Datasets:
Tauhait
/
JargonDetection

Dataset card Data Studio Files
xet
 Community
tokens
list
label_ids
list
[ "During", "our", "training", ",", "we", "find", "that", "SeqNet", "only", "needs", "about", "two", "and", "a", "half", "minutes", "to", "complete", "an", "epoch", "compared", "to", "MalConv", "which", "is", "about", "an", "hour", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "Accordingly", ",", "we", "can", "see", "that", "the", "tiny", "size", "of", "SeqNet", "leads", "to", "low", "computational", "overhead", ",", "and", "the", "convolution", "-", "based", "architecture", "accelerates", "the", "training", "and", "inference", "." ...
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "Adversarial", "attacks", "are", "the", "risk", "of", "most", "neural", "networks", ",", "and", "ours", "is", "no", "exception", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "A", "motivated", "adversary", "could", "pollute", "the", "training", "dataset", "and", "evade", "the", "detection", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Also", ",", "attacking", "based", "on", "the", "gradient", "is", "an", "effective", "way", "to", "confuse", "deep", "learning", "models", "~." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Although", "SeqNet", "could", "defend", "against", "several", "attacks", ",", "we", "still", "cannot", "completely", "guarantee", "the", "safety", "of", "SeqNet", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Also", ",", "the", "robustness", "principle", "of", "SeqNet", "needs", "us", "to", "explore", "further", "." ]
[ 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "In", "this", "work", ",", "we", "propose", "an", "efficient", "automatic", "malware", "detection", "neural", "network", "called", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0 ]
[ "SeqNet", "mainly", "aims", "at", "automatic", "and", "efficient", "detection", "and", "could", "be", "quickly", "trained", "with", "low", "training", "costs", "on", "raw", "binaries", "." ]
[ 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "One", "of", "the", "biggest", "obstacles", "to", "malware", "detection", "research", "based", "on", "deep", "learning", "is", "the", "lack", "of", "industrial", "-", "sized", "publicly", "available", "datasets", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "researchers", "require", "authoritative", "credible", "datasets", "that", "contain", "not", "only", "malicious", "features", "but", "also", "raw", "binary", "sequences", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "We", "will", "build", "a", "larger", "dataset", "to", "further", "evaluate", "the", "performance", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Also", ",", "enough", "benign", "samples", "are", "necessary", "for", "further", "study", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "We", "suppose", "that", "when", "using", "deep", "learning", "models", "for", "detection", ",", "malware", "analysis", "should", "not", "only", "focus", "on", "malicious", "samples", "but", "also", "on", "benign", "samples", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0 ]
[ "Because", "neural", "networks", "are", "black", "-", "box", "models", ",", "the", "reliability", "of", "malware", "detection", "neural", "networks", "might", "be", "suspicious", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "Although", "SeqNet", "gives", "us", "great", "results", ",", "we", "still", "cannot", "fully", "explain", "the", "reason", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Therefore", ",", "using", "deep", "learning", "algorithms", "to", "detect", "malware", "in", "practice", "still", "needs", "further", "research", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "Through", "our", "experiments", ",", "we", "suppose", "that", "neural", "networks", "might", "have", "great", "potential", "capacities", "for", "malware", "detection", ",", "and", "we", "are", "looking", "forward", "to", "the", "big", "breakthrough", "neural", ...
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "\\", "noindent", "Architecture", "Design", "." ]
[ 0, 0, 0, 1, 0 ]
[ "During", "our", "experiments", ",", "we", "also", "find", "that", "deeper", "structures", "may", "not", "perform", "better", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "We", "adjust", "the", "depth", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 1, 0 ]
[ "The", "model", "with", "the", "deeper", "architecture", "is", "called", "SeqNet", "Deep", ",", "and", "the", "more", "shallow", "architecture", "is", "called", "SeqNetShal", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0 ]
[ "Similar", "to", "our", "intuition", ",", "the", "shallower", "architecture", "significantly", "makes", "the", "model", "perform", "worse", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "However", ",", "the", "result", "in", "Table", "~", "shows", "that", "the", "deeper", "architecture", "cannot", "improve", "the", "performance", "apparently", "but", "enlarge", "the", "size", "of", "the", "model", "." ]
[ 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 ]
[ "We", "assume", "that", "this", "phenomenon", "is", "probably", "because", "of", "the", "simple", "mapping", "relationships", "from", "the", "binary", "to", "the", "possibility", "of", "malware", ",", "which", "does", "not", "need", "a", "complex", "neural", ...
[ 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, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "Another", "possible", "reason", "is", "that", "the", "deeper", "structure", "makes", "the", "network", "hard", "to", "be", "trained", ",", "which", "might", "lead", "to", "lower", "accuracy", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Another", "phenomenon", "we", "find", "is", "that", "the", "dilated", "convolution", "could", "not", "improve", "the", "performance", "effectively", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "We", "consider", "that", "it", "is", "because", "compared", "to", "the", "length", "of", "the", "input", ",", "using", "or", "not", "using", "the", "dilated", "convolution", "does", "not", "have", "much", "effect", "on", "the", "receptive", "field", "of",...
[ 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, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "We", "also", "check", "the", "robustness", "of", "SeqNet", "and", "compare", "it", "with", "MalConv", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0 ]
[ "There", "has", "abundant", "research", "about", "attacking", "deep", "models", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "However", ",", "different", "from", "traditional", "attacks", "against", "neural", "networks", "on", "image", "-", "related", "tasks", ",", "we", "could", "not", "straightforwardly", "add", "perturbation", "on", "binaries", "because", "it", "might", "make", "bi...
[ 0, 0, 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, 0 ]
[ "Also", ",", "it", "is", "hard", "for", "us", "to", "adapt", "attack", "strategy", "based", "on", "extracted", "features", "~", "to", "raw", "-", "binary", "models", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Therefore", ",", "for", "attack", "strategy", ",", "we", "apply", "the", "approach", "in", "~", "which", "injects", "a", "short", "poison", "binary", "in", "the", "padding", "part", "at", "the", "end", "of", "the", "input", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "For", "common", "convolution", ",", "the", "number", "of", "calculations", "is", "9999999999" ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "For", "DSC", ",", "the", "number", "of", "calculations", "is", "9999999998" ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "For", "SDSC", ",", "the", "number", "of", "calculations", "is", "9999999997" ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "Additionally", ",", "the", "input", "of", "SDSC", "is", "the", "one", "-", "dimensional", "data", ",", "so", "compared", "to", "the", "DSC", ",", "it", "is", "less", "susceptible", "to", "unrelated", "instructions", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "the", "experiments", ",", "we", "find", "that", "SDSC", "has", "great", "performance", "and", "successfully", "maintains", "the", "performance", "of", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Based", "on", "SDSC", ",", "we", "use", "the", "following", "two", "main", "architectures", "of", "convolution", "blocks", "in", "SeqNet", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0 ]
[ "Standard", "SDSC", "Block", "." ]
[ 0, 0, 1, 0 ]
[ "As", "shown", "in", "Figure", "~", ",", "the", "standard", "SDSC", "block", "has", "three", "parts", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "We", "use", "the", "batch", "normalization", "layers", "~", "to", "help", "the", "model", "learn", "the", "probability", "distribution", "of", "training", "samples", "better", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "ReLU", "~", "activation", "function", "could", "accelerate", "the", "training", "process", "by", "making", "the", "model", "converge", "quickly", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Residual", "SDSC", "Block", "." ]
[ 0, 0, 1, 0 ]
[ "As", "shown", "in", "Figure", "~", ",", "the", "residual", "SDSC", "block", "combines", "the", "method", "used", "in", "ResNet", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "By", "skipping", "the", "SDSC", "block", ",", "we", "could", "effectively", "prevent", "the", "gradient", "from", "vanishing", "and", "build", "much", "deeper", "architecture", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "\\end{itemize}", "Model", "Architecture" ]
[ 0, 0, 1 ]
[ "The", "construction", "of", "SeqNet", "is", "mainly", "based", "on", "SDSC", ",", "and", "Figure", "~", "explains", "the", "architecture", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "In", "order", "to", "reduce", "the", "number", "of", "parameters", ",", "we", "use", "smaller", "kernels", "and", "a", "deeper", "structure", ",", "which", "could", "also", "expand", "the", "receptive", "field", "." ]
[ 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 ]
[ "Because", "of", "the", "different", "input", "formats", "between", "MalConv", "and", "SeqNet", ",", "we", "equivalently", "adapt", "the", "poison", "binary", "generating", "approach", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Compared", "to", "selecting", "the", "closest", "embedding", "vector", "along", "the", "gradient", ",", "the", "poison", "generation", "process", "of", "SeqNet", "steps", "as", "the", "following", "formula", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "x_i^t,", "y^t", "denote", "an", "element", "of", "the", "normalized", "poison", "and", "the", "prediction", "given", "by", "SeqNet", "in", "the", "t", "-", "th", "iteration", ",", "respectively", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "order", "to", "make", "the", "attack", "strategy", "effective", "under", "the", "limited", "binary", "length", ",", "we", "randomly", "select", "500", "available", "samples", "from", "the", "validation", "malicious", "dataset", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "set", "the", "length", "of", "poison", "to", "32000", "bytes", ",", "which", "the", "whole", "binaries", "for", "MalConv", "are", "fixed", "to", "16000000", "bytes", ",", "and", "we", "progressively", "increase", "the", "poison", "generation", "iterat...
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "test", "the", "number", "of", "samples", "misclassified", "by", "SeqNet", "and", "MalConv", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 1, 0 ]
[ "We", "see", "that", "SeqNet", "has", "a", "great", "defensive", "capability", "against", "the", "poison", "binary", "attack", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "assume", "that", "this", "phenomenon", "is", "because", "of", "the", "vulnerability", "caused", "by", "the", "padding", "part", "in", "the", "file", "-", "cutting", "approach", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "File", "-", "cutting", "-", "based", "models", "often", "see", "incomplete", "binary", "while", "training", "." ]
[ 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0 ]
[ "Therefore", ",", "the", "padding", "in", "file", "-", "cutting", "gives", "attackers", "the", "chance", "to", "confuse", "models", ",", "while", "the", "models", "are", "not", "sure", "whether", "the", "poison", "binary", "is", "one", "of", "the", "parts"...
[ 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, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Deep", "Malware", "Detection" ]
[ 0, 0, 1 ]
[ "Neural", "networks", "have", "powerful", "learning", "abilities", "and", "have", "been", "widely", "used", "in", "computer", "vision", "and", "natural", "language", "processing", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "To", "our", "knowledge", ",", "we", "consider", "that", "there", "are", "two", "mainstream", "ideas", ",", "which", "is", "similar", "to", "~.", "\\noindent", "Feature", "-", "based", "Methods", "." ]
[ 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", "selected", "features", "could", "be", "API", "calls", ",", "control", "flow", "graphs", "~", "(", "CFG", ")", ",", "or", "any", "other", "information", "which", "is", "able", "to", "reflect", "the", "action", "of", "a", "program", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Deep", "learning", "algorithms", "have", "already", "been", "applied", "to", "malware", "detection", "by", "many", "researchers", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "We", "read", "the", "binary", "of", "a", "file", "and", "directly", "send", "it", "to", "the", "detection", "model", "without", "or", "with", "little", "preprocessing", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Therefore", ",", "time", "-", "consuming", "manual", "feature", "extraction", "probably", "limits", "the", "usage", "of", "the", "feature", "-", "based", "models", "and", "makes", "them", "difficult", "to", "combat", "the", "continuous", "evolution", "of", "m...
[ 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, 1, 0, 0, 0, 0, 0, 1, 0 ]
[ "Also", ",", "learning", "directly", "from", "the", "raw", "binaries", "might", "be", "better", "in", "preserving", "semantic", "and", "contextual", "information", "by", "mitigating", "the", "loss", "caused", "by", "manual", "feature", "engineering", "." ]
[ 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, 1, 0 ]
[ "This", "approach", "could", "more", "effectively", "avoid", "the", "need", "for", "people", "to", "analyze", "malware", "and", "better", "reduce", "the", "workload", "of", "analysts", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "The", "model", "will", "automatically", "find", "the", "suspicious", "part", "and", "recognize", "the", "binary", "as", "malicious", "or", "benign", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0 ]
[ "Nowadays", ",", "automatic", "feature", "extraction", "is", "one", "of", "the", "trends", "of", "neural", "networks", "with", "less", "human", "intervention", "and", "better", "performance", "than", "traditional", "feature", "engineering", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "For", "example", ",", "only", "using", "API", "calls", "as", "the", "feature", "would", "cause", "models", "to", "ignore", "the", "control", "flow", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "However", ",", "the", "specific", "-", "domain", "features", "could", "only", "well", "characterize", "the", "crucial", "information", "of", "samples", "from", "one", "perspective", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "In", "early", "works", ",", "deep", "learning", "models", "are", "trained", "from", "carefully", "crafted", "malware", "features", "~", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0, 0 ]
[ "When", "checking", "a", "suspicious", "sample", ",", "models", "need", "to", "extract", "the", "specific", "features", ",", "process", "them", "in", "specific", "ways", ",", "and", "then", "detect", "malicious", "codes", "to", "give", "their", "results", "....
[ 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, 0, 0, 0, 0 ]
[ "It", "is", "indeed", "that", "learning", "from", "manual", "features", "is", "an", "effective", "way", "for", "neural", "networks", "to", "recognize", "the", "main", "differences", "between", "malicious", "and", "benign", "samples", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "Depending", "on", "our", "method", ",", "we", "make", "SeqNet", "perform", "well", "on", "Portable", "Executable", "~", "(", "PE", ")", "malware", "detection", "without", "feature", "engineering", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 1, 0, 0, 0, 0 ]
[ "Based", "on", "this", "representation", "approach", ",", "we", "create", "a", "new", "convolution", "approach", ",", "called", "Sequence", "Depthwise", "Separable", "Convolution", "~", "(", "SDSC", ")", ",", "to", "further", "squeeze", "the", "scale", "of", ...
[ 0, 0, 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, 1, 0 ]
[ "Besides", ",", "we", "make", "our", "code", "and", "dataset", "public", "for", "further", "research", ",", "and", "we", "hope", "that", "deep", "learning", "algorithms", "will", "be", "applied", "to", "malware", "detection", "better", "." ]
[ 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, 1, 0, 0 ]
[ "\\item", "We", "devise", "a", "deep", "Convolutional", "Neural", "Network", "~", "(", "CNN", ")", ",", "called", "SeqNet", ",", "which", "has", "a", "much", "smaller", "size", "and", "shorter", "training", "process", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Based", "on", "the", "new", "representation", "method", "above", ",", "we", "propose", "SDSC", ",", "a", "novel", "convolution", "method", "for", "squeezing", "malware", "detection", "models", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "\\item", "We", "propose", "a", "novel", "approach", "to", "representing", "binary", "code", "while", "reducing", "semantic", "loss", "and", "avoiding", "contextual", "confusion", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Section", "~", "describes", "the", "major", "methodology", "we", "apply", "to", "SeqNet", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "We", "also", "demonstrate", "the", "effectiveness", "of", "our", "model", "shrinkage", "methods", "in", "further", "experiments", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0 ]
[ "We", "train", "SeqNet", "on", "a", "large", "PE", "dataset", "and", "find", "it", "has", "great", "performance", ",", "compared", "to", "many", "existing", "binary", "-", "based", "methods", "and", "models", "." ]
[ 0, 0, 1, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "To", "maintain", "accuracy", "when", "reducing", "the", "number", "of", "parameters", ",", "we", "propose", "a", "novel", "binary", "code", "representation", "method", "to", "reduce", "semantic", "loss", "and", "avoid", "contextual", "confusion", "." ]
[ 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 ]
[ "Then", "we", "discuss", "several", "problems", "of", "the", "popular", "binary", "representation", "methods", "further", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0 ]
[ "In", "this", "section", ",", "we", "introduce", "the", "background", "of", "malware", "detection", "using", "deep", "learning", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 1, 0 ]
[ "At", "last", ",", "we", "explain", "the", "depth", "wise", "separable", "convolution", "that", "one", "of", "our", "approaches", "is", "based", "on", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Convolution", "Methods" ]
[ 0, 1 ]
[ "Traditional", "convolution", "simulates", "animal", "vision", "and", "has", "great", "performance", "in", "computer", "vision", "." ]
[ 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Low", "-", "level", "convolutional", "layers", "detect", "the", "textural", "and", "simple", "features", "in", "images", ",", "and", "high", "-", "level", "convolutional", "layers", "could", "recognize", "the", "content", "and", "overall", "semantics", "~." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "However", ",", "the", "number", "of", "parameters", "needed", "in", "traditional", "convolution", "often", "makes", "deep", "learning", "models", "too", "large", "to", "apply", "on", "devices", "with", "low", "computational", "capacity", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "That", "is", "why", "computers", "can", "identify", "complex", "objects", "with", "the", "superposition", "of", "multiple", "convolutional", "layers", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0 ]
[ "For", "example", ",", "VGG", "has", "more", "than", "130", "million", "parameters", "and", "has", "been", "trained", "for", "2", "\\sim", "3", "weeks", "~", "." ]
[ 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "To", "squeeze", "the", "size", "of", "neural", "networks", ",", "advanced", "methods", "have", "been", "proposed", ",", "and", "depthwise", "separable", "convolution", "~(", "DSC", ")", "is", "one", "of", "them", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "As", "shown", "in", "Figure", "~", ",", "compared", "to", "standard", "convolution", ",", "DSC", "could", "effectively", "reduce", "the", "number", "of", "training", "parameters", "." ]
[ 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Recent", "works", "have", "revealed", "that", "the", "CNNs", "could", "effectively", "find", "the", "salient", "statistical", "differences", "between", "samples", "in", "different", "classes", "~." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "This", "method", "also", "makes", "MobileNets", "much", "smaller", "than", "many", "prior", "models", "and", "available", "to", "run", "on", "mobile", "devices", "." ]
[ 0, 0, 0, 0, 1, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0, 0 ]
[ "Compared", "to", "the", "self", "-", "attention", "mechanism", ",", "DSC", "usually", "needs", "fewer", "parameters", "and", "lower", "computation", "overhead", "while", "processing", "long", "binaries", "." ]
[ 0, 0, 0, 0, 0, 0, 1, 0, 1, 0, 0, 0, 0, 0, 0, 0, 1, 0, 0, 0, 1, 0 ]
[ "Also", ",", "several", "researchers", "point", "out", "that", "DSC", "and", "self", "-", "attention", "mechanism", "~", "which", "is", "widely", "used", "in", "natural", "language", "processing", "tasks", "have", "similar", "effects", "on", "generating", "fin...
[ 0, 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, 0, 0 ]