On the Security Relevance of Initial Weights in Deep Neural Networks.
Kathrin GrosseThomas Alexander TrostMarius MosbachMichael BackesDietrich KlakowPublished in: ICANN (1) (2020)
Keyphrases
- neural network
- relative importance
- back propagation
- information security
- pattern recognition
- artificial neural networks
- feed forward
- linear combination
- intrusion detection
- security analysis
- security requirements
- fuzzy systems
- recurrent neural networks
- neural network model
- information retrieval
- genetic algorithm
- fuzzy logic
- hidden neurons
- encryption scheme
- security systems
- security issues
- security problems
- weighted sum
- training process
- connection weights
- network security
- relevance feedback
- multilayer perceptron
- access control
- security threats
- inverse document frequency
- weight update
- computer security
- weighting scheme
- statistical databases
- multi layer
- neural nets
- self organizing maps
- test collection
- search engine