Blockchain Based Secure IOT and Federated Learning
Text
Team 12
Aadharsh Aadhithya (CB.EN.U4AIE20001)
Akshaya Jeyaprakash(CB.EN.U4AIE20003)
Abinesh Sivakumar(CB.EN.U4AIE20002)
Jayanth M (CB.EN.U4AIE20024)
Vishnu Radhakrishnan (CB.EN.U4AIE20074)
Team Members
Blockchain Based Secure IOT and Federated Learning
Text
- Internet of Things and Need for security
- Blockchain and Security for IOT Networks
- Implementation of Blockchain over a simple smart-home Application using Rpi's.
- IOT and Federated Learning: Need for Federated Learning
- What is Federated Learning?
- Simple Proof-of-Concept for Federated learning using Rpi's.
- Implementation of Federated learning of XOR using Tensorflow
Blockchain Based Secure IOT and Federated Learning
Text
IOT
Blockchain Based Secure IOT and Federated Learning
Text
IOT
IOT over Blockchain network
Blockchain Based Secure IOT and Federated Learning
Text
IOT
IOT over Blockchain network
Federated Learning
Blockchain Based Secure IOT and Federated Learning
Text
- Internet of Things and Need for security
Text
Recent research indicates that 90% of consumers lack confidence in IoT device security. A 2019 survey done in
Australia, Canada, France, Japan, the U.K., and the U.S. revealed that 63% of consumers even find connected devices "creepy."
EASY and SIMPLER WAYS OF SECURING IOT FROM USER’s SIDE
- Change default passwords
- Make sure the software is protected
- Use encrypted protocols
Blockchain Based Secure IOT and Federated Learning
Text
2. Blockchain and Security for IOT Networks
Text
2. Blockchain and Security for IOT Networks
For IoT safety, the blockchain is able to monitor the information collected by the sensors, without allowing them to be duplicated by any wrong data. Sensors can also transfer data using Blockchain technology, without the need for a trusted third party.
Every IoT node can be registered in the blockchain and will have a blockchain id which will uniquely identify a device in the universal namespace. For a device to connect another device, one will use the blockchain id as URL and will use its local blockchain wallet to raise an identity request.
HOW BC-IoT AMALGUM WORKS
THE DISADVANTAGES OF IoT-BC AMALGUM
Getting a block sometimes takes longer. Strong cryptographic processes introduce latency. The latencies are not acceptable in a near-time and real-time service situation. Hence, blockchain is not best suited in a recording of raw data at the source.
A slight improvisation may make blockchain adapted to near-time situations. An introduction of aggregation caching nodes at the closest distance of the sources can be used as a broker between source and blockchain services. However, this will be a deviation from the key strength of blockchain and must be used after careful consideration.
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network
Observables
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network
Observables
Eg. Temperature,Humidity,Isdoor Open, etc
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network
Observables
State of the Network is Saved over the network
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network is Saved over the network
Immutable
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network is Saved over the network
Immutable
Shared Across Devices
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network is Saved over the network
Immutable
Shared Across Devices
Blockchain Based Secure IOT and Federated Learning
Text
3. Implementation of Blockchain over a simple smart-home Application using Rpi's.
Local Network: Ex. Smart Home
IOT Node
Ledger
State of the Network is Saved over the network
Immutable
Shared Across Devices
Requirement of Large
Compute power in edge devices
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Local Network: Ex. Smart Home
IOT Node
Ledger
Temperature and Humidity
Soil Moisture Sensor
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Local Network: Ex. Smart Home
IOT Node
Ledger
Temperature and Humidity
Soil Moisture Sensor
Broadcast
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Local Network: Ex. Smart Home
IOT Node
Ledger
Temperature and Humidity
Soil Moisture Sensor
All Devices Involve in Consensus Algorithm (Proof of Work Here)
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Local Network: Ex. Smart Home
IOT Node
Ledger
Temperature and Humidity
Soil Moisture Sensor
All Devices Involve in Consensus Algorithm (Proof of Work Here)
Nonce Found! Block is mined and broadcast
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Disadvantages
time series data is voluminous and will not be viable to have multiple copies across all devices
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Disadvantages
time series data is voluminous and will not be viable to have multiple copies across all devices
Data is present all over the network. Not secure over a public network.
Proof of work is computationally demanding.
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Solution
Store data locally. Send only what is required. Use Encryption based Schemes.
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Solution
Store data locally. Send only what is required. Use Encryption based Schemes.
Or
Involve multiple types of nodes. Few exclusively for Data collection few exclusively for dealing with blockchain architecture
Blockchain Based Secure IOT and Federated Learning
Text
3. Our Implementation
Involve multiple types of nodes. Few exclusively for Data collection few exclusively for dealing with blockchain architecture
When Trying to Implement... We came across....
Blockchain Based Secure IOT and Federated Learning
Text
Exciting Research!
TriaBase
Blockchain Based Secure IOT and Federated Learning
Text
Exciting Research!
TriaBase
Blockchain Based Secure IOT and Federated Learning
Text
Exciting Research!
TriaBase
Proof of
Federated Learning
Blockchain Based Secure IOT and Federated Learning
Text
Exciting Research!
TriaBase
Proof of
Federated Learning
Blockchain Based Secure IOT and Federated Learning
Text
Exciting Research!
TriaBase
Designed Exclusively for Federated Learning.
Data is stored Locally and only model weights are shared over network.
Blockchain Based Secure IOT and Federated Learning
Text
4. What is Federated Learning?
Text
5. What is Federated Learning?
Text
Federated learning is all about decentralized data collection and centralized learning
THE STEPS OF FEDERATED LEARNING
personalization of mobile using local model
Many users’ updates are aggregated.
Change in local
model using aggregated one
PROPERTIES OF TYPICAL F.L PROBLEMS
Training on real-world data from mobile devices provides a distinct advantage over training on proxy data that’s generally available in the data center.
PROPERTIES OF TYPICAL F.L PROBLEMS
This data is privacy sensitive or large in size (compared to the size of the model), so it is preferable not to log it to the data center purely for the purpose of model training (in service of the focused collection principle).
PROPERTIES OF TYPICAL F.L PROBLEMS
This data is privacy sensitive or large in size (compared to the size of the model), so it is preferable not to log it to the data center purely for the purpose of model training (in service of the focused collection principle).
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Select Clients for Training
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Select Clients for Training
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Send model for local training
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Train Locally
Typicall Flow of Federated Learning
Shared model is
trained on server-side
Federated Averaging by server
Blockchain Based Secure IOT and Federated Learning
Text
5. IOT and Federated Learning: Need for Federated Learning
Text
4. IOT and Federated Learning: Need for Federated Learning
THE FL-IoT ARCHITECHTURE
COMMUNICATION IN FL-IoT
TYPES OF FL-IoT NETWORKING STRUCTURE
CHALLANGES IN FEDERATED LEARNING - IoT
data privacy
Heterogeneity in PC
Bottleneck in communication
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Consider Plant Monitoring and water irrigation application
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Consider Plant Monitoring and water irrigation application
Moderate Wheather Conditions
Extreme Weather Conditions
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
These Are Non- I.I.D data
Moderate Wheather Conditions
Extreme Weather Conditions
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
These Are Non- I.I.D data
Moderate Wheather Conditions
Extreme Weather Conditions
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Person 1 waters the plant if temperature is above 26C and humidity is less than 44
Moderate Wheather Conditions
Extreme Weather Conditions
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Person 1 waters the plant if temperature is above 26C and humidity is less than 44
Moderate Wheather Conditions
Extreme Weather Conditions
Person 2 waters the plant, if temperature is above 25C and humidity, is less than 50
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
How Are their data distributed?
Watering
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
How Are their data distributed?
Watering
Clearly, the Data is non identically distributed.
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Watering
Clearly, the Data is non identically distributed.
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Server
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Server
/start-federation
/start-federation
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Server
/start-federation
/start-federation
Train Logistic Regression
Train Logistic Regression
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Server
/start-federation
/start-federation
Blockchain Based Secure IOT and Federated Learning
Text
6. Simple Proof-of-Concept for Federated learning using Rpi's.
Take Average
Server
/completed
/completed
Blockchain Based Secure IOT and Federated Learning
Text
7. Implementation of Federated learning of XOR using PyGAD
Outline
A Simple Implementation to
- Apply the concepts of Federated Learning In Python .
- Using Socket Programing for communication between Clients and Server.
- The ML model is created using PyGAD which trains ML models using the genetic algorithm (GA)
Client Server Connection
A Secure TCP connection is established between multiple Clients and Server
Client
Client
Server
XOR Problem
The XOr problem is that we need to build a Neural Network to produce the truth table related to the XOr logical operator. This is a binary classification problem. Supervised learning is a better way to solve it.
XOR Function
Non-Linearly Separable Data Points
Client-Server
Server.py
The server app. It creates a model that is trained on the clients' devices using FL.
Client1.py
A client app which trains the model sent by the server using just 2 samples of the XOR problem.
Client2.py
A client app which trains the model sent by the server using the other 2 samples of the XOR problem.
Output
Inference
- Federated Learning depends on the aggregated updates of many users.
- As an when a new client joins we could see that the model is learning based on the spike in the model fitness score
- A change to the shared model is made according to the aggregated updates, after which the procedure is repeated.
Palette
By aie'24 amrita
