Know every vulnerabilitybefore it knows you.
DevGuard continuously monitors your dependencies and alerts you when CVEs like this one affect your stack — with real-time threat intelligence built for developers.
EEF-CVE-2026-115274
Summary
Improper Validation of Specified Quantity in Input vulnerability in the Elixir standard library's JSON module allows an unauthenticated attacker who supplies a JSON document to cause a denial of service through CPU and memory exhaustion.
JSON.decode/1 and JSON.decode!/1 convert each integer literal with erlang:binary_to_integer/1 and do not bound its length. A literal of one million digits decodes successfully but costs 100 to 300 times as much CPU as a string of the same length, so one request body of a few megabytes holds a scheduler for about a second. A literal of more than about 1.29 million digits raises SystemLimitError. Applications generated by phx.new decode request bodies with JSON in Plug.Parsers, before routing and authentication.
This issue affects elixir: from 1.18.0-rc.0 before 1.18.6, from 1.19.0-rc.0 before 1.19.7, and from 1.20.0-rc.0 before 1.20.5.
Details
1. Default integer decoder
JSON.decode/1 and JSON.decode!/1 call JSON.decode/3 without an integer decoder. Up to the removal of Erlang/OTP 26 support, JSON.decode/3 used the vendored elixir_json module, whose default is fun erlang:binary_to_integer/1. It now calls OTP's :json.decode/3, which has the same default. The JSON.decode/3 documentation names &String.to_integer/1 as the default, which converts the same way.
2. Unbounded conversion
The decoder scans the digits of an integer literal and passes the whole run to the conversion. Nothing limits the run's length, so the cost grows faster than linearly with the number of digits. At one million digits an integer literal costs 100 to 300 times as much CPU as a string literal of the same length. Above about 1.29 million digits the conversion raises SystemLimitError, which JSON.decode/3 does not catch.
3. Reach
phx.new sets config :phoenix, :json_library, JSON, and the generated endpoint passes Phoenix.json_library() to Plug.Parsers as json_decoder. Plug.Parsers runs before routing and authentication and accepts bodies up to 8 MB by default. It rescues the SystemLimitError into a 400 response after the CPU cost is spent. Jason, the previous default, rejects integer literals longer than its decoding_integer_digit_limit (1024 by default) before conversion.
Proof of concept
- Build
body = "[" <> Enum.map_join(1..7, ",", fn _ -> String.duplicate("9", 1_000_000) end) <> "]", a body of about 7 MB. JSON.decode(body)returns{:ok, ...}after about one second of CPU.JSON.decode(String.duplicate("9", 1_000_000))takes 100 to 300 times as long asJSON.decode("\"" <> String.duplicate("9", 1_000_000) <> "\"").JSON.decode(String.duplicate("9", 1_400_000))raisesSystemLimitError.
Impact
An unauthenticated client that sends a JSON body with integer literals of about a million digits makes the server spend about a second of CPU per request and allocate large bignums. Sustained requests exhaust CPU on applications that decode untrusted JSON with JSON, including new Phoenix applications by default.
Workarounds
Decode untrusted JSON with an integer decoder that rejects long digit runs before conversion:
defmodule SafeJSON do
def decode!(binary) do
case JSON.decode(binary, :ok, integer: &safe_integer/1) do
{decoded, :ok, ""} -> decoded
_ -> raise ArgumentError, "invalid JSON"
end
end
defp safe_integer(digits) when byte_size(digits) <= 1024, do: String.to_integer(digits)
defp safe_integer(_digits), do: raise(ArgumentError, "integer too long")
end
In Phoenix, pass SafeJSON to Plug.Parsers as json_decoder instead of Phoenix.json_library().
Upload your own SBOM in CycloneDX 1.6 or higher (JSON) directly here to check your vulnerabilities.
Drag and drop some file here, or click to select
The vulnerability can be exploited over the network without needing physical access. It is easy for an attacker to exploit this vulnerability. An attacker does not need any special privileges or access rights. No user interaction is needed for the attacker to exploit this vulnerability.
Limited exploitation activity has been observed. Close monitoring and planned remediation are recommended.
The exploit probability is very low. The vulnerability is unlikely to be exploited in the next 30 days.
We did not find any exploit available. Neither in GitHub repositories nor in the Exploit-Database.
Browse More
Continuously monitor your dependencies and get alerted when vulnerabilities like this one affect your stack.
Checkout DevGuard