1 / 78

SoK: Security and Privacy in Machine Learning

This paper explores the major security threats in machine learning and proposes solutions to mitigate them. It also discusses the challenges of maintaining confidentiality in machine learning models and user data.

suej
Download Presentation

SoK: Security and Privacy in Machine Learning

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. SoK: Security and Privacy in Machine Learning Nicolas Papernot∗, Patrick McDaniel∗, Arunesh Sinha†, and Michael P. Wellman† ∗ Pennsylvania State University † University of Michigan {ngp5056,mcdaniel}@cse.psu.edu, {arunesh,wellman}@umich.edu Presented by Jonny Dowdall CSE 914 2 Feb 2019

  2. What is machine learning? Relevant terms A few math proofs Threats What are the major security threats in machine learning? Solutions What can we do about these threats? Machine Learning Objectives

  3. Machine Learning An Over-Simplification

  4. Machine Learning A computer’s ability to infer something based on experience rather than being explicitly programmed. An Over-Simplification

  5. Machine Learning Given some input, produce some output. Supervised Learning

  6. Machine Learning Train by providing explicit input and output pairs. Supervised Learning

  7. Machine Learning Identify patterns in unlabeled data. Unsupervised Learning

  8. Machine Learning No training in the conventional sense. Unsupervised Learning

  9. Machine Learning Given a state, perform an action that leads you to an eventual desired state. Reinforcement Learning

  10. Machine Learning No explicitly correct output for any input. Reinforcement Learning

  11. Machine Learning Train by having an agent interact with an environment. Reinforcement Learning

  12. Machine Learning Reward/penalize agent for reaching desired states. Reinforcement Learning

  13. Machine Learning Typically, input data is represented as a vector (or matrix/tensor) of values. Call these input values features. Features

  14. Machine Learning Some function transforms the input features into target output(s). Parameters

  15. Machine Learning Function is typically a weighted sum of input features. Parameters

  16. Machine Learning Call the weights parameters. Parameters

  17. Machine Learning Start with random weights. Loss function

  18. Machine Learning Grab a sample from the training data and compute some (bogus) output value(s). Loss function

  19. Machine Learning The loss function tells us how far predicted output is from ground-truth output. Loss function

  20. Machine Learning The gradient of the loss function tells us how to change our weights. Gradient

  21. Machine Learning Next time we see that training sample, the loss should be smaller. Training

  22. Machine Learning Rinse and repeat with all training samples until loss is low and weights stop changing. Training

  23. Machine Learning Now, use these weights to predict values for new, unobserved inputs. This is our model. Inference

  24. Confidentiality/Privacy Uncovering ML model itself, which may be confidential intellectual privacy. Uncovering users’ data, which may be private (model input/output). Integrity Exploiting the model to produce outputs that don’t correspond to patterns in training data. Availability Preventing access to an output or action induced by a model output. Threats (CIA)

  25. Confidentiality/Privacy Uncovering ML model itself, which may be confidential intellectual privacy. Uncovering users’ data, which may be private (model input/output). Exploiting the model to produce outputs that don’t correspond to patterns in training data. Preventing access to an output or action induced by a model output. Integrity Threats (CIA)

  26. White-box vs. Black-box Adversaries White-box Black-box • Access to the machine-learning model internals • Can see parameters/architecture • Only able to interact with model • Can provide input and view output

  27. Papernot et al.

  28. Confidentiality

  29. Model extraction • Recover model parameters by testing many input output pairs (Tramer et al). • Requires access to class probabilities.

  30. Privacy

  31. Membership Attack • Test whether or not a data point is in the training set. • Exploit differences on model’s confidence to identify points that were trained on (Shokri). • Generate synthetic data until model produces an output with very high confidence.

  32. Model Inversion • Fredrikson et al. present the model inversion attack. • Given the output of a model, predict the input. • For a medicine dosage prediction task, they show that given access to the model and auxiliary information about the patient’s stable medicine dosage, they can recover genomic information about the patient.

  33. Analyzes the privacy guarantees of algorithms. Differential Privacy Framework

  34. An algorithm’s output should not differ significantly statistically for two versions of the data differing by only one record (Dwork et al). Differential Privacy Framework

  35. Add randomization to the machine learning pipeline. Solutions?

  36. Randomly add noise to every input while training. Solutions? Local Privacy

  37. E.g. Google Chrome collects user data with a probability q that the data is real and 1-q that the data is random. Resulted in meaningful and privacy-preserving statistics (Erlingsson) Solutions? Local Privacy

  38. It was shown that adding random noise to the loss function during training provides differential privacy (Chaudhuri). Solutions? Perturbing loss

  39. Randomly perturbing gradients before applying parameter updates guarantees even stronger differential privacy (Abadi). Solutions? Perturbing gradients

  40. Randomly perturb output values. Solutions? Noisy output

  41. This method degrades performance. Solutions? Noisy output

  42. “Precisely quantifying the learning algorithm’s sensitivity to training points is necessary to establish differential privacy guarantees. For nonconvex models (e.g., neural nets), current loose bounds on sensitivity require that learning be heavily randomized to protect data—often at the expense of utility.”

  43. Integrity

  44. Training attack • Poisoning: input/label manipulation • Test distribution is different from training distribution • Model learns to behave in an unintended way • Example • Intentionally label genuine emails as spam

  45. Assumption: Poisoned data points are typically outside the expected input distribution. Solution: Reduce the influence of outliers in training. Solutions?

  46. Use a modified PCA algorithm: • Maximize the variance between training samples while reducing variance of outliers. (Rubinstien et al.) Solutions? Reduce outlier sensitivity

  47. Adding a regularization term to the loss function, which in turn reduces the model sensitivity to outliers (Stempfel and L. Ralaivola). Solutions? Reduce outlier sensitivity

  48. Inference attack • Adversarial example • Exploit weakness in model • Perturb a sample until it is misclassified • E.g. How can we get past malware detection by changing our malware as little as possible?

  49. Inference attack • The more information provided in the output, the easier this is to exploit. • Especially easy when the system is an oracle. • Adversary can issue queries for any chosen input and observe the model output.

  50. Inference attack • Lowd and Meek introduce ACRE learnability, which poses the problem of finding the least cost modification to have a malicious input classified as benign using a polynomial number of queries to the ML oracle.

More Related