# Democratizing Cellular Access with AnyCell

### Presentation 

#### Outline 

* Why a new cellular architecture?
* Cellbricks: overall approach 
* Prototype implementation 

#### Growing the tent in cellular 

* Today's cellular ecosystem is dominated by a small number of mobile network operators (MNOs) 
* Our goal: lower the barrier to entry for new entrants 
* Key: ensure MNOs with a small/mid-scale footprint can play
  * City government, university campus --> equal participant 
* This is difficult with today's architecture! 

#### Today's architecture relies heavily on trust and in-network coordination 

* Authentication / accounting built on trust between the user and its MNO 
* Seamless mobility built on coordination between towers (handovers)
* Broad coverage built on coordination / trust between MNOs (roaming) 

#### Pre-establishing trust doesn't scale! 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2FNENPvB4wQCYnsx84rMe8%2Fimage.png?alt=media\&token=6622fd87-3c0d-49cd-9e57-5f22f906d893)

#### Seamless mobility via in-network coordination does not scale 

#### The same problems exist with MVNOs 

* User signs up for MVNOs 
* MVNO still needs to go out and have trust agreements with all MNOs 

#### Cellbricks: a new cellular architecture

* Goal: enable users to consume access on-demand from any available infrastructure operator without requiring pre-established trust or in-network coordination 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2FbzSChDwYT5V3VbkeW5By%2Fimage.png?alt=media\&token=21ea594f-8887-4d74-8064-087b60925883)

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2FAmY7O6Br47iI7OHoQvpK%2Fimage.png?alt=media\&token=82c295f3-1a6b-49af-a0e4-841034e85f29)

* Potential benefits:
  * Lower barrier to entry for new operators
  * More efficient use of infrastructure
  * Simpler infrastructure 
* Two main challenges 
  * Secure attachment and billing (without mutual trust)
  * Seamless mobility (without in-network coordination) 

#### Secure attachment 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2FY6YzBFx17xQ3vNN3AWZG%2Fimage.png?alt=media\&token=dd945b97-1b7f-40eb-91a3-a50f06cc83f8)

* Today: share keys, cellbricks: public keys 
  * How any online transaction works (use exactly the same authentication)
  * 3-way authentication 
  * Evidence: serving a valid user that belongs to a particular broker, and that the MMO is actually the one the user wanted to connect to 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2F02NCpOlKNDXGmsktzTkk%2Fimage.png?alt=media\&token=408733cb-7f1f-47da-ba01-c1cf38c117aa)

* Broker to pay the MNO accordingly 
* Broker and MNO: reputation system 

#### Seamless mobility 

* Today, handovers require cooperation between towers 
  * Allows a UE's IP address to stay unchanged (TCP connection stays unchanged, application sessions persist)
* In CellBricks, handovers may involve different MNOs
  * Different to preserve a UE's IP address across admin boundaries 
* Solution: leverage modern transport protocols (MPTCP, QUIC)
  * L4 connection is maintained even if the UE's IP address changes
* Moving mobility support out of the network and into the transport layer 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2FyxaSxOqWqkt9xVBONPSX%2Fimage.png?alt=media\&token=21d13a54-0d89-4f8a-8cd8-99118baf0e8f)

#### Prototype 

![](https://2097630930-files.gitbook.io/~/files/v0/b/gitbook-x-prod.appspot.com/o/spaces%2F-MVORxAomcgtzVVUqmws%2Fuploads%2Fh7srJBZCq8KpOceEjmhs%2Fimage.png?alt=media\&token=3f077797-100e-4349-a6a5-471324bc9b8c)

* USRPs
  * Provide ratio connectivity (no changes to RAM)
* srsLTE
  * UE and eNodeB
* Magma 
  * AGW: extended to support secure attachment (3-way secure attachment)
  * Orc8r: added a Brokerd service 
* Performance 
  * End-to-end latency of attachment protocol
  * UE, eNodeB, AGW in local machines 
  * Vary locations for SubscriberDB (S6a) and brokered 

Host-driven mobility is a promising direction to explore further 

Conclusion 

* Technical pieces for a radically different cellular ecosystem 
  * Public key based 3-way authentication 
  * Secure counters in device hardware 
  * Promote the adoption of MPTCP, QUIC 
* Status
  * Early prototype on Facebook's Magma 
  * Real-world measurements of overheads (MPTCP, handovers)