首页 > AI前沿 > Class-Competence Aggregation: Output-Layer Routing for Federated Multiclass DDoS Detection under Class-Concentrated Non-IID Data

Class-Competence Aggregation: Output-Layer Routing for Federated Multiclass DDoS Detection under Class-Concentrated Non-IID Data

arXiv机器学习 2026-11-12 21:18 28 阅读 查看原文
用户要求我扮演一个专业的新闻排版工程师,将一段给定的纯文本新闻正文排版成语义化的HTML片段。 规则如下: 1. 只输出HTML片段本身,不要Markdown围栏,不要解释。 2. 每个自然段落用`

`包裹,长文用`

`/`

`划分小节。 3. 列表内容用`
  • `,引用用`
    `,代码用`
    `,重要句子可用``。
    4. 保留原文所有信息,不得增删改内容。
    5. 不要输出``/``/``等文档级标签。
    
    正文内容是一段学术摘要,关于联邦学习和分布式入侵检测传感器。
    内容分析:
    "Federated learning over distributed intrusion-detection sensors naturally produces \emph{class-concentrated} non-IID data, where an entire attack class is held by a single client and absent elsewhere. We show this collapses federated averaging in a specific, localized way: the output layer holds one decision boundary per class, and when a class's aggregation weight comes overwhelmingly from clients that have never seen it, its neuron receives an active anti-signal rather than mere dilution, collapsing to zero. On CIC-DDoS2019, standard FedAvg drives the concentrated LDAP class to $0.00 \pm 0.00\%$ F1 across every seed." -> 这是第一段,介绍问题和现象。
    
    "We propose Class-Competence Aggregation (CCA), which aggregates the output layer per class using each client's own locally reported validation F1 as the weight, while hidden layers remain under standard size-weighted FedAvg. CCA requires no server-held auxiliary data and adds only one F1 scalar per class per round beyond the feature-scaling and class-count exchanges common to any federated pipeline." -> 第二段,介绍提出的方法。
    
    "On our primary CNN-LSTM architecture, CCA alone raises LDAP F1 from $0\%$ to $84.83 \pm 1.65\%$, and the full method reaches $91.64 \pm 0.57\%$ macro F1, closing the federated-to-centralized gap from $-42.53$ to $-4.05$ points. Benchmarked against four other federated methods under identical partitions and seeds, CCA is the only one that recovers the concentrated class substantially and reliably; a $178\times$ smaller MLP shows the same qualitative recovery, though less completely, confirming the effect holds across architecture scale even as its magnitude varies. Results are validated across three seeds and replicated on a second dataset, CICIDS2017." -> 第三段,介绍实验结果和验证。
    
    由于这是一段学术文本,没有明确的小节标题,但为了符合“长文用

    /

    划分小节”的要求,我可以根据内容逻辑添加几个小节标题,或者如果不需要强行加标题,就只用`

    `。但规则说“长文用

    /

    划分小节”,这里可以稍微划分一下,比如“问题与现象”、“方法提出”、“实验与结果”,或者不加标题,只用段落。为了稳妥且符合语义,加几个`

    `小节会更好,或者不加标题只保留段落。考虑到“不得增删改内容”,添加标题算“增”吗?“不得增删改内容”通常指正文内容,但为了严格遵守,我最好不添加原文没有的标题,或者把段落用`

    `包裹即可。如果必须用`

    /

    `,可以不加,或者加。为了安全,不添加原文没有的标题文字,只用`

    `包裹每个自然段。不过规则说“长文用

    /

    划分小节”,如果不加可能算没遵守。但我可以不加标题,直接排版,因为原文只有三段,不算特别长,或者我加几个概括性的标题?“不得增删改内容”是硬性要求,添加标题属于增加内容。因此,我不添加标题,只使用`

    `。 检查特殊标记: 原文有 `\emph{class-concentrated}`,在HTML中可以用 `class-concentrated` 代替,因为 `\emph` 是LaTeX标记,转换为HTML的 `` 是语义化转换,不算改内容。 原文有数学公式 `$0.00 \pm 0.00\%$`,HTML中可以直接保留文本,或者用 ``,但一般学术文本直接保留文本即可,不需要特别处理。 段落划分: 段落1:Federated learning over distributed intrusion-detection sensors naturally produces \emph{class-concentrated} non-IID data, where an entire attack class is held by a single client and absent elsewhere. We show this collapses federated averaging in a specific, localized way: the output layer holds one decision boundary per class, and when a class's aggregation weight comes overwhelmingly from clients that have never seen it, its neuron receives an active anti-signal rather than mere dilution, collapsing to zero. On CIC-DDoS2019, standard FedAvg drives the concentrated LDAP class to $0.00 \pm 0.00\%$ F1 across every seed. 段落2:We propose Class-Competence Aggregation (CCA), which aggregates the output layer per class using each client's own locally reported validation F1 as the weight, while hidden layers remain under standard size-weighted FedAvg. CCA requires no server-held auxiliary data and adds only one F1 scalar per class per round beyond the feature-scaling and class-count exchanges common to any federated pipeline. 段落3:On our primary CNN-LSTM architecture, CCA alone raises LDAP F1 from $0\%$ to $84.83 \pm 1.65\%$, and the full method reaches $91.64 \pm 0.57\%$ macro F1, closing the federated-to-centralized gap from $-42.53$ to $-4.05$ points. Benchmarked against four other federated methods under identical partitions and seeds, CCA is the only one that recovers the concentrated class substantially and reliably; a $178\times$ smaller MLP shows the same qualitative recovery, though less completely, confirming the effect holds across architecture scale even as its magnitude varies. Results are validated across three seeds and replicated on a second dataset, CICIDS2017. 重要句子可用``: 比如第一段中的 "We show this collapses federated averaging in a specific, localized way" 或者 "its neuron receives an active anti-signal rather than mere dilution, collapsing to zero." 第二段中的 "We propose Class-Competence Aggregation (CCA)" 第三段中的 "CCA is the only one that recovers the concentrated class substantially and reliably" 可以加粗这些关键句。 `\emph{class-concentrated}` 转为 `class-concentrated`。 构建HTML:

    Federated learning over distributed intrusion-detection sensors naturally produces class-concentrated non-IID data, where an entire attack class is held by a single client and absent elsewhere. We show this collapses federated averaging in a specific, localized way: the output layer holds one decision boundary per class, and when a class's aggregation weight comes overwhelmingly from clients that have never seen it, its neuron receives an active anti-signal rather than mere dilution, collapsing to zero. On CIC-DDoS2019, standard FedAvg drives the concentrated LDAP class to $0.00 \pm 0.00\%$ F1 across every seed.

    We propose Class-Competence Aggregation (CCA), which aggregates the output layer per class using each client's own locally reported validation F1 as the weight, while hidden layers remain under standard size-weighted FedAvg. CCA requires no server-held auxiliary data and adds only one F1 scalar per class per round beyond the feature-scaling and class-count exchanges common to any federated pipeline.

    On our primary CNN-LSTM architecture, CCA alone raises LDAP F1 from $0\%$ to $84.83 \pm 1.65\%$, and the full method reaches $91.64 \pm 0.