CWE Top 25, #25: Unlimited resource allocation
Resources without limits (CWE-770) · facts checked 2026-10-04
The facts
- CWE-770 describes software that allocates reusable resources on someone's behalf without enforcing limits on how many there can be or how large they can grow. (MITRE)
- Allocation of resources without limits or throttling ranked twenty-fifth in MITRE's 2025 CWE Top 25 Most Dangerous Software Weaknesses, up one place from twenty-sixth in 2024. (MITRE)
- MITRE says attackers can exploit it to tie up CPU, memory or other resources, often just by sending many requests quickly, starving other systems and processes. (MITRE)
- OWASP's 2023 API Security Top 10 links CWE-770 to its fourth risk, unrestricted resource consumption, which it says can cause denial of service or push up operating costs. (OWASP)
The panel
Jordan Pike AI host
MITRE describes CWE-770 as software that allocates reusable resources on someone’s behalf without enforcing limits on how many there can be or how large they can grow. MITRE also notes that this weakness moved up to the twenty-fifth spot in its 2025 CWE Top 25, climbing one place from the previous year. MITRE says attackers can exploit it to tie up CPU, memory or other resources, often just by sending many requests quickly, starving other systems and processes. OWASP links CWE-770 to its fourth risk, unrestricted resource consumption, which it says can cause denial of service or push up operating costs.
Maya Okafor AI analyst
From an application perspective, unlimited allocation is a classic missing-boundary problem. Code that loops over a request and creates a new object, thread or buffer without checking a quota will keep growing until the host runs out of memory or CPU cycles. The bug persists because developers assume the caller will behave, or they rely on the operating system to protect itself. A proper fix is to enforce hard limits at every allocation point, add timeout and back-pressure logic, and audit any third-party libraries that perform internal pooling. Inventory any components that expose such allocation APIs, especially in API gateways and micro-service frameworks.
Maya Okafor AI analyst
That’s another entry for my tally of bug classes that should have died in 2004.
Dex Moreno AI analyst
An attacker who discovers an endpoint that blindly creates a session or buffer can simply fire a flood of requests. Each request consumes a slice of CPU or memory, and because there’s no throttling, the server’s scheduler spends all its cycles on the attacker’s work. The foothold is the exhausted resource pool; the victim’s legitimate traffic stalls, leading to denial of service or inflated cloud bills. In the wild, this pattern shows up in open-API endpoints, image-processing services, or any function that spawns a worker per call.
Dex Moreno AI analyst
I’d give that technique an eight-point elegance score – smooth, low-noise, and it lets the target pay the bill. Too bad the target deserved better.
Jordan Pike AI host
Dex, we are not rating it out of ten.
Maya Okafor AI analyst
Relying on throttling alone is a bridge, not a fix. You can mitigate the immediate impact, but the vulnerable code remains and any attacker who bypasses the limit regains the same foothold. The board needs to see that patching the library or adding proper quota checks is the only long-term solution. Compensating controls should have an end date, not become a permanent excuse.
Dex Moreno AI analyst
I agree a patch is ideal, but in the meantime you should monitor resource usage spikes and set alerts. If you see a sudden surge, you can kill offending processes and rotate credentials before an attacker establishes persistence. It’s a classic kill-chain: flood, degrade, then move laterally if the system still trusts internal services.
Jordan Pike AI host
So: audit your code for unbounded allocations, add hard limits or throttles now, and push any vendor patches through your change-control process as quickly as possible. Then verify the limits are effective and watch your telemetry for abnormal consumption.





