From b412f18903ed4aed4fea383e0c9768ea34252112 Mon Sep 17 00:00:00 2001 From: Gary0901 Date: Thu, 24 Oct 2024 18:04:05 +0800 Subject: [PATCH] 10/24 update --- public/featureSelection.js | 2 +- public/userIdentification.js | 259 +++++++++++++++++++---------------- server.js | 2 +- 3 files changed, 141 insertions(+), 122 deletions(-) diff --git a/public/featureSelection.js b/public/featureSelection.js index 778fdd0..584a2e2 100644 --- a/public/featureSelection.js +++ b/public/featureSelection.js @@ -1,7 +1,7 @@ const _ = require('lodash'); // 接受指紋數據和一個唯一值閾值作為參數 -function selectFeatures(fingerprintData, uniquenessThreshold = 0.1, stabilityThreshold = 0.7, minFeatures = 5 /* , maxComputationTime = 100 */) { +function selectFeatures(fingerprintData, uniquenessThreshold = 0.1, stabilityThreshold = 0.7, minFeatures = 7 /* , maxComputationTime = 100 */) { //步驟一　：　收集所有用戶的所有屬性 reduce方法遍歷所有fingerprintData const allFeatures = fingerprintData.reduce((acc,fp)=>{ fp.components.forEach(comp=>{ diff --git a/public/userIdentification.js b/public/userIdentification.js index 9218618..0cb4ac9 100644 --- a/public/userIdentification.js +++ b/public/userIdentification.js @@ -1,112 +1,80 @@ // 實現了 K nearest neighbors 的算法。它接受已知的指紋，新的指紋，選定的特徵值和K值作為參數。 const _= require('lodash'); //添加這行來引入 lodash -function determineOptimalK(totalFingerprints) { - // 常見的經驗法則 : k = √n，其中n為樣本總數 - const k = Math.round(Math.sqrt(totalFingerprints.length)); - - // 設置合理的上下限 - const minK = 3; // 至少要3個鄰居 - const maxK = 15; // 至少要15個鄰居 - - return Math.min(Math.max(k,minK),maxK); -} - function identifyUser(knownFingerprints, newFingerprint, selectedFeatures){ - const k = determineOptimalK(knownFingerprints); if(knownFingerprints.length === 0 || selectedFeatures.length === 0) { console.log("No known fingerprints of selected features.Treating as new user"); return { isSameUser : false, predictedUserId : null, // 使用 _id - confidence : 0, - nearestDistance : Infinity, - nearestNeighbors:[] // 添加以便追蹤 + distance: Infinity, }; } - // 計算指紋與所有已知指紋的距離 - const distances = knownFingerprints.map(known=>{ - const distance = calculateDistance(known, newFingerprint, selectedFeatures); - return { - userId: known._id, // 儲存 _id - distance : distance, - fingerprint: known.fingerprint // 保留fingerprint 以便除錯 - }; - }); - - // 按距離排序並選擇k個最近鄰 - distances.sort((a,b) => a.distance - b.distance); - const nearestNeighbors = distances.slice(0, Math.min(k, distances.length)); - - // 分析最近鄰中的用戶ID分布 - const userCounts = {}; - nearestNeighbors.forEach(n=>{ - userCounts[n.userId] = (userCounts[n.userId] || 0) +1; - }) - - // 選擇最多的用戶ID - const [mostFrequentUserId, count] = Object.entries(userCounts) - .sort((a,b)=>b[1] - a[1])[0]; - - // 計算 confidence - const confidence = count / k; - - // 計算距離閾值 - let meanDistance = 0; - let stdDevDistance = 0; - let threshold = Infinity; - - if (nearestNeighbors.length > 0) { - meanDistance = _.mean(nearestNeighbors.map(n => n.distance)); - stdDevDistance = Math.sqrt(_.mean(nearestNeighbors.map(n => Math.pow(n.distance - meanDistance, 2)))); - threshold = meanDistance + stdDevDistance; - } + try { + // 計算指紋與所有已知指紋的距離 + const distances = knownFingerprints.map(known=>{ + const distance = calculateDistance(known, newFingerprint, selectedFeatures); + return { + userId: known._id, // 儲存 _id + distance : distance, + fingerprint: known.fingerprint // 保留fingerprint 以便除錯 + }; + }); + + // 按距離排序並選擇最近的匹配 + distances.sort((a,b) => a.distance - b.distance); + const nearestMatch = distances[0]; + + // 設定閾值 (這個閾值需要根據實際數據調整) + const threshold = 0.2; + + // 判斷是否為同一用戶。 + const isSameUser = nearestMatch.distance < threshold; + + console.log("Distance analysis:",{ + nearestDistance : nearestMatch.distance, + threshold : threshold, + allDistances : distances + }); - console.log("Nearest neighbors:",nearestNeighbors.map(n=>({ - distance:n.distance, - isAuthentic:n.fingerprint.isAuthentic, - fingerprintId:n.fingerprint._id - }))); - console.log("Mean distance:",meanDistance,"StdDev:",stdDevDistance,"Threshold:",threshold); - - - return { - isSameUser : confidence >= 0.6 && distances[0].distance < threshold, - predictedUserId : mostFrequentUserId, - confidence : confidence, - nearestDistance : distances[0].distance, - threshold : threshold, - nearestNeighbors:nearestNeighbors // 返回最近鄰資訊以便除錯 - }; + return { + isSameUser : isSameUser, + predictedUserId : isSameUser ? nearestMatch.userId : null, + distance : nearestMatch.distance, + matches:distances // 返回所有距離供分析 + } + } catch (error) { + console.error("Error in user identification:",error); + return { + isSameUser: false, + predictedUserId: null, + distance: Infinity, + error: error.message + } + } + } -function calculateDistance(fp1,fp2,selectedFeatures){ - // 先印出debug的資訊 - console.log("Calculating distance between",selectedFeatures); +function calculateDistance(fp1, fp2, selectedFeatures) { + console.log("Calculating distance between", selectedFeatures); - return selectedFeatures.reduce((sum,feature)=>{ - const comp1 = fp1.components.find(c=>c.key === feature); - const comp2 = fp2.components.find(c=>c.key === feature); + return selectedFeatures.reduce((sum, feature) => { + const comp1 = fp1.components.find(c => c.key === feature); + const comp2 = fp2.components.find(c => c.key === feature); if(!comp1 || !comp2) { - console.log(`Waring :Feature${feature} not found in one of the fingerprints`); - return sum + 1; // Treat missing features as maximum difference + console.log(`Warning: Feature ${feature} not found in one of the fingerprints`); + return sum + 1; } - /* console.log(`Comparing feature ${feature}:`, { - value1: comp1.value, - type1: typeof comp1.value, - value2: comp2.value, - type2: typeof comp2.value - }); */ - - // 針對不同類型的特徵使用不同的比較方法 try { + let featureDistance; + switch (feature) { + // 螢幕相關 case 'screenResolution': case 'availableScreenResolution': - // 確保值是字串或可以轉換為字串 const str1 = String(comp1.value); const str2 = String(comp2.value); if (str1.includes('x') && str2.includes('x')) { @@ -114,20 +82,27 @@ function calculateDistance(fp1,fp2,selectedFeatures){ const [width2, height2] = str2.split('x').map(Number); const widthDiff = Math.abs(width1 - width2) / Math.max(width1, width2); const heightDiff = Math.abs(height1 - height2) / Math.max(height1, height2); - return sum + (widthDiff + heightDiff) / 2; + featureDistance = (widthDiff + heightDiff) / 2; + } else { + featureDistance = str1 === str2 ? 0 : 1; } - return sum + (str1 === str2 ? 0 : 1); - + break; + + // 語言相關 + case 'language': + case 'languages': + const langs1 = Array.isArray(comp1.value) ? comp1.value : [comp1.value]; + const langs2 = Array.isArray(comp2.value) ? comp2.value : [comp2.value]; + const langIntersection = langs1.filter(l => langs2.includes(l)); + featureDistance = 1 - (langIntersection.length / Math.max(langs1.length, langs2.length)); + break; + + // 插件相關 case 'plugins': - // 安全地處理插件資訊 const getPluginList = (value) => { - if (Array.isArray(value)) { - return value; - } else if (typeof value === 'string') { - return value.split(','); - } else if (typeof value === 'object' && value !== null) { - return Object.values(value); - } + if (Array.isArray(value)) return value; + if (typeof value === 'string') return value.split(','); + if (typeof value === 'object' && value !== null) return Object.values(value); return []; }; @@ -135,35 +110,45 @@ function calculateDistance(fp1,fp2,selectedFeatures){ const plugins2 = new Set(getPluginList(comp2.value)); const intersection = new Set([...plugins1].filter(x => plugins2.has(x))); const union = new Set([...plugins1, ...plugins2]); - return sum + (1 - (intersection.size / union.size)); + featureDistance = 1 - (intersection.size / union.size); + break; + // 渲染相關 case 'webgl': - // 處理 WebGL 資訊 + case 'canvas': + case 'webglVendor': + case 'webglRenderer': if (typeof comp1.value === 'object' && typeof comp2.value === 'object') { - // 如果是物件，比較其字串表示 - return sum + (JSON.stringify(comp1.value) === JSON.stringify(comp2.value) ? 0 : 1); + featureDistance = JSON.stringify(comp1.value) === JSON.stringify(comp2.value) ? 0 : 1; + } else { + featureDistance = comp1.value === comp2.value ? 0 : 1; } - return sum + (comp1.value === comp2.value ? 0 : 1); + break; + // 硬體相關 case 'touchSupport': - // 處理觸控支援資訊 if (typeof comp1.value === 'object' && typeof comp2.value === 'object') { const touch1 = comp1.value.maxTouchPoints || 0; const touch2 = comp2.value.maxTouchPoints || 0; - return sum + (touch1 === touch2 ? 0 : 0.5); + featureDistance = touch1 === touch2 ? 0 : 0.5; + } else { + featureDistance = comp1.value === comp2.value ? 0 : 1; } - return sum + (comp1.value === comp2.value ? 0 : 1); + break; + case 'hardwareConcurrency': + case 'deviceMemory': + const value1 = Number(comp1.value) || 0; + const value2 = Number(comp2.value) || 0; + featureDistance = Math.abs(value1 - value2) / Math.max(value1, value2, 1); + break; + + // 字體相關 case 'fonts': - // 處理字體列表 const getFontList = (value) => { - if (Array.isArray(value)) { - return value; - } else if (typeof value === 'string') { - return value.split(','); - } else if (typeof value === 'object' && value !== null) { - return Object.values(value); - } + if (Array.isArray(value)) return value; + if (typeof value === 'string') return value.split(','); + if (typeof value === 'object' && value !== null) return Object.values(value); return []; }; @@ -171,23 +156,57 @@ function calculateDistance(fp1,fp2,selectedFeatures){ const fonts2 = new Set(getFontList(comp2.value)); const fontIntersection = new Set([...fonts1].filter(x => fonts2.has(x))); const fontUnion = new Set([...fonts1, ...fonts2]); - return sum + (1 - (fontIntersection.size / fontUnion.size)); - + featureDistance = 1 - (fontIntersection.size / fontUnion.size); + break; + + // 時區相關 + case 'timezone': + case 'timezoneOffset': + featureDistance = comp1.value === comp2.value ? 0 : 1; + break; + + // 瀏覽器/系統功能 + case 'sessionStorage': + case 'localStorage': + case 'indexedDb': + case 'addBehavior': + case 'openDatabase': + featureDistance = comp1.value === comp2.value ? 0 : 1; + break; + + // 用戶代理相關 + case 'userAgent': + case 'platform': + case 'vendor': + featureDistance = 1 - stringSimilarity(String(comp1.value), String(comp2.value)); + break; + + // 預設比較邏輯 default: - // 預設比較邏輯 if (typeof comp1.value === 'string' && typeof comp2.value === 'string') { - return sum + (1 - stringSimilarity(comp1.value, comp2.value)); + featureDistance = 1 - stringSimilarity(comp1.value, comp2.value); } else if (typeof comp1.value === 'number' && typeof comp2.value === 'number') { - return sum + Math.abs(comp1.value - comp2.value) / Math.max(Math.abs(comp1.value), Math.abs(comp2.value), 1); + featureDistance = Math.abs(comp1.value - comp2.value) / Math.max(Math.abs(comp1.value), Math.abs(comp2.value), 1); } else if (typeof comp1.value === 'object' && typeof comp2.value === 'object') { - return sum + (JSON.stringify(comp1.value) === JSON.stringify(comp2.value) ? 0 : 1); + featureDistance = JSON.stringify(comp1.value) === JSON.stringify(comp2.value) ? 0 : 1; } else { - return sum + (comp1.value === comp2.value ? 0 : 1); + featureDistance = comp1.value === comp2.value ? 0 : 1; } + break; } + + // 在計算完距離後，印出比較資訊 + console.log(`Comparing ${feature}:`, { + value1: comp1.value, + value2: comp2.value, + distance: featureDistance + }); + + return sum + featureDistance; + } catch (error) { console.error(`Error comparing feature ${feature}:`, error); - return sum + 1; // 如果比較出錯，視為最大差異 + return sum + 1; } }, 0) / selectedFeatures.length; } diff --git a/server.js b/server.js index d258fba..b9ca1a8 100644 --- a/server.js +++ b/server.js @@ -42,7 +42,7 @@ app.post('/login', async (req, res) => { } */ // Select features - const selectedFeatures = allFingerprints.length > 0 ? selectFeatures(allFingerprints,0.1,0.7,5):[]; // 可以進一步降低閾值 + const selectedFeatures = allFingerprints.length > 0 ? selectFeatures(allFingerprints,0.1,0.7,7):[]; // 可以進一步降低閾值 console.log("Selected Features:",selectedFeatures); // 用戶識別