ARCHIVES

E-mail remains a primary mode of digital communication, and attachments play a crucial role in transmitting documents, reports, and media files. However, these attachments often serve as an entry point for cyberattacks, where malicious PDFs, Office documents, or scripts exploit hidden vulnerabilities the moment a user opens them. Traditional defenses depend heavily on signaturebased malware scanners and user awareness training, which are increasingly inadequate against zero-day exploits, polymorphic malware, and sophisticated spear-phishing attachments. As attackers continue to embed harmful payloads within seemingly legitimate documents, there is a pressing need for a more intelligent and proactive defense mechanism. To address this challenge, this project introduces Opaline Attachment, an advanced attachment-screening system powered by the DistilBERT deep learning model. DistilBERT analyzes both structural features and embedded textual semantics from incoming email attachments to detect suspicious patterns that may indicate hidden threats. Instead of relying solely on detection, the system incorporates an additional protective layer: potentially harmful attachments are isolated in a secured sandbox and transformed into safe, static PNG renderings. This rendering process removes executable components and embedded scripts, allowing users to preview the content without interacting with the original file. The system automatically substitutes the user’s attachment view with its sanitized PNG version and provides cautionary alerts before granting access to the original file. By combining intelligent threat classification, secure content rendering, and controlled access, Opaline Attachment ensures robust protection against both known and unknown attachment-based attacks. This approach significantly reduces user exposure to harmful files and strengthens overall email security without disrupting normal workflow.