Machine Learning & Security Seminar

Welcome to Machine Learning & Security Seminar, organized by Samuel Conte Professor Xiangyu Zhang, in Department of Computer Science at Purdue University.

• 2023 [Spring 2023]
• 2022 [Fall 2022], [Summer 2022], [Spring 2022]
• 2021 [Fall 2021], [Summer 2021]

Spring 2023

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
01/12	Guanhong Tao	The "Beatrix'' Resurrections: Robust Backdoor Detection via Gram Matrices	NDSS 2023
01/20	Xuan Chen	Provable Defense against Backdoor Policies in Reinforcement Learning	NeurIPS 2022
01/27	Zian Su	Chain-of-Thought Reasoning & TypeT5: Seq2seq Type Inference using Static Analysis	ICLR 2023
02/03	Canceled
02/10	Kaiyuan Zhang	FedRecover: Recovering from Poisoning Attacks in Federated Learning using Historical Information	S&P 2023
02/24	Canceled
03/03	Lu Yan	A Study of the Attention Abnormality in Trojaned BERTs	NAACL 2023
03/10	Yunshu Mao	PoisonedEncoder: Poisoning the Unlabeled Pre-training Data in Contrastive Learning	USENIX 2022
03/17	Siyuan Cheng	Backdoor defense
03/24	Canceled
03/31	Guangyu Shen	Single Image Backdoor Inversion via Robust Smoothed Classifiers	CVPR 2023
04/07	Canceled
04/14	Qiuling Xu	RecSys 2020 Tutorial on Conversational Recommendation Systems	RecSys 2020
04/21	Canceled
04/28	Shengwei An	Consistency Models	arXiv 2023
05/05	Canceled	Backdoor Attacks and Defenses in Machine Learning (BANDS) Workshop Day!	ICLR 2023

Fall 2022

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
08/26	Shengwei An	Blacklight: Scalable Defense for Neural Networks against Query-Based Black-Box Attacks	USENIX 2022
09/02	Yingqi Liu	Not All Poisons are Created Equal: Robust Training against Data Poisoning	ICML 2022
09/09	Canceled
09/16	Canceled
09/23	Kaiyuan Zhang	Auditing Privacy Defenses in Federated Learning via Generative Gradient Leakage	CVPR 2022
09/30	Zian Su	Repository-Level Prompt Generation for Large Language Models of Code	arXiv 2022
10/07	Xuan Chen	Robust Reinforcement Learning with Alternating Training of Learned Adversaries	ICLR 2021
10/14	Canceled
10/21	Lu Yan	Data Isotopes for Data Provenance in DNNs	arXiv 2022
10/28	Siyuan Cheng	Patch-Fool: Are Vision Transformers Always Robust Against Adversarial Perturbations?	ICLR 2022
11/04	Guangyu Shen	Quarantine: Sparsity Can Uncover the Trojan Attack Trigger for Free	CVPR 2022
11/11	Canceled
11/18	Qiuling Xu	Denoising Diffusion Implicit Models	ICLR 2021
11/25	Canceled	Happy Thanksgiving
12/02	Linyi Li (UIUC)	Certifiable Deep Learning at Scale towards Trustworthy Machine Learning
12/09	Shengwei An	Membership Inference Attacks From First Principles	S&P 2022
12/16	Canceled	Final Week
12/23	Canceled	Happy holidays!

Summer 2022

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
05/20	Yingqi Liu	Practice talk
	Guanhong Tao	Practice talk
05/27	Siyuan Cheng	Class-Disentanglement and Applications in Adversarial Detection and Defense	NeurIPS 2021
06/03	Canceled
06/10	Guanhong Tao	Backdoor defense
06/17	Canceled
06/24	Kaiyuan Zhang	CAFE: Catastrophic Data Leakage in Vertical Federated Learning	NeurIPS 2021
07/01	Zian Su	Black-Box Tuning for Language-Model-as-a-Service	ICML 2022
07/08	Xuan Chen	Robust Deep Reinforcement Learning through Adversarial Loss	NeurIPS 2021
07/15	Lu Yan	Fight Poison with Poison: Detecting Backdoor Poison Samples via Decoupling Benign Correlations	arXiv 2022
		Circumventing Backdoor Defenses That Are Based on Latent Separability	arXiv 2022
07/22	Guanhong Tao	Excess Capacity and Backdoor Poisoning	NeurIPS 2021
07/29	Canceled
08/05	Guangyu Shen	SPECTRE: Defending Against Backdoor Attacks Using Robust Statistics	ICML 2021
08/12	Siyuan Cheng	FLAME: Taming Backdoors in Federated Learning	USENIX 2022
08/19	Canceled

Spring 2022

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
01/07	Zian Su	Masked Autoencoders Are Scalable Vision Learners	arXiv 2021
01/14	Canceled
01/21	Shiwei Feng	Adversarial Neuron Pruning Purifies Backdoored Deep Models	NeurIPS 2021
01/28	Canceled
02/04	Shengwei An	Get a Model! Model Hijacking Attack Against Machine Learning Models	NDSS 2022
02/11	Siyuan Cheng	Anti-Backdoor Learning: Training Clean Models on Poisoned Data	NeurIPS 2021
02/18	Guanhong Tao	Traceback of Data Poisoning Attacks in Neural Networks	arXiv 2021
02/25	Canceled
03/04	Kaiyuan Zhang	Label Inference Attacks Against Vertical Federated Learning	USENIX 2022
03/11	Yingqi Liu	Back to the Drawing Board: A Critical Evaluation of Poisoning Attacks on Production Federated Learning	S&P 2022
03/18	Xuan Chen	Who Is the Strongest Enemy? Towards Optimal and Efficient Evasion Attacks in Deep RL	ICLR 2022
03/25	Canceled
04/01	Canceled
04/08	Guangyu Shen	Backdoor defense
04/15	Qiuling Xu	Few-shot learning in adversarial learning
04/22	Shengwei An	Practice talk: MIRROR: Model Inversion for Deep Learning Network with High Fidelity	NDSS 2022
		Model inversion defense
04/29	Canceled
05/06	Canceled
05/13	Canceled

Fall 2021

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
08/27	Zhiyuan Cheng	Dirty Road Can Attack: Security of Deep Learning based Automated Lane Centering under Physical-World Attack	USENIX 2021
09/03	Canceled
09/10	Guangyu Shen	Adversarial Attacks are Reversible with Natural Supervision	ICCV 2021
		Poisoning and Backdooring Contrastive Learning	arXiv 2021
09/17	Xuan Chen	Adversarial Policies: Attacking Deep Reinforcement Learning	ICLR 2020
		BACKDOORL: Backdoor Attack against Competitive Reinforcement Learning	IJCAI 2021
09/24	Guanhong Tao	BadEncoder: Backdoor Attacks to Pre-trained Encoders in Self-Supervised Learning	S&P 2022
10/01	Qiuling Xu	Stronger and Faster Wasserstein Adversarial Attacks	ICML 2020
10/08	Canceled
10/15	Zian Su	Prefix-Tuning: Optimizing Continuous Prompt for Generation	ACL 2021
10/22	Siyuan Cheng	Backdoor detection
10/29	Canceled
11/05	Guanhong Tao	Backdoor defense
11/12	Canceled
11/19	Shengwei An	Neural code editing
11/26	Canceled
12/03	Kaiyuan Zhang	Model-Contrastive Federated Learning	CVPR 2021
12/10	Yingqi Liu	Hidden Backdoors in Human-Centric Language Models	CCS 2021
		Backdoor Pre-trained Models Can Transfer to All	CCS 2021
12/17	Guangyu Shen	Rethinking Stealthiness of Backdoor Attack against NLP Models	ACL 2021

Summer 2021

This semester the seminar is co-organized by Kaiyuan Zhang.

Date	Discussion Leader	Title	Source
05/13	Guanhong Tao	What Doesn't Kill You Makes You Robust(er): Adversarial Training against Poisons and Backdoors	arXiv 2021
		PatchGuard: A Provably Robust Defense against Adversarial Patches via Small Receptive Fields and Masking	arXiv 2021
05/21	Yingqi Liu	Intrinsic Certified Robustness of Bagging against Data Poisoning Attacks	arXiv 2021
		Deep Partition Aggregation: Provable Defenses against General Poisoning Attacks	ICLR 2021
05/28	Canceled
06/04	Guangyu Shen	T-Miner: A Generative Approach to Defend Against Trojan Attacks on DNN-based Text Classification	USENIX 2021
06/11	Kaiyuan Zhang	Manipulating the Byzantine: Optimizing Model Poisoning Attacks and Defenses for Federated Learning	NDSS 2021
06/18	Shengwei An	Detecting and Simulating Artifacts in GAN Fake Images	WIFS 2019
		CNN-generated images are surprisingly easy to spot... for now	CVPR 2020
06/25	Yingqi Liu	Hidden Killer: Invisible Textual Backdoor Attacks with Syntactic Trigger	ACL 2021
		Turn the Combination Lock:Learnable Textual Backdoor Attacks via Word Substitution	arXiv 2021
07/02	Guanhong Tao	Backdoor learning
07/09	Guangyu Shen	See through Gradients: Image Batch Recovery via GradInversion	CVPR 2021
07/16	Canceled
07/23	Kaiyuan Zhang	How To Backdoor Federated Learning	PMLR 2020
		Attack of the Tails: Yes, You Really Can Backdoor Federated Learning	NeurIPS 2020
07/30	Shengwei An	Improving the Efficiency and Robustness of Deepfakes Detection through Precise Geometric Features	CVPR 2021
08/06	Yingqi Liu	Detecting AI Trojans Using Meta Neural Analysis	S&P 2021
08/13	Guanhong Tao	Double-Cross Attacks: Subverting Active Learning Systems	USENIX 2021