Network security and related apparatuses, methods, and security systems
Network security and related apparatuses, methods, and security systems are disclosed. An apparatus includes a variational autoencoder trained to reconstruct a benign packet flow representation of a benign packet flow corresponding to a benign stream of packets. The processing circuitry is configured to apply a packet flow representation of a packet flow corresponding to a received stream of packets to the variational autoencoder to generate a reconstructed packet flow representation. The packet flow representation includes one or more of a determined transfer entropy corresponding to the received stream of packets, flow derived metadata, or a Granger causality of the packet flow. The processing circuitry is also configured to determine a reconstruction loss of the reconstructed packet flow representation and determine whether the received stream of packets is anomalous responsive to the determined reconstruction loss.
1 . An apparatus, comprising:
an input terminal configured to receive a stream of packets from a network; and
processing circuitry implementing a variational autoencoder trained to reconstruct a benign packet flow representation of a benign packet flow corresponding to a benign stream of packets, the processing circuitry configured to:
apply a packet flow representation of a packet flow corresponding to the received stream of packets to the variational autoencoder to generate a reconstructed packet flow representation, the packet flow representation comprising one or more of a determined transfer entropy corresponding to the received stream of packets or a Granger causality of the packet flow;
determine a reconstruction loss of the reconstructed packet flow representation based, at least in part, on the determined transfer entropy corresponding to the received stream of packets or the determined Granger causality of the packet flow; and
determine whether the received stream of packets is anomalous responsive to the determined reconstruction loss.
2 . The apparatus of claim 1 , wherein the variational autoencoder includes an input layer, an encoder layer, a latent layer, a decoder layer, and an output layer, the encoder layer configured to compress the packet flow representation into a latent space corresponding to the latent layer, the latent space corresponding to a random multivariable normal distribution, the decoder layer configured to generate the reconstructed packet flow representation responsive to information sampled from the latent space.
3 . The apparatus of claim 1 , wherein the processing circuitry is further configured to deliver the received stream of packets to a destination device responsive to a determination that the received stream of packets is not anomalous.
4 . The apparatus of claim 1 , wherein the processing circuitry is further configured to block the received stream of packets from a destination device for the received stream of packets responsive to a determination that the received stream of packets is anomalous.
5 . The apparatus of claim 1 , wherein the processing circuitry is configured to determine the packet flow representation responsive to the stream of packets itself.
6 . The apparatus of claim 1 , wherein the processing circuitry is configured to determine the packet flow representation responsive to summarized data indicating the packet flow.
7 . A method of operating a security system, the method comprising:
training a variational autoencoder to reconstruct a benign packet flow representation of a benign packet flow corresponding to a benign stream of packets;
determining a packet flow representation of a packet flow corresponding to a received stream of packets, the packet flow representation comprising one or more of a determined transfer entropy corresponding to the received stream of packets or a Granger causality of the packet flow;
applying the packet flow representation to the trained variational autoencoder to generate a reconstructed packet flow representation;
determining a reconstruction loss of the reconstructed packet flow representation based, at least in part, on the determined transfer entropy corresponding to the received stream of packets or the determined Granger causality of the packet flow; and
determining whether the received stream of packets is anomalous responsive to the reconstructed loss.
8 . The method of claim 7 , further comprising blocking the received stream of packets from delivery to a destination device responsive to a determination that the received stream of packets is anomalous.
9 . The method of claim 7 , further comprising delivering the received stream of packets to a destination device responsive to a determination that the received stream of packets is not anomalous.
10 . The method of claim 7 , wherein training the variational autoencoder comprises determining the benign packet flow representation of the benign packet flow responsive to the benign stream of packets itself.
11 . The method of claim 7 , wherein training the variational autoencoder comprises determining the benign packet flow representation responsive to summarized data indicating the benign packet flow.
12 . The method of claim 7 , wherein determining the packet flow representation comprises determining the packet flow representation of the packet flow responsive to the received stream of packets itself.
13 . The method of claim 7 , wherein determining the packet flow representation comprises determining the packet flow representation responsive to summarized data indicating the packet flow.
14 . The method of claim 7 , wherein training the variational autoencoder includes:
feeding the benign packet flow representation to the variational autoencoder;
building a directed acyclic graph hierarchy;
encoding the benign packet flow representation, by a separate encoder layer, responsive to latent dimension selection;
creating a latent distribution, using a random multivariable normal distribution, based on user-specified latent dimensions; and
decoding information sampled from the latent distribution to generate new examples from the latent distribution.
15 . A security system, comprising:
a destination device;
a network interface configured to receive a stream of packets from a network; and
processing circuitry configured to:
determine whether the received stream of packets is anomalous using a variational autoencoder trained to reconstruct a benign transfer entropy of a benign packet flow corresponding to a benign stream of packets; and
deliver the received stream of packets to the destination device responsive to a determination that the received stream of packets is not anomalous, the determination based, at least in part, on the benign transfer entropy of a benign packet flow corresponding to a benign stream of packets.
16 . The security system of claim 15 , wherein the processing circuitry is further configured to block the received stream of packets from delivery to the destination device responsive to a determination that the received stream of packets is anomalous.
17 . The security system of claim 15 , wherein the processing circuitry is further configured to determine a transfer entropy of a packet flow corresponding to the received stream of packets.
18 . The security system of claim 17 , wherein the processing circuitry is configured to determine the transfer entropy responsive to the received stream of packets itself.
19 . The security system of claim 17 , wherein the processing circuitry is configured to determine the transfer entropy responsive to summarization information indicating the packet flow.
20 . The security system of claim 17 , wherein the processing circuitry is configured to determine whether the received stream of packets is anomalous by applying the determined transfer entropy to the variational autoencoder and determining a reconstruction loss of the reconstructed transfer entropy relative to the determined transfer entropy.