Datasets:
Tauhait
/
JargonDetection

Dataset card Viewer Files Community
tokens
sequence
label_ids
sequence
[ "Using", "a", "discrete", "logarithm", "oracle" ]
[ 0, 0, 0, 0, 1 ]
[ "In", "this", "subsection", ",", "we", "investigate", "how", "being", "able", "to", "compute", "the", "discrete", "logarithm", "may", "help", "break", "the", "delay", "of", "our", "scheme", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "We", "let", "such", "an", "adversary", "have", "access", "to", "an", "oracle", "that", "can", "compute", "the", "discrete", "logarithms", "in", "\\mathbb", "Z_p^*", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "Using", "only", "the", "available", "results", ",", "we", "suggest", "the", "cautious", "approach", "of", "assuming", "a", "20", "times", "speed", "-", "up", "when", "using", "a", "GPU", "compared", "to", "mid-range", "CPUs", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0 ]
[ "Here", ",", "we", "use", "the", "standard", "circuit", "model", "where", "we", "bound", "the", "input", "on", "any", "logic", "gate", "by", "two", "bits", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "There", "is", "a", "circuit", "of", "depth", "O(\\log", "n", "+", "\\log", "k)", "and", "size", "O(nk", "+", "n^2)", "that", "computes", "\\sum_{i=1}^k", "x_i", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "There", "is", "a", "circuit", "of", "depth", "O(\\log", "n)", "and", "size", "O(n^2)", "that", "computes", "xy", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "For", "any", "\\alpha", ">", "\\frac", "1", "{\\log", "n}", ",", "there", "is", "a", "circuit", "of", "depth", "O(n\\frac", "{\\alpha+1}\\alpha)", "and", "size", "O(n^{3+", "2\\alpha}/(\\alpha", "\\log", "n))", "that", "computes", "x^y", "\\mmod", "m", ".", "9999999963" ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "All", "our", "experiments", "are", "conducted", "on", "a", "computer", "with", "a", "single", "GPU", "GeForce", "RTX", "3090", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "Building", "a", "good", "training", "dataset", "is", "crucial", "for", "evaluating", "the", "performance", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Labeled", "correctly", "and", "having", "enough", "samples", "are", "necessary", "for", "showing", "the", "learning", "ability", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "For", "the", "type", "of", "samples", ",", "we", "consider", "that", "PE", "malware", "is", "one", "of", "the", "main", "threats", "to", "electronic", "systems", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Also", ",", "there", "are", "many", "PE", "malicious", "samples", ",", "and", "it", "is", "easy", "to", "get", "enough", "PE", "samples", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Therefore", ",", "the", "following", "experiments", "are", "applied", "to", "a", "set", "of", "PE", "files", "because", "of", "their", "prevalence", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "In", "this", "work", ",", "all", "the", "malicious", "samples", "come", "from", "VirusShare", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0 ]
[ "About", "10,000", "benign", "samples", "are", "provided", "by", "QI", "-", "ANXIN", "Corporation", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "also", "collect", "many", "benign", "samples", "from", "real", "personal", "computers", "to", "simulate", "the", "real", "environment", "in", "our", "daily", "lives", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "To", "ensure", "that", "no", "virus", "is", "mixed", "in", "the", "benign", "samples", ",", "we", "detect", "all", "the", "files", "using", "VirusTotal", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "If", "no", "AV", "engine", "regards", "it", "as", "malware", "in", "the", "VirusTotal", "report", ",", "we", "consider", "it", "a", "benign", "sample", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Operating", "system", "files", "and", "malware", "often", "have", "similar", "behaviors", ",", "which", "probably", "confuse", "the", "detection", "models", "and", "even", "a", "trained", "analyst", "~", "." ]
[ 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "Therefore", ",", "in", "order", "to", "help", "SeqNet", "observe", "the", "general", "differences", "between", "malicious", "and", "benign", "programs", "and", "make", "SeqNet", "more", "robust", ",", "we", "add", "about", "10,000", "system", "files", "as", "benign", "data", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "system", "files", "are", "also", "checked", "by", "VirusTotal", "~", "to", "ensure", "they", "are", "benign", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0 ]
[ "All", "the", "files", "in", "the", "datasets", "are", "PE", "files", ",", "and", "we", "remove", "the", "duplication", "by", "comparing", "SHA256", "values", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "In", "total", ",", "we", "have", "obtained", "a", "training", "dataset", "of", "72,329", "binaries", ",", "with", "37,501", "malicious", "and", "34,828", "benign", ",", "and", "a", "validation", "dataset", "of", "24,110", "binaries", ",", "with", "12,501", "malicious", "and", "11,609", "benign", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0 ]
[ "Therefore", ",", "we", "apply", "CNN", "architecture", "to", "SeqNet", "and", "adapt", "the", "input", "form", "of", "DSC", "from", "2D", "to", "1D", ",", "which", "better", "reduces", "the", "number", "of", "training", "parameters", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "this", "section", ",", "we", "will", "introduce", "the", "details", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Secondly", ",", "we", "introduce", "the", "sequence", "characterization", "that", "could", "reduce", "semantic", "loss", "and", "avoid", "contextual", "confusion", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "The", "third", "part", "describes", "how", "SDSC", "compresses", "the", "scale", "of", "the", "model", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "During", "the", "whole", "training", "and", "detection", "processes", ",", "the", "operator", "does", "not", "need", "professional", "malware", "analysis", "knowledge", "to", "perform", "manual", "domain", "-", "specific", "feature", "engineering", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "In", "practice", ",", "we", "directly", "input", "raw", "binaries", "into", "SeqNet", ",", "and", "SeqNet", "will", "automatically", "analyze", "the", "sequences", "and", "extract", "the", "features", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Depending", "on", "this", "theory", ",", "one", "of", "our", "main", "design", "outlines", "of", "SeqNet", "is", "to", "reduce", "contextual", "confusion", "and", "semantic", "loss", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "We", "use", "the", "raw", "binary", "sequence", "as", "the", "input", "of", "SeqNet", ",", "which", "can", "avoid", "contextual", "confusion", "and", "reduce", "semantic", "loss", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "Detecting", "accurately", "is", "an", "essential", "requirement", "of", "malware", "detection", "models", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "consider", "that", "malware", "detection", "is", "different", "from", "image", "classification", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "The", "output", "of", "SeqNet", "is", "the", "malicious", "possibility", "of", "a", "suspicious", "sample", ",", "and", "whether", "the", "input", "sample", "is", "malware", "or", "not", "is", "decided", "by", "the", "possibility", "given", "by", "the", "model", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Malware", "detection", "might", "need", "more", "attention", "to", "several", "crucial", "malicious", "codes", ",", "while", "image", "classification", "might", "focus", "more", "on", "the", "whole", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "goal", "of", "SeqNet", "is", "to", "achieve", "efficient", "and", "automatic", "malware", "detection", "with", "low", "training", "costs", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "At", "last", ",", "we", "elaborate", "on", "the", "architecture", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Further", "Evaluation", "\\&", "Ablation", "Study" ]
[ 0, 0, 0, 0, 1 ]
[ "Next", ",", "we", "will", "describe", "further", "experiments", "to", "discuss", "contextual", "confusion", "avoidance", ",", "model", "shrinkage", ",", "and", "architecture", "design", "in", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0 ]
[ "We", "apply", "several", "simple", "changes", "to", "SeqNet", "to", "test", "our", "assumptions", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "\\", "noindent", "Contextual", "Confusion", "." ]
[ 0, 0, 0, 1, 0 ]
[ "To", "prove", "that", "the", "contextual", "confusion", "exists", ",", "we", "change", "the", "SeqNet", "into", "SeqNet2D", "which", "could", "input", "512\\times512", "images", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "The", "architecture", "of", "SeqNet2D", "which", "uses", "depth", "wise", "separable", "convolution", "layers", "is", "highly", "similar", "to", "SeqNet", "." ]
[ 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "Table", "~", "makes", "sure", "that", "the", "image", "converting", "method", "causes", "confusion", "that", "could", "confuse", "the", "network", "when", "learning", "features", "from", "programs", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "MobileNet", "uses", "images", "as", "the", "input", ",", "so", "Table", "~", "also", "reflects", "the", "existence", "of", "contextual", "confusion", "through", "the", "comparison", "among", "SeqNet", "and", "ResNet", "." ]
[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0 ]
[ "\\", "noindent", "Model", "Shrinkage", "." ]
[ 0, 0, 0, 1, 0 ]
[ "SDSC", "layers", "reduce", "the", "amount", "of", "the", "parameters", "by", "both", "decreasing", "input", "dimension", "and", "factorizing", "convolution", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "replace", "all", "SDSC", "layers", "with", "common", "convolutional", "layers", "in", "SeqNet", "and", "name", "the", "new", "model", "SeqNetConv", "to", "reflect", "the", "effect", "of", "factorizing", "convolution", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "To", "validate", "the", "role", "of", "dimension", "decreasing", "in", "model", "shrinkage", ",", "we", "use", "SeqNet2D", "as", "the", "experimental", "subject", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "We", "also", "apply", "common", "convolutional", "layers", "to", "SeqNet2D", "and", "call", "the", "new", "model", "SeqNet2DConv", "to", "show", "the", "reduction", "effect", "of", "the", "two", "methods", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "As", "shown", "in", "Table", "~", ",", "we", "can", "see", "that", "the", "dimension", "effectively", "decreasing", "reduces", "the", "number", "of", "parameters", ",", "and", "the", "convolution", "factorizing", "reduces", "further", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "Through", "the", "result", ",", "it", "is", "obvious", "that", "the", "SDSC", "layers", "effectively", "shrink", "the", "model", "with", "performance", "maintained", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Due", "to", "the", "structure", "and", "the", "scale", "of", "the", "models", ",", "retraining", "might", "be", "time", "-", "and", "-", "computation", "consuming", ",", "and", "such", "a", "high", "cost", "might", "make", "model", "updating", "difficult", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Furthermore", ",", "malware", "detection", "is", "a", "usual", "operation", "in", "almost", "every", "electronic", "system", ",", "and", "it", "is", "necessary", "for", "devices", "with", "a", "low", "computational", "capacity", "to", "perform", "detection", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "This", "problem", "probably", "causes", "lower", "malware", "detection", "accuracy", "in", "small", "deep", "learning", "models", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0 ]
[ "Without", "artificial", "feature", "selection", ",", "SeqNet", "could", "automatically", "analyze", "samples", "and", "find", "the", "differences", "between", "malicious", "and", "benign", "programs", "only", "based", "on", "raw", "binaries", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "It", "might", "be", "because", "smaller", "models", "are", "often", "more", "challenging", "to", "fit", "the", "complex", "mapping", "from", "the", "raw", "binaries", "to", "the", "malicious", "possibility", "domain", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "The", "smaller", "neural", "networks", "usually", "have", "fewer", "parameters", ",", "which", "might", "lead", "to", "lower", "learning", "capacity", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "this", "paper", ",", "we", "propose", "an", "efficient", "automatic", "malware", "detection", "model", "with", "only", "about", "136", "K", "parameters", "and", "refer", "to", "it", "as", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "The", "detection", "model", "should", "be", "automatic", "enough", ",", "and", "it", "requires", "little", "manual", "feature", "engineering", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "These", "requirements", "illustrate", "that", "the", "detection", "model", "should", "be", "small", "and", "efficient", "enough", "to", "make", "it", "more", "practical", "so", "that", "we", "could", "quickly", "retrain", "or", "perform", "detection", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "For", "example", ",", "it", "is", "difficult", "for", "laptops", "or", "other", "mobile", "devices", "to", "run", "a", "huge", "model", "to", "scan", "all", "the", "files", "for", "malware", "detection", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "The", "new", "convolution", "method", ",", "called", "Sequence", "Depthwise", "Separable", "Convolution", "~", "(", "SDSC", ")", ",", "helps", "SeqNet", "meet", "this", "requirement", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "The", "input", "of", "SeqNet", "is", "the", "raw", "binary", "sequences", "which", "are", "resized", "to", "the", "same", "length", "by", "the", "linear", "interpolation", "algorithm", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Also", ",", "by", "using", "the", "linear", "interpolation", "algorithm", ",", "we", "can", "ensure", "the", "length", "of", "input", "sequences", "is", "the", "same", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Resampling", "Noise", "Reduction", "." ]
[ 0, 0, 1, 0 ]
[ "The", "disappearance", "of", "interruptions", "also", "effectively", "protects", "the", "semantics", "of", "the", "raw", "binary", "sequence", "because", "the", "instructions", "in", "close", "proximity", "are", "not", "separated", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "edge", "loss", "could", "be", "avoided", "because", "the", "sequence", "only", "has", "two", "edges", ",", "the", "beginning", "and", "the", "end", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "\\noindent", "Edge", "Loss", "Avoidance", "." ]
[ 0, 0, 0, 1, 0 ]
[ "Sequence", "Characterization" ]
[ 0, 1 ]
[ "The", "input", "format", "is", "critical", "for", "neural", "network", "performance", "and", "model", "size", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "Therefore", ",", "squeezing", "the", "scale", "of", "SeqNet", "is", "necessary", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "Larger", "input", "often", "leads", "to", "larger", "models", ",", "and", "proper", "input", "format", "could", "effectively", "improve", "the", "learning", "effect", "of", "neural", "networks", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Raw", "binary", "sequence", "input", "needs", "little", "human", "intervention", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "Without", "converting", "to", "images", ",", "it", "is", "obvious", "that", "we", "could", "avoid", "contextual", "information", "confusion", "and", "reduce", "semantic", "loss", ",", "as", "shown", "in", "Figure", "~." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "Lightweight", "models", "often", "have", "a", "wider", "range", "of", "application", "scenarios", "and", "faster", "detection", "performance", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "sequence", "format", "conforms", "to", "the", "spatial", "structure", "of", "codes", ",", "so", "there", "is", "no", "break", "in", "any", "instruction", ",", "which", "makes", "all", "instructions", "intact", "when", "inputting", "into", "the", "model", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "To", "better", "understand", "what", "SeqNet", "has", "learned", ",", "we", "randomly", "select", "four", "samples", "and", "use", "the", "Grad", "-", "CAM", "~", "explanation", "technique", "to", "generate", "heatmaps", "so", "that", "we", "could", "visualize", "which", "part", "affects", "the", "result", "most", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Also", ",", "we", "manually", "analyze", "the", "corresponding", "samples", "to", "verify", "whether", "SeqNet", "finds", "the", "right", "malicious", "codes", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "In", "manual", "analysis", ",", "we", "disassemble", "the", "samples", "by", "IDA", "Pro", "~", "and", "precisely", "locate", "the", "malicious", "functions", "or", "codes", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "In", "order", "to", "better", "plot", "the", "results", ",", "we", "extract", "the", "key", "part", "of", "the", "heatmaps", "and", "apply", "the", "following", "normalization", "formula", "to", "the", "snippets", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "The", "activation", "maps", "used", "for", "heatmaps", "are", "generated", "by", "the", "last", "convolution", "and", "ReLU", "layer", "of", "SeqNet", ",", "because", "of", "the", "remained", "spatial", "information", "encoded", "by", "convolutional", "layers", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "also", "mark", "the", "manual", "location", "results", "on", "raw", "binaries", "for", "better", "comparison", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "Figure", "~", "implies", "the", "comparison", "between", "manual", "location", "and", "Grad", "-", "CAM", "-", "based", "explanation", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "see", "that", "the", "local", "activated", "positions", "are", "closed", "to", "the", "malicious", "parts", "located", "by", "analysts", ",", "which", "reflects", "that", "SeqNet", "might", "find", "the", "malicious", "codes", "and", "give", "its", "reliable", "detection", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "However", ",", "it", "is", "a", "pity", "that", "we", "find", "it", "hard", "to", "collect", "more", "benign", "samples", "due", "to", "the", "neglect", "of", "benign", "files", "in", "the", "academic", "community", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "We", "choose", "several", "state", "-", "of", "-", "the", "-", "art", "binary", "-", "based", "methods", "as", "our", "baselines", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "To", "reflect", "the", "total", "performance", "of", "image", "-", "converting", "-", "based", "models", ",", "we", "choose", "the", "well", "-", "known", "MobileNet", "~", "as", "the", "representative", "model", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "We", "convert", "programs", "to", "RGB", "images", "as", "the", "input", "of", "MobileNet", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "One", "byte", "maps", "one", "pixel", "in", "one", "channel", "during", "the", "converting", "process", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "For", "file", "-", "cutting", "-", "based", "models", ",", "we", "select", "MalConv", "~", "and", "MalConvGCT", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0 ]
[ "Both", "ResNet", "and", "MobileNet", "are", "implemented", "by", "Pytorch", "~", ",", "and", "we", "use", "ResNet18", "as", "ResNet", "while", "MobileNetV2", "as", "MobileNet", "." ]
[ 0, 1, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 1, 0, 1, 0, 1, 0 ]
[ "We", "use", "the", "source", "codes", "provided", "by", "the", "author", "of", "MalConv", "and", "MalConvGCT", "and", "apply", "them", "to", "our", "experiments", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "We", "add", "a", "softmax", "layer", "at", "the", "end", "of", "each", "model", "to", "translate", "the", "result", "into", "the", "possibility", "format", "when", "validating", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "According", "to", "the", "number", "of", "parameters", "in", "Table", "~", ",", "we", "could", "find", "that", "SeqNet", "has", "the", "minimum", "parameters", "which", "are", "only", "about", "one", "-", "tenth", "of", "MalConv", "and", "MalConvGCT", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0 ]
[ "What", "is", "more", ",", "SeqNet", "maintains", "its", "performance", "in", "malware", "detection", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0 ]
[ "The", "precision", "also", "shows", "that", "SeqNet", "has", "a", "low", "possibility", "of", "misunderstanding", "benign", "samples", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "recall", "implies", "that", "SeqNet", "might", "have", "the", "ability", "to", "prevent", "evasion", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "the", "training", "process", ",", "We", "find", "that", "most", "models", "achieve", "90", "\\", "%", "accuracy", "after", "the", "first", "epoch", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]