Research map / Federated learning and privacy
Privacy attacks and secure learning: research map
428 papers on Privacy attacks and secure learning in Federated learning and privacy, from ICML, NeurIPS, ICLR, CVPR, AAAI, ICCV and ECCV, and the journals TPAMI, IJCV and JMLR (2016–2026), grouped into 2 clusters and 6 approaches. The busiest year so far is 2025.
What it is
Attacks that leak or corrupt data in collaborative learning, such as reconstructing training images from gradients or poisoning a federated model, and defences based on encryption, secure aggregation and auditing.
Terms you will see on this page:
- Gradient inversion
- reconstructing training data from shared gradients.
- Secure aggregation
- combining client updates so the server sees only the sum.
- Homomorphic encryption
- computing on encrypted data without decrypting it.
Start reading with the most cited papers here:
- CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy (ICML 2016 · 1,289 citations)
- Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware (ICLR 2019 · 245 citations)
Within Federated learning and privacy, its share shrank from 17.6% in 2023–24 to 15.6% in 2025–26 (131 → 142 papers at ICML, NeurIPS, ICLR, CVPR and AAAI, the venues with data for all four years).
Explore Privacy attacks and secure learning in the interactive map
Working on something in this topic? Describe your idea in scime atlas to see which approach it falls under, the closest papers by meaning and how crowded the spot has become.
What changed recently
Approaches whose share of the topic moved most between 2023–24 and 2025–26 (ICML, NeurIPS, ICLR, CVPR and AAAI):
- Defending federated learning against poisoning rising: 22% → 28% of the topic (29 → 40 papers)
- Membership inference and auditing rising: 11% → 15% of the topic (15 → 21 papers)
- Visual privacy and face de-identification rising: 8% → 11% of the topic (10 → 15 papers)
- Gradient inversion attacks falling: 15% → 8% of the topic (19 → 12 papers)
- Private federated aggregation falling: 19% → 13% of the topic (25 → 19 papers)
Papers from 2025–2026 that are already cited most:
- Model Poisoning Attacks to Federated Learning via Multi-Round Consistency (CVPR 2025 · 28 citations)
- Federated Unlearning with Gradient Descent and Conflict Mitigation (AAAI 2025 · 23 citations)
Approaches and key papers
Encryption and secure learning · 229 papers
Keywords: preserving, vfl, encryption
Approaches in this cluster:
- Encryption-based private learning (111 papers)
Use homomorphic encryption and private synthetic data for privacy-preserving learning. - Private federated aggregation (50 papers)
Secure aggregation and encrypted gradients for efficient, fair federated learning. - Visual privacy and face de-identification (68 papers)
Obfuscate faces and private content in images while keeping utility.
Most cited and most cited since 2024:
- CryptoNets: Applying Neural Networks to Encrypted Data with High Throughput and Accuracy (ICML 2016 · 1,289 citations)
- Slalom: Fast, Verifiable and Private Execution of Neural Networks in Trusted Hardware (ICLR 2019 · 245 citations)
- Privacy-Preserving Face Recognition Using Trainable Feature Subtraction (CVPR 2024 · 33 citations)
- No Prejudice! Fair Federated Graph Neural Networks for Personalized Recommendation (AAAI 2024 · 32 citations)
Attacks and defenses in FL · 199 papers
Keywords: attack, defense, backdoor
Approaches in this cluster:
- Defending federated learning against poisoning (89 papers)
Detect and mitigate malicious clients and backdoors in federated learning with provable defenses. - Gradient inversion attacks (58 papers)
Reconstruct training data from shared gradients and defend against such leakage. - Membership inference and auditing (52 papers)
Attack and audit models for privacy leakage with membership inference and statistical tests.
Most cited and most cited since 2024:
- Inverting Gradients - How easy is it to break privacy in federated learning? (NeurIPS 2020 · 206 citations)
- Soteria: Provable Defense Against Privacy Leakage in Federated Learning From Representation Perspective (CVPR 2021 · 168 citations)
- Revamping Federated Learning Security from a Defender's Perspective: A Unified Defense with Homomorphic Encrypted Data Space (CVPR 2024 · 43 citations)
- Model Poisoning Attacks to Federated Learning via Multi-Round Consistency (CVPR 2025 · 28 citations)
