Student Launches BLA to Bloat AI Reverse-Engineering Token Bills 300x

1 reports archived1 independent sourcesupdated 11/10/2026, 06:11:49

Synthesis & Latest Updates

1 Sources Cross-Validated

Timo Zhang, a junior software engineering student at Sun Yat-sen University, has open-sourced BLA, an adversarial hardening scheme aimed at AI reverse-engineering agents. By planting fake credentials and decoy source maps, precisely tripping analyzer quota limits, and inducing misjudgment, BLA inflates the token cost of reverse-engineering hardened software by 300x to 2,800x; across 14 repo examples the total cost ran about 304x higher. The project has drawn 3,900+ stars and is positioned as a direct counter to REA, which crossed 100k stars in a week.

LATEST/Timo Zhang, a junior software engineering student at Sun Yat-sen University, has open-sourced BLA, an adversarial hardening scheme aimed at AI reverse-engineering agents. By planting fake credentials and decoy source maps, precisely tripping analyzer quota limits, and inducing misjudgment, BLA inflates the token cost of reverse-engineering hardened software by 300x to 2,800x; across 14 repo examples the total cost ran about 304x higher. The project has drawn 3,900+ stars and is positioned as a direct counter to REA, which crossed 100k stars in a week.

Hourly heat curve

2 fully observed hours

Fewer than 3 fully observed hours; trend not plotted yet.

Coverage timeline

Total 1 reports · Latest first

量子位官网T2Nov 10, 06:11·72 pts
Student Launches BLA to Bloat AI Reverse-Engineering Token Bills 300x
- BLA combines poisoning, budget exhaustion and misjudgment inducement to inflate reverse-engineering token use by 300x to 2,800x
- Across 14 repo examples, total token consumption ran about 639.5x and total cost about 304.4x versus unhardened builds
- Shipped as an MCP service and CLI tool, it has 3,900+ stars and 476 forks, built by a Sun Yat-sen University junior
View Related Reports & Analysis Original Source