the number of interactions. We assume to know only a database of possible IoT protocol messages, whose semantics are however unknown. RL-IoT exchanges messages with the target IoT device, learning those commands that are useful to reach the given goal. Our results show that RL-IoT is able to solve both simple and complex tasks. With properly tuned parameters, RL-IoT learns how to perform actions with the target device, a Yeelight smart bulb in our case study, completing non-trivial patterns with as few as 400 interactions. RL-IoT paves the road for automatic interactions with poorly documented IoT protocols, thus enabling interoperable systems.

cs.LG↗

arXiv2021

You, the Web and Your Device: Longitudinal Characterization of Browsing Habits

Understanding how people interact with the web is key for a variety of applications, e.g., from the design of effective web pages to the definition of successful online marketing campaigns. Browsing behavior has been traditionally represented and studied by means of clickstreams, i.e., graphs whose vertices are web pages, and edges are the paths followed by users. Obtaining large and representative data to extract clickstreams is however challenging. The evolution of the web questions whether browsing behavior is changing and, by consequence, whether properties of clickstreams are changing. This paper presents a longitudinal study of clickstreams in from 2013 to 2016. We evaluate an anonymized dataset of HTTP traces captured in a large ISP, where thousands of households are connected. We first propose a methodology to identify actual URLs requested by users from the massive set of requests automatically fired by browsers when rendering web pages. Then, we characterize web usage patterns and clickstreams, taking into account both the temporal evolution and the impact of the device used to explore the web. Our analyses precisely quantify various aspects of clickstreams and uncover interesting patterns, such as the typical short paths followed by people while navigating the web, the fast increasing trend in browsing from mobile devices and the different roles of search engines and social networks in promoting content. Finally, we contribute a dataset of anonymized clickstreams to the community to foster new studies (anonymized clickstreams are available to the public at http://bigdata.polito.it/clickstream).

cs.CY↗

arXiv2021

Towards Understanding Political Interactions on Instagram

Online Social Networks (OSNs) allow personalities and companies to communicate directly with the public, bypassing filters of traditional medias. As people rely on OSNs to stay up-to-date, the political debate has moved online too. We witness the sudden explosion of harsh political debates and the dissemination of rumours in OSNs. Identifying such behaviour requires a deep understanding on how people interact via OSNs during political debates. We present a preliminary study of interactions in a popular OSN, namely Instagram. We take Italy as a case study in the period before the 2019 European Elections. We observe the activity of top Italian Instagram profiles in different categories: politics, music, sport and show. We record their posts for more than two months, tracking "likes" and comments from users. Results suggest that profiles of politicians attract markedly different interactions than other categories. People tend to comment more, with longer comments, debating for longer time, with a large number of replies, most of which are not explicitly solicited. Moreover, comments tend to come from a small group of very active users. Finally, we witness substantial differences when comparing profiles of different parties.

cs.SI↗