You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
{{ message }}
Repository navigation
[ROS] Consider further mitigation of resource exhaustion with excessive argon2 parameters #708
Quick context: RFC9580 adds an intentionally expensive String-to-Key (S2K) conversion method based on Argon2, which can be triggered without additional user interaction as part of parsing message inputs. Previously, this allowed an attacker to specify extra-expensive derivations in terms of RAM memory and CPU time. To a large degree, this problem is inherent in the Argon2 design and S2K purpose.
a69992c introduced some upper limit ARGON2_MEMORY_LIMIT_KIB, but the potential issue remains, especially on setups that are sensitive to slow input handling or that have a limited amount of free memory.
Relevant steps to close this ticket could involve giving the library caller more control over the maximum allowed memory footprint for Argon2 key conversion attempts.
@hko-s side note: a potential expert API option to raise or lower the Argon2 limits could also be beneficial for more some fuzzing configurations, where slow input handling of Argon2 messages costs performance and leads to false-positive detection (for example, warnings on slow inputs).
Activity
ros-cr commentedon Apr 9, 2026
Quick context: RFC9580 adds an intentionally expensive String-to-Key (S2K) conversion method based on Argon2, which can be triggered without additional user interaction as part of parsing message inputs. Previously, this allowed an attacker to specify extra-expensive derivations in terms of RAM memory and CPU time. To a large degree, this problem is inherent in the Argon2 design and S2K purpose.
a69992c introduced some upper limit
ARGON2_MEMORY_LIMIT_KIB, but the potential issue remains, especially on setups that are sensitive to slow input handling or that have a limited amount of free memory.Relevant steps to close this ticket could involve giving the library caller more control over the maximum allowed memory footprint for Argon2 key conversion attempts.
ros-cr commentedon Apr 9, 2026
@hko-s side note: a potential expert API option to raise or lower the Argon2 limits could also be beneficial for more some fuzzing configurations, where slow input handling of Argon2 messages costs performance and leads to false-positive detection (for example, warnings on slow inputs).