Description

New API is a large language mode (LLM) gateway and artificial intelligence (AI) asset management system. Prior to 1.0.0-rc.18, user-controlled image n, video seconds and duration, max_tokens, max_completion_tokens, maxOutputTokens, audio duration, and billing-expression quantities can overflow conversions in common/quota_math.go and related settlement paths, allowing a low-privileged account with positive balance or an active subscription to turn a negative charge into account credit and potentially drain upstream funds. This issue is fixed in version 1.0.0-rc.18.

Executive Summary

Technical Details

Mitigations

References

Additional Metadata

{
  "score": "9.1",
  "pubDate": "2026-08-17T16:17:44.307Z",
  "pubdate": "2026-08-17T16:17:44.307Z",
  "executiveSummary": "A numeric overflow vulnerability exists within the billing and quota calculation logic of New API prior to version 1.0.0-rc.18. The flaw resides in the processing of user-controlled request parameters such as image count n, video seconds, video duration, max_tokens, max_completion_tokens, maxOutputTokens, audio duration, and billing-expression quantities. These inputs are inadequately validated and handled during settlement paths, leading to arithmetic overflows within common/quota_math.go. An attacker with low privileges, possessing a positive account balance or an active subscription, can exploit this vulnerability to manipulate transaction costs. By deliberately triggering integer overflows, the system converts negative charge calculations into positive account credits, thereby allowing unauthorized financial enrichment and the potential depletion of upstream provider funds. Exploitation requires authenticated access with minimal privileges and malicious manipulation of supported API payload parameters.",
  "technicalDetails": "The vulnerability stems from unsafe integer arithmetic operations within the quota settlement and calculation mechanisms of New API. Specifically, the vulnerable component is located in common/quota_math.go and associated settlement execution paths. The root cause is the lack of robust bounds checking and overflow protection when computing costs derived from user-controlled variables. Affected parameters include image quantities (n), video duration metrics (seconds and duration), token constraints (max_tokens, max_completion_tokens, maxOutputTokens), audio duration values, and custom billing-expression quantities.\nDuring API request processing, the gateway calculates resource consumption and deducts costs from the user's quota balance. An authenticated attacker leverages low-privileged access—such as an account with a positive balance or an active subscription—to inject excessively large values into the aforementioned parameters. When these values are multiplied or aggregated during the execution of common/quota_math.go, the resulting numerical value exceeds the maximum limit of the underlying integer data type, causing an arithmetic overflow wrap-around.\nThe step-by-step attack flow proceeds as follows: First, the attacker authenticates to the New API gateway using a low-privileged account. Second, the attacker crafts a malicious API request targeting LLM or media generation endpoints, supplying hyper-inflated numerical inputs for parameters like max_tokens or video duration. Third, the gateway processes the request and passes these parameters to the settlement and quota calculation routines. Fourth, the arithmetic operation overflows, inverting the expected positive cost deduction into a negative deduction or an erroneous negative charge. Fifth, the system interprets this calculation anomaly as a refund or credit addition, depositing funds directly into the attacker's account balance. Finally, the attacker repeats the payload sequence to accumulate illicit credits, which can subsequently be utilized to drain upstream AI model provider funds."
}