Research map / Architectures and efficiency / Attention and transformers
Classifying Sequences of Extreme Length with Constant Memory Applied to Malware Detection
Edward Raff, William Fleshman, Richard Zak, Hyrum S. Anderson, Bobby Filar, Mark McLean
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 |
|---|---|---|---|
| Add global channel gating | MalConv · Ember2018 | accuracy: 91.14 -> 93.29 | Improved |
| Add global channel gating | MalConv · Ember2018 | AUC: 97.29 -> 98.04 | Improved |
| Add global channel gating | MalConv · PDF corpus | accuracy: 99.16% -> 99.42% | Improved |
| Add global channel gating | MalConv · PDF corpus | AUC: 99.76% -> 99.80% | Improved |
| Train on full files with fixed-memory pooling instead of truncating at 2MB | MalConv · Ember2018 | accuracy: 91.27 -> 91.14 | No clear effect |
| Use fixed-memory temporal max pooling | MalConv · Ember2018 | GPU RAM: 128 GB -> 1.1 GB | Improved |
| Use fixed-memory temporal max pooling | MalConv · Ember2018 | time per epoch: 21 hr 29 min -> 1 hr 10 min | Improved |
| Replace global average pooling with fixed-memory temporal max pooling in AvastConv (ablation) | AvastConv · Ember2018 | accuracy | Hurt |
| Replace global average pooling with fixed-memory temporal max pooling in AvastConv (ablation) | AvastConv · Ember2018 | AUC | Hurt |
| Increase convolution depth beyond one layer (ablation) | MalConv and MalConv with GCG · Ember2018 | accuracy | Hurt |
| Increase MalConv filter count (ablation) | MalConv · Ember2018 | accuracy | No clear effect |
Closest papers by meaning
- Dynamic Malware Analysis with Feature Engineering and Feature Learning (AAAI 2020 · similarity 0.84)
- MalCL: Leveraging GAN-Based Generative Replay to Combat Catastrophic Forgetting in Malware Classification (AAAI 2025 · similarity 0.82)
- MalDetectFormer: Leveraging Sparse SpatioTemporal Information for Effective Malicious Traffic Detection (AAAI 2025 · similarity 0.81)
- Parallelizing non-linear sequential models over the sequence length (ICLR 2024 · similarity 0.79)
- DRSM: De-Randomized Smoothing on Malware Classifier Providing Certified Robustness (ICLR 2024 · similarity 0.79)
- FlashFFTConv: Efficient Convolutions for Long Sequences with Tensor Cores (ICLR 2024 · 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.
