notesum.ai

Published at November 12

SCORE: Syntactic Code Representations for Static Script Malware Detection

cs.CR

cs.AI

cs.LG

Released Date: November 12, 2024

Authors: Ecenaz Erdemir¹, Kyuhong Park¹, Michael J. Morais¹, Vianne R. Gao¹, Marion Marschalek¹, Yi Fan¹

Aff.: ¹Amazon Web Services

Arxiv: http://arxiv.org/abs/2411.08182v1

Refer to caption

Malware Detectors	Features	Variations	F1	FPR	Precision	TPR/Recall	AUROC
ClamAV	Files	$-$	$0.28993$	$0$	$1$	$0.16954$	$-$
BitDefender	Files	$-$	$0.85590$	$0$	$1$	$0.74810$	$-$
[18] MalConv2	Byte-strings	$-$	$0.97016$	$0.00188$	$0.99813$	$0.94371$	$0.99756$
[18] MalConv2	SCORE-T	BFT	$0.97350$	$0.00180$	$0.99810$	$0.95010$	$0.99784$
[12] SAST	AST	BFT	$0.95276$	$0.00184$	$0.99791$	$0.91152$	$0.99781$
[12] GAST	AST	-	$0.69786$	$0.00192$	$0.99642$	$0.53698$	$0.99443$
[12] UAST	AST	BFT	$0.95967$	$0.00185$	$0.99813$	$0.92391$	$0.99690$
CodeBERT XGBoost	CodeBERT embed.	$-$	$0.96665$	$0.00192$	$0.99795$	$0.93725$	$0.99060$
$[$ Ours $]$ SM	SCORE-H	$-$	$0.97732$	$0.00185$	$0.99805$	$0.95744$	$0.99922$
$[$ Ours $]$ SM	SCORE-T	DFT	$0.97932$	$0.00180$	$0.99806$	$0.96128$	$0.99909$
$[$ Ours $]$ SM	SCORE-T	BFT	$\mathbf{0.98949}$	$\mathbf{0.00172}$	$\mathbf{0.99825}$	$\mathbf{0.98088}$	$\mathbf{0.99909}$
$[$ Ours $]$ GRL	SCORE-T	Label-wise	$0.98175$	$0.00185$	$0.99807$	$0.96595$	$0.99917$
$[$ Ours $]$ GRL	SCORE-T	Threat-wise	$0.98823$	$0.00185$	$0.99810$	$0.97855$	$0.99901$