Research map / Label-efficient and robust learning / Recognition and long-tail
EMBERSim: A Large-Scale Databank for Boosting Similarity Search in Malware Analysis
Dragos Georgian Corlatescu, Alexandru Dinu, Mihaela Gaman, Paul Sumedrea
Read the paperCheck your idea against 137,217 papers
AI-extracted. The measured effects below were read by AI from the paper's full text (no summary yet). Numbers and claims can be wrong or missing context, so check the paper before you cite or build on it.
Measured effects reported in the paper
| What was changed | Where | Measured | Effect |
|---|---|---|---|
| use XGBoost leaf similarity instead of ssdeep | similarity search · EMBER label homogeneity, K=100, all | mean same-label hits: 40.42 -> 94.53 | Improved |
| use leaf similarity instead of ssdeep | similarity search · EMBER label homogeneity, K=10, all | mean same-label hits: 5.43 -> 9.74 | Improved |
| use leaf similarity instead of ssdeep | similarity search · EMBER label homogeneity, K=100, malicious | mean same-label hits: 61 -> 96.28 | Improved |
| use leaf similarity instead of ssdeep | similarity search · EMBER label homogeneity, K=100, benign | mean same-label hits: 19.84 -> 92.80 | Improved |
| increase top-K from 1 to 100 | leaf similarity · EMBER MAP, counterfactual, malicious | mean average precision: 0.960 -> 0.783 | Hurt |
| evaluate malicious instead of benign queries | leaf similarity · EMBER relevance EM, counterfactual, K=100 | relevance: 0.933 -> 0.641 | Hurt |
| use IoU instead of exact match as relevance function | leaf similarity · EMBER counterfactual, malicious, K=100 | relevance: 0.641 -> 0.759 | Improved |
| use NES instead of IoU as relevance function | leaf similarity · EMBER counterfactual, malicious, K=100 | relevance: 0.759 -> 0.752 | No clear effect |
| counterfactual (out-of-time) evaluation instead of unsupervised labelling | leaf similarity · EMBER MAP, malicious, K=1 | mean average precision: 0.716 -> 0.960 | Improved |
| apply leaf similarity to DREBIN instead of EMBER | leaf similarity · label homogeneity, K=100, all | mean same-label hits: 94.53 -> 98.30 | Improved |
| DREBIN malicious queries at K=100 | leaf similarity · DREBIN label homogeneity | mean same-label hits: 98.30 -> 71.21 | Hurt |
Closest papers by meaning
- Is Function Similarity Over-Engineered? Building a Benchmark (NeurIPS 2024 · similarity 0.83)
- DRSM: De-Randomized Smoothing on Malware Classifier Providing Certified Robustness (ICLR 2024 · similarity 0.82)
- MalCL: Leveraging GAN-Based Generative Replay to Combat Catastrophic Forgetting in Malware Classification (AAAI 2025 · similarity 0.81)
- SRDC: Semantics-based Ransomware Detection and Classification with LLM-assisted Pre-training (AAAI 2025 · similarity 0.80)
- Assemblage: Automatic Binary Dataset Construction for Machine Learning (NeurIPS 2024 · similarity 0.80)
- Pre-trained Behavioral Model for Malicious User Prediction on Social Platform (AAAI 2025 · similarity 0.79)
Writing a paper in this area? scime draft is a free LaTeX editor with AI included, real-time co-authors and a check of your references and venue rules before you submit. It also imports OpenAI Prism projects.
