# A Cross-Platform Study on IoT Malware --- ## Abstract 作者提出多階段的分析 : (事實上跟一般惡意軟體分析沒差多少) **Multi-level** analysis of IoT malware programs based on static/dynamic analysis. First >Entropy-based method to differentiate packing malware from non-packed ones Second >動態、靜態分析的 Characterizing information 做 t-SNE (視覺化) provides a visual hint on the interpretability of different features Third >SVM 做 malware 分類 --- ## Data IoTPOT: IOT的 Honey Pot ,虛擬化Linux 系統,具有一些滲透漏洞,可以記錄惡意軟體的指令和動作 9,085 samples collected 以SHA256 來 label 他們 最後限制在以三大類 Bashlite 、Mirai、 Tsunami 共 2,931 samples 作為Dataset --- [A] preprocess: packing entropy analysis => Lyda et al. Entropy => 不確定性問題所包含「不確定」(uncertainty) 的程度可以用數學來定量 [B]Feature Engineering ###### tags: `thesis`