arXiv subjects

Mohamed Nabeel

Publications and source records attributed to Mohamed Nabeel.

15 recordsLinked to original sources

arXiv2026

Can LLMs Reason About Brand Ownership? An Empirical Study of Domain Attribution Intelligence

When a new domain resembling a popular brand appears, defenders face a fundamental ambiguity: it may be an attacker-created squatting site for phishing, or it may be a domain the brand itself registered, either defensively, to block attackers, or legitimately, for a new product or service launch. Incorrectly flagging a brand-owned domain as malicious produces a false positive that harms end users and damages the brand's reputation. Resolving this ambiguity requires brand intelligence: the ability to determine, at scale, whether a given domain belongs to a brand. Large language models (LLMs), with their broad knowledge of brand domain relationships, offer a promising zero configuration approach to this problem, but their reliability for brand intelligence tasks remains unknown. We present the first systematic empirical evaluation of LLM brand intelligence across three tasks: domain enumeration (Q1), open ended brand attribution (Q2), and binary ownership classification (Q3). We evaluate four models, Gemini 2.5 Flash, Gemini 3.5 Flash, Claude Sonnet 4.5, and Claude Sonnet 4.6, across four retrieval settings (in context, web search, WHOIS lookup, and combined) on 36 of the most phished brands. Our results reveal a stark dichotomy: models achieve up to 82% precision enumerating brand domains from memory alone, yet fail at ownership verification without external tools, with macro F1 at most 0.37 in ICL mode. WHOIS augmentation lifts Q3 macro F1 by up to 0.65 points, yielding near perfect precision (<= 0.99), dramatically reducing the false positive risk for defenders. We provide concrete recommendations for deploying LLMs in brand protection pipelines.

cs.CR↗

arXiv2026

A Longitudinal Study of Recently Observed Malicious Domains: Characteristics, Infrastructure, and Abuse Patterns

We present a longitudinal study of approximately 1.52 million malicious domains observed on VirusTotal (VT) between January and May 2026. Domains were selected on the basis of detection by at least five independent VT scanning engines and a first-seen date within the study window. We group the dataset into compromised domains and attacker created domains, which account for approximately 89.3% of the dataset. Combining WHOIS registration records and passive DNS (PDNS) data with the VT dataset, we characterise attacker behaviour across eight dimensions: temporal distribution, compromisedvs.attack classification, domain age at first detection, registrar and TLD preferences, DNS query volume as a damage proxy, hosting infrastructure concentration (IP and ASN level), bulk registration patterns, and brand impersonation. Key findings include: the majority of attacker created domains are short lived registrations used within weeks of creation; a small number of registrars and TLDs account for most abuse; Cloudflare infrastructure is heavily exploited for domain fronting; bulk registration events involving thousands of domains from a single registrar on a single day are widespread; and several global brands, particularly WhatsApp and Google, are heavily impersonated. We share the annotated dataset in the GitHub repo https://github.com/mufimash/malicious_domains for further research.

cs.CR↗

arXiv2026

Deep Dive into the Abuse of DL APIs To Create Malicious AI Models and How to Detect Them

According to Gartner, more than 70% of organizations will have integrated AI models into their workflows by the end of 2025. In order to reduce cost and foster innovation, it is often the case that pre-trained models are fetched from model hubs like Hugging Face or TensorFlow Hub. However, this introduces a security risk where attackers can inject malicious code into the models they upload to these hubs, leading to various kinds of attacks including remote code execution (RCE), sensitive data exfiltration, and system file modification when these models are loaded or executed (predict function). Since AI models play a critical role in digital transformation, this would drastically increase the number of software supply chain attacks. While there are several efforts at detecting malware when deserializing pickle based saved models (hiding malware in model parameters), the risk of abusing DL APIs (e.g. TensorFlow APIs) is understudied. Specifically, we show how one can abuse hidden functionalities of TensorFlow APIs such as file read/write and network send/receive along with their persistence APIs to launch attacks. It is concerning to note that existing scanners in model hubs like Hugging Face and TensorFlow Hub are unable to detect some of the stealthy abuse of such APIs. This is because scanning tools only have a syntactically identified set of suspicious functionality that is being analysed. They often do not have a semantic-level understanding of the functionality utilized. After demonstrating the possible attacks, we show how one may identify potentially abusable hidden API functionalities using LLMs and build scanners to detect such abuses.

cs.CR↗

arXiv2025

Malicious GenAI Chrome Extensions: Unpacking Data Exfiltration and Malicious Behaviours

The rapid proliferation of AI and GenAI tools has extended to the Chrome Web Store. Cybercriminals are exploiting this trend, deploying malicious Chrome extensions posing as AI tools or impersonating popular GenAI models to target users. These extensions often appear legitimate while secretly exfiltrating sensitive data or redirecting users web traffic to attacker-controlled domains. To examine the impact of this trend on the browser extension ecosystem, we curated a dataset of 5,551 AI-themed extensions released over a nine-month period to the Chrome Web Store. Using a multi-signal detection methodology that combines manifest analysis, domain reputation, and runtime network behavior, supplemented with human review, we identified 154 previously undetected malicious Chrome extensions. Together with extensions known from public threat research disclosures, this resulted in a final set of 341 malicious extensions for analysis. Of these, 29 were GenAI-related, forming the focus of our in-depth analysis and disclosure. We deconstruct representative GenAI cases, including Supersonic AI, DeepSeek AI | Free AI Assistant, and Perplexity Search, to illustrate attacker techniques such as Adversary-in-the-Browser, impersonation, bait-and-switch updates, query hijacking, and redirection. Our findings show that threat actors are leveraging GenAI trends and exploiting browser extension APIs and settings for malicious purposes. This demonstrates that the browser extension threat landscape is directly evolving alongside the rapid adoption of GenAI technologies.

cs.CR↗

arXiv2025

MANTIS: Detection of Zero-Day Malicious Domains Leveraging Low Reputed Hosting Infrastructure

Internet miscreants increasingly utilize short-lived disposable domains to launch various attacks. Existing detection mechanisms are either too late to catch such malicious domains due to limited information and their short life spans or unable to catch them due to evasive techniques such as cloaking and captcha. In this work, we investigate the possibility of detecting malicious domains early in their life cycle using a content-agnostic approach. We observe that attackers often reuse or rotate hosting infrastructures to host multiple malicious domains due to increased utilization of automation and economies of scale. Thus, it gives defenders the opportunity to monitor such infrastructure to identify newly hosted malicious domains. However, such infrastructures are often shared hosting environments where benign domains are also hosted, which could result in a prohibitive number of false positives. Therefore, one needs innovative mechanisms to better distinguish malicious domains from benign ones even when they share hosting infrastructures. In this work, we build MANTIS, a highly accurate practical system that not only generates daily blocklists of malicious domains but also is able to predict malicious domains on-demand. We design a network graph based on the hosting infrastructure that is accurate and generalizable over time. Consistently, our models achieve a precision of 99.7%, a recall of 86.9% with a very low false positive rate (FPR) of 0.1% and on average detects 19K new malicious domains per day, which is over 5 times the new malicious domains flagged daily in VirusTotal. Further, MANTIS predicts malicious domains days to weeks before they appear in popular blocklists.

cs.CR↗

arXiv2022

GREENER: Graph Neural Networks for News Media Profiling

We study the problem of profiling news media on the Web with respect to their factuality of reporting and bias. This is an important but under-studied problem related to disinformation and "fake news" detection, but it addresses the issue at a coarser granularity compared to looking at an individual article or an individual claim. This is useful as it allows to profile entire media outlets in advance. Unlike previous work, which has focused primarily on text (e.g.,~on the text of the articles published by the target website, or on the textual description in their social media profiles or in Wikipedia), here our main focus is on modeling the similarity between media outlets based on the overlap of their audience. This is motivated by homophily considerations, i.e.,~the tendency of people to have connections to people with similar interests, which we extend to media, hypothesizing that similar types of media would be read by similar kinds of users. In particular, we propose GREENER (GRaph nEural nEtwork for News mEdia pRofiling), a model that builds a graph of inter-media connections based on their audience overlap, and then uses graph neural networks to represent each medium. We find that such representations are quite useful for predicting the factuality and the bias of news media outlets, yielding improvements over state-of-the-art results reported on two datasets. When augmented with conventionally used representations obtained from news articles, Twitter, YouTube, Facebook, and Wikipedia, prediction accuracy is found to improve by 2.5-27 macro-F1 points for the two tasks.

cs.LG↗

arXiv2022

Exploration of Enterprise Server Data to Assess Ease of Modeling System Behavior

Enterprise networks are one of the major targets for cyber attacks due to the vast amount of sensitive and valuable data they contain. A common approach to detecting attacks in the enterprise environment relies on modeling the behavior of users and systems to identify unexpected deviations. The feasibility of this approach crucially depends on how well attack-related events can be isolated from benign and mundane system activities. Despite the significant focus on end-user systems, the background behavior of servers running critical services for the enterprise is less studied. To guide the design of detection methods tailored for servers, in this work, we examine system event records from 46 servers in a large enterprise obtained over a duration of ten weeks. We analyze the rareness characteristics and the similarity of the provenance relations in the event log data. Our findings show that server activity, in general, is highly variant over time and dissimilar across different types of servers. However, careful consideration of profiling window of historical events and service level grouping of servers improve rareness measurements by 24.5%. Further, utilizing better contextual representations, the similarity in provenance relationships could be improved. An important implication of our findings is that detection techniques developed considering experimental setups with non-representative characteristics may perform poorly in practice.

cs.CR↗

arXiv2022

A Large Scale Study and Classification of VirusTotal Reports on Phishing and Malware URLs

VirusTotal (VT) provides aggregated threat intelligence on various entities including URLs, IP addresses, and binaries. It is widely used by researchers and practitioners to collect ground truth and evaluate the maliciousness of entities. In this work, we provide a comprehensive analysis of VT URL scanning reports containing the results of 95 scanners for 1.577 Billion URLs over two years. Individual VT scanners are known to be noisy in terms of their detection and attack type classification. To obtain high quality ground truth of URLs and actively take proper actions to mitigate different types of attacks, there are two challenges: (1) how to decide whether a given URL is malicious given noisy reports and (2) how to determine attack types (e.g., phishing or malware hosting) that the URL is involved in, given conflicting attack labels from different scanners. In this work, we provide a systematic comparative study on the behavior of VT scanners for different attack types of URLs. A common practice to decide the maliciousness is to use a cut-off threshold of scanners that report the URL as malicious. However, in this work, we show that using a fixed threshold is suboptimal, due to several reasons: (1) correlations between scanners; (2) lead/lag behavior; (3) the specialty of scanners; (4) the quality and reliability of scanners. A common practice to determine an attack type is to use majority voting. However, we show that majority voting could not accurately classify the attack type of a URL due to the bias from correlated scanners. Instead, we propose a machine learning-based approach to assign an attack type to URLs given the VT reports.

cs.CR↗

arXiv2022

PDNS-Net: A Large Heterogeneous Graph Benchmark Dataset of Network Resolutions for Graph Learning

In order to advance the state of the art in graph learning algorithms, it is necessary to construct large real-world datasets. While there are many benchmark datasets for homogeneous graphs, only a few of them are available for heterogeneous graphs. Furthermore, the latter graphs are small in size rendering them insufficient to understand how graph learning algorithms perform in terms of classification metrics and computational resource utilization. We introduce, PDNS-Net, the largest public heterogeneous graph dataset containing 447K nodes and 897K edges for the malicious domain classification task. Compared to the popular heterogeneous datasets IMDB and DBLP, PDNS-Net is 38 and 17 times bigger respectively. We provide a detailed analysis of PDNS-Net including the data collection methodology, heterogeneous graph construction, descriptive statistics and preliminary graph classification performance. The dataset is publicly available at https://github.com/qcri/PDNS-Net. Our preliminary evaluation of both popular homogeneous and heterogeneous graph neural networks on PDNS-Net reveals that further research is required to improve the performance of these models on large heterogeneous graphs.

cs.LG↗

arXiv2022

PhishChain: A Decentralized and Transparent System to Blacklist Phishing URLs

Blacklists are a widely-used Internet security mechanism to protect Internet users from financial scams, malicious web pages and other cyber attacks based on blacklisted URLs. In this demo, we introduce PhishChain, a transparent and decentralized system to blacklisting phishing URLs. At present, public/private domain blacklists, such as PhishTank, CryptoScamDB, and APWG, are maintained by a centralized authority, but operate in a crowd sourcing fashion to create a manually verified blacklist periodically. In addition to being a single point of failure, the blacklisting process utilized by such systems is not transparent. We utilize the blockchain technology to support transparency and decentralization, where no single authority is controlling the blacklist and all operations are recorded in an immutable distributed ledger. Further, we design a page rank based truth discovery algorithm to assign a phishing score to each URL based on crowd sourced assessment of URLs. As an incentive for voluntary participation, we assign skill points to each user based on their participation in URL verification.

cs.CR↗

arXiv2022

CGraph: Graph Based Extensible Predictive Domain Threat Intelligence Platform

Ability to effectively investigate indicators of compromise and associated network resources involved in cyber attacks is paramount not only to identify affected network resources but also to detect related malicious resources. Today, most of the cyber threat intelligence platforms are reactive in that they can identify attack resources only after the attack is carried out. Further, these systems have limited functionality to investigate associated network resources. In this work, we propose an extensible predictive cyber threat intelligence platform called cGraph that addresses the above limitations. cGraph is built as a graph-first system where investigators can explore network resources utilizing a graph based API. Further, cGraph provides real-time predictive capabilities based on state-of-the-art inference algorithms to predict malicious domains from network graphs with a few known malicious and benign seeds. To the best of our knowledge, cGraph is the only threat intelligence platform to do so. cGraph is extensible in that additional network resources can be added to the system transparently.

cs.CR↗

arXiv2021

Uncovering IP Address Hosting Types Behind Malicious Websites

Hundreds of thousands of malicious domains are created everyday. These malicious domains are hosted on a wide variety of network infrastructures. Traditionally, attackers utilize bullet proof hosting services (e.g. MaxiDed, Cyber Bunker) to take advantage of relatively lenient policies on what content they can host. However, these IP ranges are increasingly being blocked or the services are taken down by law enforcement. Hence, attackers are moving towards utilizing IPs from regular hosting providers while staying under the radar of these hosting providers. There are several practical advantages of accurately knowing the type of IP used to host malicious domains. If the IP is a dedicated IP (i.e. it is leased to a single entity), one may blacklist the IP to block domains hosted on those IPs as welll as use as a way to identify other malicious domains hosted the same IP. If the IP is a shared hosting IP, hosting providers may take measures to clean up such domains and maintain a high reputation for their users.

cs.CR↗

arXiv2021

Malicious and Low Credibility URLs on Twitter during the AstraZeneca COVID-19 Vaccine Development

We investigate the link sharing behavior of Twitter users following the temporary halt of AstraZeneca COVID-19 vaccine development in September 2020. During this period, we show the presence of malicious and low credibility information sources shared on Twitter messages in multiple languages. The malicious URLs, often in shortened forms, are increasingly hosted in content delivery networks and shared cloud hosting infrastructures not only to improve reach but also to avoid being detected and blocked. There are potential signs of coordination to promote both malicious and low credibility URLs on Twitter. Our findings suggest the need to develop a system that monitors the low-quality URLs shared in times of crisis.

cs.SI↗

arXiv2019

DeviceWatch: Identifying Compromised Mobile Devices through Network Traffic Analysis and Graph Inference

In this paper, we propose to identify compromised mobile devices from a network administrator's point of view. Intuitively, inadvertent users (and thus their devices) who download apps through untrustworthy markets are often allured to install malicious apps through in-app advertisement or phishing. We thus hypothesize that devices sharing a similar set of apps will have a similar probability of being compromised, resulting in the association between a device being compromised and apps in the device. Our goal is to leverage such associations to identify unknown compromised devices (i.e., devices possibly having yet currently not having known malicious apps) using the guilt-by-association principle. Admittedly, such associations could be quite weak as it is often hard, if not impossible, for an app to automatically download and install other apps without explicit initiation from a user. We describe how we can magnify such weak associations between devices and apps by carefully choosing parameters when applying graph-based inferences. We empirically show the effectiveness of our approach with a comprehensive study on the mobile network traffic provided by a major mobile service provider. Concretely, we achieve nearly 98\% accuracy in terms of AUC (area under the ROC curve). Given the relatively weak nature of association, we further conduct in-depth analysis of the different behavior of a graph-inference approach, by comparing it to active DNS data. Moreover, we validate our results by showing that detected compromised devices indeed present undesirable behavior in terms of their privacy leakage and network infrastructure accessed.

cs.CR↗

arXiv2017

Killing Two Birds with One Stone: Malicious Domain Detection with High Accuracy and Coverage

Inference based techniques are one of the major approaches to analyze DNS data and detecting malicious domains. The key idea of inference techniques is to first define associations between domains based on features extracted from DNS data. Then, an inference algorithm is deployed to infer potential malicious domains based on their direct/indirect associations with known malicious ones. The way associations are defined is key to the effectiveness of an inference technique. It is desirable to be both accurate (i.e., avoid falsely associating domains with no meaningful connections) and with good coverage (i.e., identify all associations between domains with meaningful connections). Due to the limited scope of information provided by DNS data, it becomes a challenge to design an association scheme that achieves both high accuracy and good coverage. In this paper, we propose a new association scheme to identify domains controlled by the same entity. Our key idea is an in-depth analysis of active DNS data to accurately separate public IPs from dedicated ones, which enables us to build high-quality associations between domains. Our scheme identifies many meaningful connections between domains that are discarded by existing state-of-the-art approaches. Our experimental results show that the proposed association scheme not only significantly improves the domain coverage compared to existing approaches but also achieves better detection accuracy. Existing path-based inference algorithm is specifically designed for DNS data analysis. It is effective but computationally expensive. As a solution, we investigate the effectiveness of combining our association scheme with the generic belief propagation algorithm. Through comprehensive experiments, we show that this approach offers significant efficiency and scalability improvement with only minor negative impact of detection accuracy.

cs.CR↗