Unbounded response body buffering and decompression in hackney

Summary

Allocation of Resources Without Limits or Throttling vulnerability in hackney_conn in benoitc hackney allows a malicious HTTP server to exhaust client memory via an oversized or highly compressed response body. hackney:request/5 and hackney:body/1 return the whole response body, and nothing bounds how much the connection process buffers. Any server the client requests can trigger it.

hackney_conn:read_full_body/2 and the HTTP/2 and HTTP/3 accumulators append every chunk until the body ends. With {auto_decompress, true} the buffered body is then inflated in one step with zlib:gunzip/1 or zlib:inflate/2, which return the whole output at once.

This issue affects hackney: from 2.0.0-beta.1 before 4.9.0.

Details

1. Unbounded accumulation. hackney:request/5 with a body, and hackney:body/1 after start_response/1, read the response in src/hackney_conn.erl through read_full_body/2, which appends each chunk to a binary accumulator until the body ends. The HTTP/2 and HTTP/3 synchronous paths accumulate stream data the same way. No option caps the size.

2. One-step decompression. With {auto_decompress, true} hackney sends Accept-Encoding: gzip, deflate. Once the body is complete, maybe_decompress_body/2 dispatches on Content-Encoding to decompress_gzip/1 (zlib:gunzip/1) or decompress_deflate/1 (zlib:inflate/2). Both return the whole output at once, so the inflated size is bounded only by the compression ratio. The report measured a 1 MB gzip body inflating to 1 GB on hackney 4.8.4.

Proof of concept

1. Build a gzip stream whose inflated size is 1 GB from a reused 64 KB chunk of zero bytes (deflate with window bits 31, best compression). The stream is about 1 MB on the wire.
2. Start a TCP listener on loopback that accepts one connection, reads the request, and answers HTTP/1.1 200 OK with Content-Encoding: gzip, a Content-Length matching the stream, Connection: close, and the stream as body.
3. From a node running a vulnerable hackney, call hackney:request(get, Url, [], <<>>, [{auto_decompress, true}, with_body]) against that listener.
4. Observe the returned body of 1,048,576,000 bytes and the VM memory growing by about 1.8 GB for a 1,019,197 byte response. A larger stream, or an endless body without auto_decompress, drives the node out of memory.

Impact

A server the client requests can make the client allocate memory proportional to the body it sends, or to its decompressed size, until the node runs out of memory.

Workarounds

Leave auto_decompress disabled (the default). Without it hackney_conn:maybe_decompress_body/2 never inflates the body, which closes the amplification half. The uncapped raw body read has no workaround.