Intelligent Feature Engineering for Cybersecurity

Using Side Channel Information and Artificial Intelligence for Malware Detection

“…Research in this area also derives many of its techniques from the image processing realm which is not always effective. As Maxwell et al point out [6], there is more work to be done to determine the best pre-processing techniques for network data.…”

“…Further limiting the usefulness of AI-based approaches is the increasing application of adversarial AI. These techniques can cause the malfunction of AI systems through data poisoning, creation of altered inputs, and more [4]- [6]. Systems that rely upon data that can be altered by an adversary are subject to reduced effectiveness and reliability.…”

Algorithm selection framework for cyber attack detection

“…Maxwell et al [12] and Viegas [24] describe IDS tools that incorporate ML models. Unfortunately, raw network traffic data is not a suitable input for building accurate and efficient ML models.…”

“…These emerging systems are susceptible to attack by malicious actors wishing to degrade the system or steal proprietary information [7]. Intrusion detection systems (IDS) are central to maintaining the security of modern computer networks from malicious actors [12]. IDS have been successfully demonstrated in both the wired and wireless domain of IoT [10].…”

“…Numerous studies [10] have explored meta-models to detect malicious behavior in computer networks. Maxwell et al [12] further focused on intelligent cybersecurity feature engineering for various meta-models.…”