Steganography offers techniques to send a secure message without revealing its presence, sometimes called “hiding in plain sight.” With the increased use of mobile phones in many forensic investigations, detection of steganography, or steganalysis, has become an area of interest in forensic science. Currently, there is no dataset established to provide a standard, authenticated image dataset for benchmarking steganalysis tools by forensic scientists.
Statistics, Computer Science, Mathematics, Computer Engineering, Big Data
- Collecting thousands of stego images on the computer using the computer-equivalent of mobile phone stego apps, such as PixelKnot, and using data to run novel steganalysis experiments on.
- Creating mobile camera apps to collect images and populate an image database that allows the academic and forensic community to evaluate existing steg detection software, including commercial and in-house lab software. The CSAFE database is novel as all images originate from mobile phones, with many images in RAW format and a variety of ISO and exposure time settings.
Benefits of Research
CSAFE researchers, through innovative data acquisition that includes authoring their own camera apps, are using camera data to answer such questions as:
- Can machine learning/artificial intelligence be used to detect stego images taken from the same model of cell phone, but different individual devices than used in the training? (appeared in Electronic Imaging, Media Watermarking, Security, and Forensics, 2018)
- Can stego images produced using mobile apps be reliably detected? (submitted to DFRWS 2018 as journal paper)
- Can the effect of sensor noise be quantized to provide more insight into reliable steganalysis and digital camera identification? (submitted to ICIP 2018)
Answers to these questions can help create standardized steg detection software for forensic practitioners and companies interested in producing commercial software.
Select Publications, Conference Papers, Presentations and/or Tools
A StegoDB dataset will be publically available soon and housed on CSAFE datasets and tools.
Newman, L. Lin, W. Chen, S. Reinders, Y. Wang, M. Wu, Y. Guan. “StegoAppDB: A steganography apps forensics image database,” IS&T Int’l. Symp. on Electronic Imaging, Media Watermarking, Security, and Forensics 2019, Burlingame, CA, 2019.
Reinders, J. Newman, L. Lin, M. Wu, Y. Guan. “Algorithm mismatch in spatial steganalysis,” IS&T Int’l. Symp. on Electronic Imaging, Media Watermarking, Security, and Forensics 2019, Burlingame, CA, 2019.
Lin, W. Chen, Y. Wang, S. Reinders, M. Wu, Y. Guan, J. Newman. “The Impact of Exposure Settings in Digital Image Forensics,”2018 25th IEEE International Conference on Image Processing (ICIP), Athens, Greece, pp. 540-544, 2018.
Lin, S. Reinders, Y. Guan, M. Wu, J. Newman. “Domain Adaptation in Steganalysis for the Spatial Domain,” IS&T Int’l. Symp. on Electronic Imaging (EI), Burlingame, CA, v. 7, pp.1-9, 2018.
Chen, L. Lin, M. Wu, Y. Guan, and J. Newman. “Tackling Android Stego Apps in the Wild,” 2018 Asia-Pacific Signal and Information Processing Association Annual Summit and Conference (APSIPA ASC), Honolulu, HI, 2018.
Chen, Y. Wang, Y. Guan, J. Newman, L. Lin, and S. Reinders. “Forensic analysis of android steganography apps,” In G. Peterson and S. Shenoi, eds., Advances in Digital Forensics XIV, Cham. Springer Int’l. Publishing, pp. 293-312, 2018.
Presentation, “Can stego images from a mobile phone stego app be detected?” NIST/FBI second International Symposium on Forensic Science Error Management, Gaithersburg, MD, July 24-27, 2017.
Invited presentation, “Criticality of data for more practical steganography detection in images,” Workshop on Role of Data, Databases and Expert Knowledge in Forensic Inference, at the Department of Statistics and Data Science, Carnegie Mellon University, September 10-11, 2017.