Enabling Trust in Deep Learning Models: A Digital Forensics Case Study

Aditya, K; Grzonkowski, Slawomir; Le-Khac, Nhien-An

doi:10.1109/trustcom/bigdatase.2018.00172

Cited by 8 publications

(6 citation statements)

References 6 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…predictions? These question needs to be answered while designing the application [31]. Hence, there is need of framework which test the security and robustness of these forensic tools.…”

Section: Discussion and Future Workmentioning

confidence: 99%

See 1 more Smart Citation

Review on effectiveness of deep learning approach in digital forensics

Ekhande

Patil

Kulhalli

2022

IJECE

View full text Add to dashboard Cite

Cyber forensics is use of scientific methods for definite description of cybercrime activities. It deals with collecting, processing and interpreting digital evidence for cybercrime analysis. Cyber forensic analysis plays very important role in criminal investigations. Although lot of research has been done in cyber forensics, it is still expected to face new challenges in near future. Analysis of digital media specifically photographic images, audio and video recordings are very crucial in forensics This paper specifically focus on digital forensics. There are several methods for digital forensic analysis. Currently deep learning (DL), mainly convolutional neural network (CNN) has proved very promising in classification of digital images and sound analysis techniques. This paper presents a compendious study of recent research and methods in forensic areas based on CNN, with a view to guide the researchers working in this area. We first, defined and explained preliminary models of DL. In the next section, out of several DL models we have focused on CNN and its usage in areas of digital forensic. Finally, conclusion and future work are discussed. The review shows that CNN has proved good in most of the forensic domains and still promise to be better.

show abstract

“…predictions? These question needs to be answered while designing the application [31]. Hence, there is need of framework which test the security and robustness of these forensic tools.…”

Section: Discussion and Future Workmentioning

confidence: 99%

“…It tries to find history of the digital content by recognizing acquisition device that is used to produce the data. It also performs validation of integrity and captured information from multimedia signals [31]. This sub-section concentrates on audio and video forensic.…”

Section: Multimedia Forensicsmentioning

confidence: 99%

Review on effectiveness of deep learning approach in digital forensics

Ekhande

Patil

Kulhalli

2022

IJECE

View full text Add to dashboard Cite

show abstract

“…By employing this technique, it can aid in both anti-crime efforts and swift responses when criminal acts or behaviors are detected. In future endeavors, we aim to establish stronger connections between newly discovered evidence items and existing ones, further strengthening the investigative process [24], [25].…”

Section: Deep Learningmentioning

confidence: 99%

Enhancements in the world of digital forensics

Vaddi,

Kamble,

Vaingankar

et al. 2024

IJ-AI

View full text Add to dashboard Cite

Currently, the rapid advancement of computer systems and mobile phones has resulted in their utilization in unlawful acts. Ensuring adequate and effective security measures poses a difficult task due to the intricate nature of these devices, thereby exacerbating the challenges associated with investigating crimes involving them. Digital forensics, which involves investigating cyber crimes, plays a crucial role in this realm. Extensive research has been conducted in this field to aid forensic investigations in addressing contemporary obstacles. This paper aims to explore the progress made in the applications of digital forensics and security, encompassing various aspects, and provide insights into the evolution of digital forensics over the past five years.

show abstract

“…The model was made purposely suboptimal (with regular ROC curve far from upper-left corner) in order to better observe the manifestations of the theory. It contains a deep convolutional neural network with {embedding, 4 convolutional, 3 dense, softmax} layers summarized in Appendix A4, trained on half a million raw executable files (originally aggregated from a mix of clean and malicious customer submissions and vendor feeds), XORed with a common byte for inoculation at rest, One of the simplest adversarial attacks for binaries to circumvent malware protection is to append a crafted payload at the end of the file (Kuppa, Grzonkowski, & LeKhac, 2018). These methods can append a binary string, backdoor legitimate files by adding a new section to the executable (either as data or code), or use the resource part of the file when modifying already compiled code.…”

Section: Malware Detection From Raw Bytesmentioning

confidence: 99%

Deep Detector Health Management under Adversarial Campaigns

Echauz

Kenemer

Hussein

et al. 2019

IJPHM

View full text Add to dashboard Cite

Machine learning models are vulnerable to adversarial inputs that induce seemingly unjustifiable errors. As automated classifiers are increasingly used in industrial control systems and machinery, these adversarial errors could grow to be a serious problem. Despite numerous studies over the past few years, the field of adversarial ML is still considered alchemy, with no practical unbroken defenses demonstrated to date, leaving PHM practitioners with few meaningful ways of addressing the problem. We introduce turbidity detection as a practical superset of the adversarial input detection problem, coping with adversarial campaigns rather than statistically invisible one-offs. This perspective is coupled with ROCtheoreticdesign guidance that prescribes an inexpensive domain adaptation layer at the output of a deep learning model during an attack campaign. The result aims to approximate the Bayes optimal mitigation that ameliorates the detection model’s degraded health. A proactively reactive type of prognostics is achieved via Monte Carlo simulation of various adversarial campaign scenarios, by sampling from the model’s own turbidity distribution to quickly deploy the correct mitigation during a real-world campaign.

show abstract

Enabling Trust in Deep Learning Models: A Digital Forensics Case Study

Cited by 8 publications

References 6 publications

Review on effectiveness of deep learning approach in digital forensics

Review on effectiveness of deep learning approach in digital forensics

Enhancements in the world of digital forensics

Deep Detector Health Management under Adversarial Campaigns

Contact Info

Product

Resources

About