From Attachments to SEO: Click Here to Learn More about Clickbait PDFs!

Giada Stivala, Sahar Abdelnabi, Andrea Mengascini, Mariano Graziano, Mario Fritz, Giancarlo Pellegrino

TL;DR

Clickbait PDFs, which trick users into visiting malicious websites instead of embedding malware, are a growing threat that has been largely overlooked. This paper provides the first comprehensive study of clickbait PDFs, revealing that they are increasingly distributed via Search Engine Optimization (SEO) attacks, moving beyond traditional email phishing.

What Problem Does This Paper Address?

Traditionally, PDF files in the context of cyber threats were primarily associated with embedding malware or exploits within email attachments (MalPDFs). However, recent trends indicate a surge in "clickbait PDFs" – documents that don't contain malicious code themselves but instead feature enticing visuals or text designed to trick users into clicking links that lead to malicious websites. These websites can then be used for credential theft, drive-by downloads, or other attacks. Despite this emerging threat, prior research has largely neglected clickbait PDFs, focusing instead on MalPDFs or traditional email phishing. This paper aims to fill this knowledge gap by systematically investigating the clickbait PDF landscape, understanding their characteristics, and identifying their distribution vectors.

Methods

The researchers employed a multi-stage methodology:

  • Dataset Creation: They collected a large dataset of 176,208 PDF files from two industrial partners.
  • Clustering: To manage the large dataset and identify visually similar documents, they utilized the DeepCluster framework, which combines Convolutional Neural Networks (CNNs) with k-means clustering. This approach leverages visual features from the first page of PDFs, as this is the most likely area to contain a bait.
  • URL Extraction and Analysis: Clickable "bait URLs" from the first page of PDFs were extracted. These URLs were then analyzed using VirusTotal and manual inspection to determine if they led to malicious activity.
  • Distribution Vector Analysis: They investigated two primary distribution methods:
  • Email Attachments: By checking against corporate spam traps and analyzing VirusTotal tags.
  • SEO Attacks: By searching for clickbait PDFs via popular search engines (Google and Bing) using specific keywords and analyzing the structure of the PDFs to identify SEO-related techniques like keyword stuffing and cross-linking.
  • Characterization: Identified clickbait PDF clusters were characterized based on their volumetric (size), temporal (activity over time), visual baits, and language distribution.

Datasets

  • Primary Dataset: A newly created dataset of 176,208 clickbait PDF files, collected from December 16th, 2020, to June 23rd, 2021, from two industrial partners (Cisco and InQuest Labs). This dataset is unique in its focus on clickbait PDFs rather than solely MalPDFs.
  • External Data: VirusTotal (for URL analysis and file metadata), Google Vision API (for language detection), and search engine APIs (Google and Bing) were used.

Key Insights and Ideas

  • Beyond Email Attachments: Clickbait PDFs are not solely distributed via email attachments. A significant portion, particularly the largest clusters, are being distributed through Search Engine Optimization (SEO) attacks. This represents a major shift in attack vectors.
  • SEO as a Primary Vector: The three largest clickbait PDF clusters, accounting for 89% of the dataset, were found to be distributed via SEO attacks. Attackers leverage keyword stuffing and cross-linking to increase the visibility of these PDFs in search engine results.
  • Visual Deception is Key: Clickbait PDFs rely heavily on visual elements to lure victims. Many mimic familiar web UI elements like CAPTCHA challenges or search engine result pages, exploiting users' familiarity with these interfaces.
  • VirusTotal's Limited Effectiveness: Standard AV scoring systems like VirusTotal often rank clickbait PDFs low, creating a blind spot for detection. Their effectiveness is further hampered by the dynamic nature of web attacks and the fact that many URLs are not flagged as malicious until after the attack has ended.
  • Persistence of Clickbait PDFs: Unlike short-lived email phishing campaigns, some clickbait PDF clusters exhibit long-lasting activity, persisting for months.

Key Results

  • 44 Malicious Clusters Identified: Out of 80 document categories, 44 clusters were identified as leading to malicious web attacks.
  • Dominance of Large Clusters: The top three clickbait PDF clusters constitute 89% of the entire dataset, highlighting a concentration of activity.
  • SEO Attack Confirmation: Experiments confirmed that clickbait PDFs are actively distributed via SEO attacks, with thousands of matches found through search engine queries.
  • Limited URL Blocklist Coverage: URL blocklists show limited effectiveness, with only 18% of malicious URLs being blocklisted by Google SafeBrowsing.
  • Multi-regional Targeting: Most large clickbait PDF clusters are multi-regional, with English being the most common language, but Chinese, a dominant internet language, is notably absent.

Limitations and Future Work

  • Dataset Bias: The study acknowledges potential biases due to the sampling policies of data providers (e.g., Cisco's cap on daily samples).
  • URL Analysis Challenges: The dynamic nature of web attacks and the limitations of URL analysis services make it challenging to definitively label all URLs as malicious.
  • In-browser Defenses: While ad-blockers show some effectiveness, they fail to protect against all types of attacks delivered by clickbait PDFs.
  • Future Work: The authors suggest further research into the role and limitations of anti-phishing ecosystem players, the development of more effective detection mechanisms that consider structural features and cross-linked PDF patterns, and a deeper understanding of the geographical reach and evolution of these attacks. They also plan to release their dataset metadata to foster further research.

Summary updated

Abstract

Clickbait PDFs are PDF documents that do not embed malware but trick victims into visiting malicious web pages leading to attacks like password theft or drive-by download. While recent reports indicate a surge of clickbait PDFs, prior works have largely neglected this new threat, considering PDFs only as accessories of email phishing campaigns.This paper investigates the landscape of clickbait PDFs and presents the first systematic and comprehensive study of this phenomenon. Starting from a real-world dataset, we identify 44 clickbait PDF clusters via clustering and characterize them by looking at their volumetric, temporal, and visual features. Among these, we identify three large clusters covering 89% of the dataset, exhibiting significantly different volumetric and temporal properties compared to classical email phishing, and relying on web UI elements as visual baits. Finally, we look at the distribution vectors and show that clickbait PDFs are not only distributed via attachments but also via Search Engine Optimization attacks, placing clickbait PDFs outside the email distribution ecosystem.Clickbait PDFs seem to be a lurking threat, not subjected to any form of content-based filtering or detection: AV scoring systems, like VirusTotal, rank them considerably low, creating a blind spot for organizations. While URL blocklists can help to prevent victims from visiting the attack web pages, we observe that they have a limited coverage.
BibTeX
@inproceedings{stivala2023from,
  author = {Giada Stivala and Sahar Abdelnabi and Andrea Mengascini and Mariano Graziano and Mario Fritz and Giancarlo Pellegrino},
  title = {{From Attachments to SEO: Click Here to Learn More about Clickbait PDFs!}},
  booktitle = {ACSAC},
  year = {2023},
  url = {https://openreview.net/forum?id=wvFaUYzFa7},
}
Figure