System One 决策模型用于 Windows 恶意软件检测:S1Mal 集成方案首次评估
A First Look at System One Decision Models for Malware Detection
这篇论文把 System One 决策模型第一次用到了 Windows 恶意软件检测上。单个判定问题效果很差,但拆成 32 个原子行为问题后 AUROC 平均提升 54.61%,S1Mal 在 1.45% 误报率下能抓住 93.47% 恶意软件。做安全检测或部署分类模型的人可以看看它的读取设计,比直接让 LLM 输出置信度靠谱得多。
一篇 arXiv 论文首次评估了 System One(S1)决策模型在 Windows PE 恶意软件检测中的表现,并提出集成方案 S1Mal。S1 模型在单次前向传播中回答类型化问题并给出概率,论文在隔离虚拟机中用 Ghidra 和 capa 解析每个二进制文件,将其渲染为包含头部摘要、导入表、MITRE ATT&CK 能力和字符串的规范化文本状态。
论文设计了三种读取方式:A0 为单一判定问题的概率,A1 为 32 个原子行为问题加权得分,A2 为模型状态嵌入上的分类头。在 BODMAS 时间划分、SOREL-20M 分布外恶意样本和 14,534 个匹配良性文件上,A0 的 AUROC 仅 0.4025-0.8590,A1 使所有模型 AUROC 平均提升 54.61%,A2 将 CLM-8B 提升至 0.9943。
最终 S1Mal 结合原子答案与 CLM-8B 的 A2 分数,AUROC 达 0.9946,在 1.45% 实测误报率下检出 93.47% 的恶意软件,0.1% FPR 下检出 59.15%。相比传统特征检测器给出校准分数、LLM 给出未校准置信文本,S1Mal 同时提供校准决策和可追溯的原子答案,便于分析师核查。
A First Look at System One Decision Models for Malware Detection
Malware detection needs scores calibrated at a fixed false positive rate (FPR) and decisions an analyst can read: feature-based detectors give the first, large language models (LLMs) on decompiled code the second, through generated text of uncalibrated confidence. System One decision models answer a typed question about a text with a probability over its declared answers in one forward pass. We present the first evaluation of these models for Windows Portable Executable malware detection, analyze how they decide, and propose S1Mal, an ensemble on the typed answers and the embedding of CLM-8B. In an isolated virtual machine, Ghidra and capa decompose each binary, and the host renders the records into a canonical textual state, a digest of headers, imports, MITRE ATT&CK capabilities and strings, optionally with decompiled functions, without class shortcuts. Three readouts score each model on the state: A0, the probability of one verdict question; A1, 32 atomic behavioral questions weighed by gradient-boosted trees; A2, heads on the model's state embedding. S1Mal feeds gradient-boosted trees with the atomic answers and the A2 score of CLM-8B. We evaluate CLM-8B, ten open decision models and two LLMs read through logits on BODMAS temporal splits, SOREL-20M out-of-distribution malware and 14,534 matched benign files. On BODMAS, A0 wastes the static evidence: its area under the receiver operating characteristic curve (AUROC) spans 0.4025-0.8590 across 13 models. A1 recovers it, raising every model's AUROC by +54.61% on average; A2 lifts CLM-8B by +0.2749 to 0.9943, above a combiner on digest facts (0.9901), and reaches 0.9853-0.9948 on nine open models. S1Mal reaches 0.9946 and detects 93.47% of the malware at a realized FPR of 1.45%, and 59.15% at 0.1% FPR on the test ROC curve, against 46.59% with A2. S1Mal gives analysts calibrated decisions, each with the atomic answers behind it.
本条情报归属 AITOP「论文」核心栏目,重点聚焦北美前沿实验室与硅谷科技巨头。在跨区域产业观察坐标中,该动态直接关联「AI 开发者工具链、Coding Agent 与工程化实践」研究主线,反映出当前技术演进与商业落地的最新风向。涉及核心实体与技术议题包括:System One、CLM-8B、恶意软件检测、MITRE ATT&CK,读者可通过下方标签进入对应专题页查阅历史脉络与关联对比。
信源核验说明:本文原始线索采编自「arXiv cs.AI」(@Pier Mattia Basciano, Giovanni Farina, Salvatore Monteleone),经 AITOP 多维评分管线结构化提炼。