University of Twente Student Theses
StegaScanMail : Enhancing Email Security with Deep Learning Cloud-Based Image Steganography Detection
Basarabă, Radu-Cristian (2025) StegaScanMail : Enhancing Email Security with Deep Learning Cloud-Based Image Steganography Detection.
|
PDF
5MB |
| Abstract: | As security technologies and practices evolve, so do attackers’ techniques to exploit potential informatic vulnerabilities. One recent trend that is on the rise is steganogra- phy. Digital steganography studies communication in which secret messages are concealed in other unsuspicious forms, such as images, audio files, or documents. Steganalysis is a study focused on detecting the existence of secret messages inserted into digital media using steganography. When steganography is used, attacks can occur in several forms. Image steganography is the most commonly used technique because of the frequent usage of images on the Internet and a wide variety of formats and compression techniques. An attack can take the form of a hidden executable file inside an image, which starts an at- tack in the background when the image is clicked. This type of cyberattack has targeted higher governing bodies and international companies, such as energy companies in Central Asia (Kazakhstan and Mongolia), public sector entities in Southeast Asia (Japan, Viet- nam, and Indonesia), and the Azerbaijan government. The attacks were initiated through electronic mail, which was used to transmit the modified digital media and payload to vic- tims’ workstations. Existing image steganalysis tools, such as Aperi’Solve or StegoHunt, do not cover email extensively and usually require manually passing the image, increasing the associated threats, such as malware infections or data breaches. Email providers, such as Outlook or Gmail, do not specifically mention performing image steganalysis, although these services are used by most businesses worldwide. This research proposes a deep learn- ing Python-based tool named StegaScanMail, which can be deployed in the cloud as a Docker container. StegaScanMail performs steganalysis on images in received emails using a convolutional neural network (CNN) that classifies both greyscale and RGB images. Ste- gaScanMail’s functionality targets and has been trained mainly with the portable network graphics (PNG) format of images encoded with the least significant bit (LSB) technique because malicious actors frequently use this technique due to its simplicity and effectiveness in avoiding detection. The training and testing of StegaScanMail were performed mainly using the Stego-Images-Dataset, which contains 44 000 PNG images embedded with mali- cious code. The PNG image format is often found in emails because of its frequent use in logos and web graphics. Email, which is now widely adopted and is increasingly hosted in the cloud, is highly susceptible to cyberattacks. Its extensive use, particularly in business and government, makes it a prime target. This vulnerability underscores the need for en- hanced security measures, particularly for companies and administrative bodies that rely on email as the primary mode of communication. |
| Item Type: | Essay (Master) |
| Faculty: | EEMCS: Electrical Engineering, Mathematics and Computer Science |
| Subject: | 54 computer science |
| Programme: | Computer Science MSc (60300) |
| Link to this item: | https://purl.utwente.nl/essays/106241 |
| Export this item as: | BibTeX EndNote HTML Citation Reference Manager |
Show download statistics for this publication
Show download statistics for this publicationRepository Staff Only: item control page