Transaction clustering using network traffic analysis for Bitcoin and derived Transaction clustering using network traffic blockchains analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction Network privacy Alex Biryukov, Sergei Tikhomirov Our transaction clustering method Parallel connections SnT, University of Luxembourg Weighting timestamp vectors Clustering the correlation matrix 29 April 2019 Metrics Cryblock Experimental results Paris, France Discussion Future work 1/30
Transaction Outline clustering using network traffic analysis for Bitcoin and derived blockchains Introduction Biryukov, Tikhomirov Network-level privacy of Bitcoin and derivatives Introduction Network privacy Our transaction clustering method Our transaction Parallel connections clustering method Parallel connections Weighting timestamp vectors Weighting timestamp vectors Clustering the correlation matrix Clustering the correlation matrix Metrics Metrics Experimental results Experimental results Discussion Future work Discussion Future work 2/30
Transaction Outline clustering using network traffic analysis for Bitcoin and derived blockchains Introduction Biryukov, Tikhomirov Network-level privacy of Bitcoin and derivatives Introduction Network privacy Our transaction clustering method Our transaction Parallel connections clustering method Parallel connections Weighting timestamp vectors Weighting timestamp vectors Clustering the correlation matrix Clustering the correlation matrix Metrics Metrics Experimental results Experimental results Discussion Future work Discussion Future work 3/30
Transaction Privacy in cryptocurrencies clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ Transactions not linked to ”real-world” identity Network privacy Our transaction ◮ False sense of privacy: blockchain can be analyzed clustering method Parallel connections Weighting timestamp vectors ◮ Taint analysis, various heuristics Clustering the correlation matrix Metrics ◮ Countermeasures: mixing, cryptography (Monero, Experimental results Zcash, ...) Discussion Future work 4/30
Transaction Our focus: network-level privacy clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction Network privacy ◮ How do messages propagate through the network? Our transaction clustering method Parallel connections ◮ What information does the traffic leak? Weighting timestamp vectors Clustering the correlation matrix ◮ Is it possible to link txs by the same user? Metrics Experimental results Discussion Future work 5/30
Transaction Outline clustering using network traffic analysis for Bitcoin and derived blockchains Introduction Biryukov, Tikhomirov Network-level privacy of Bitcoin and derivatives Introduction Network privacy Our transaction clustering method Our transaction Parallel connections clustering method Parallel connections Weighting timestamp vectors Weighting timestamp vectors Clustering the correlation matrix Clustering the correlation matrix Metrics Metrics Experimental results Experimental results Discussion Future work Discussion Future work 6/30
Transaction Transaction propagation in Bitcoin clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ Alice: INV (I know an object with hash H) Network privacy Our transaction clustering method ◮ Bob: GETDATA (I want to get this object) Parallel connections Weighting timestamp vectors Clustering the correlation matrix ◮ Alice: TX (Here it is) Metrics Experimental Bob announces to his neighbors, etc. results Discussion Future work 7/30
Transaction Broadcast randomization clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction Network privacy Privacy issue: well-connected adversary infers the original IP. Our transaction Countermeasures: clustering method Parallel connections ◮ trickling: send to a subset once a period Weighting timestamp vectors Clustering the correlation matrix ◮ diffusion: send to all after random delays Metrics Experimental results Discussion Future work 8/30
Transaction Previous work clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ Biryukov, Khovratovich, Pustogarov (2014) - Network privacy Our transaction ”Deanonymisation of clients in Bitcoin P2P network” clustering method proposed a method for linking Bitcoin txs to IPs Parallel connections Weighting timestamp vectors Clustering the correlation matrix ◮ Key idea: nodes connect to 8 random ”entry nodes”, Metrics the ”entry set” is a fingerprint Experimental results Discussion Future work 9/30
Transaction Outline clustering using network traffic analysis for Bitcoin and derived blockchains Introduction Biryukov, Tikhomirov Network-level privacy of Bitcoin and derivatives Introduction Network privacy Our transaction clustering method Our transaction Parallel connections clustering method Parallel connections Weighting timestamp vectors Weighting timestamp vectors Clustering the correlation matrix Clustering the correlation matrix Metrics Metrics Experimental results Experimental results Discussion Future work Discussion Future work 10/30
Transaction Understanding relationships between transactions clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ Connect to many nodes Network privacy Our transaction clustering method ◮ Log timestamps of received tx announcements Parallel connections Weighting timestamp vectors Clustering the correlation matrix ◮ Intuition: we will hear of new txs from Alice or her Metrics entry nodes faster than from other nodes Experimental results Discussion Future work 11/30
Transaction Parallel connections clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov ◮ Nodes maintain 8 outgoing and 117 (optional) incoming Introduction connections Network privacy Our transaction clustering method ◮ Txs propagate to some neighbors with random delays Parallel connections Weighting timestamp vectors Clustering the correlation ◮ If we connect to a node once, the probability of getting matrix Metrics a new tx quickly is low Experimental results Discussion ◮ Can we connect to nodes many times in parallel? Future work 12/30
Transaction Saturating connection slots clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov ◮ bcclient tool connects to Bitcoin nodes with many Introduction parallel connections Network privacy Our transaction clustering method ◮ We occupy all available slots (avg 64 slots / peer on Parallel connections Weighting timestamp Bitcoin testnet) vectors Clustering the correlation matrix Metrics ◮ Nodes don’t distinguish incoming and outgoing Experimental results connections for tx propagation! Occupy 50% of slots – Discussion 50% chance of getting a new txs first. Future work 13/30
Transaction Weighting timing vectors clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ Earlier work only considered the first IP to relay a tx Network privacy Our transaction clustering method ◮ We consider the vector of the first 3 – 7 IPs to relay a Parallel connections tx, and assign them exponentially decreasing weights Weighting timestamp vectors Clustering the correlation matrix Metrics ◮ High correlation between vectors indicate the same Experimental originator results Discussion Future work 14/30
Transaction Weighting formula clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction IPs p i get decreasing weights; median IP gets weight 0.5: Network privacy Our transaction w ( p i ) = e − ( t i / k ) 2 clustering method Parallel connections Weighting timestamp vectors where Clustering the correlation t median matrix k = Metrics � − ln(0 . 5) Experimental results Discussion Future work 15/30
Transaction Weighting timing vectors: example clustering using network traffic analysis for Bitcoin High values indicate higher probability of an IP to be the and derived blockchains originator or one of its entry nodes. Biryukov, Tikhomirov Introduction Network privacy Our transaction clustering method Parallel connections Weighting timestamp vectors Clustering the correlation matrix Metrics Experimental results Discussion Future work Figure: Weight function for 3 vectors of timestamps 16/30
Transaction Clustering of vectors clustering using network traffic analysis for Bitcoin and derived blockchains Biryukov, Tikhomirov Introduction ◮ For each pair of txs, calculate correlation of weight Network privacy vectors Our transaction clustering method ◮ Hypothesis: correlation matrix has a block-diagonal Parallel connections Weighting timestamp vectors structure Clustering the correlation matrix Metrics ◮ Related transactions form clusters along the main Experimental results diagonal Discussion Future work 17/30
Recommend
More recommend
