Yuhao Zhang, Aws Albarghouthi, Loris Antoni
We lifted 8 functions out of this paper's own repositories and ran 5 of them in a sandbox. "Ran" means the function executed on a synthesized input and returned a value. It is not a reproduction of the paper's results.
| Repository | Role | Ran |
|---|---|---|
| foreverzyh/defend_framework | canonical | 5 of 8 |
| Function | Status | Where it lives |
|---|---|---|
| get_abstain_DPA | Ran | foreverzyh/defend_framework/certified_eval.py code served (permissive licence) · get_code("0375769d8cc30148") |
| get_abstain_bagging_replace | Ran | foreverzyh/defend_framework/certified_eval.py code served (permissive licence) · get_code("404cf124c90a1ef3") |
| get_count | Ran | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("62f8cc312a6e9038") |
| my_comb | Ran | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("3e9d553c69e15339") |
| my_powe | Ran | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("97822baffffe7539") |
| BoundCalculator | Not yet run | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("f0a49775ce104f77") |
| FlipBoundCalculator | Not yet run | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("112c567f6ed569b6") |
| process_count | Not yet run | foreverzyh/defend_framework/utils/cal_bound.py code served (permissive licence) · get_code("ee9fe1dbf9fb50bf") |
Some links come from the archived Papers with Code dataset (CC BY-SA 4.0): attribution and licence.
Machine learning models are vulnerable to data-poisoning attacks, in which an attacker maliciously modifies the training set to change the prediction of a learned model. In a trigger-less attack, the attacker can modify the training set but not the test inputs, while in a backdoor attack the attacker can also modify test inputs. Existing model-agnostic defense approaches either cannot handle backdoor attacks or do not provide effective certificates (i.e., a proof of a defense). We present BagFlip, a model-agnostic certified approach that can effectively defend against both trigger-less and backdoor attacks. We evaluate BagFlip on image classification and malware detection datasets. BagFlip is equal to or more effective than the state-of-the-art approaches for trigger-less attacks and more effective than the state-of-the-art approaches for backdoor attacks. 36th Conference on Neural Information Processing Systems (NeurIPS 2022).
The same record, over MCP at https://syntology.ai/mcp:
get_harvested_code_for_paper("2205.13634")
get_code_for_paper("2205.13634")
have("2205.13634")
Connect an agent — have() is free.