Using deep learning to detect digitally encoded DNA trigger for Trojan malware in Bio-Cyber attacks

Area of Science:

• Bioinformatics
• Cybersecurity
• Molecular Biology

Background:

• DNA sequencing pipelines are vulnerable to hybrid bio-cyber attacks.
• Attackers encode malicious payloads in synthetic DNA to hijack sequencing resources.
• Existing methods struggle to detect sophisticated encoding techniques like fragmentation and steganography.

Purpose of the Study:

• To develop and validate a Deep Learning approach for detecting Trojan malware attacks in DNA sequencing.
• To enhance the security of DNA sequencing pipelines against novel bio-cyber threats.
• To demonstrate the effectiveness of Deep Learning in identifying hidden malicious data within DNA sequences.

Main Methods:

• Utilized Deep Learning algorithms to analyze DNA sequences for malicious patterns.
• Simulated hybrid attack scenarios involving encoded IP addresses and port numbers within DNA.
• Applied fragmentation encryption and steganography to obfuscate malicious data.
• Validated the feasibility of DNA payload synthesis through wet lab experiments.

Main Results:

• Achieved nearly 100% accuracy in detecting encoded Trojan attack data.
• Demonstrated successful identification of malicious DNA sequences even after advanced obfuscation techniques.
• Confirmed the practical viability of creating DNA-based Trojan payloads.

Conclusions:

• Deep Learning offers a robust defense against advanced bio-cyber attacks in DNA sequencing.
• The proposed method effectively safeguards sequencing pipelines from sophisticated hybrid threats.
• This research validates the potential for securing genomic data and infrastructure.