cs527 mancill presentation

Joint Compression and Encryption for Wireless Sensor Networks: 

Joint Compression and Encryption for Wireless Sensor Networks Computer Security CS 427/527 Tony Mancill May 2nd, 2007

What’s a WSN?: 

What’s a WSN? Self-organizing network of wireless motes deployed in-situ to monitor the environment. Wide variety of sensing applications. Low-bandwidth “event detection” Target tracking High-bandwidth waveform capture Runs small embedded systems OS – e.g. TinyOS (3.4kB ROM, 700 bytes RAM).

Resource Constraints: 

Resource Constraints CPU 8MHz 8/16-bit in second generation iMote2 with Intel Xscale (13-416MHz) Memory RAM: 512-4kB RAM (32MB on iMote2) ROM: 128kB for program (32MB on iMote2) Radio 250kbps max; 30m – 100m range @ 1mW CC2420 – IEEE 802.15.4 + AES Power supply – up to a year from 2000mAh

Encryption for WSNs: 

Encryption for WSNs DES, RSA, AES, and ECC Sizzle – HTTPS in a matchbox Execution Times on Mote-class Hardware

Data Compression Overview: 

Data Compression Overview Entropy coders Huffman-coding/Adaptive Huffman coding Dictionary coders LZ77, LZ78, LZW (1984) Arithmetic coding Burrows-Wheeler Transform and Structured Transpose

LZW – Lempel-Ziv-Welch: 

LZW – Lempel-Ziv-Welch Replace recurring strings from input with symbols from dictionary. Dictionary is built on the fly. Order of LZW dictionary determines number of bits per symbol in output. Dictionary reset and other control words

TEA, XTEA, XBTEA: 

TEA, XTEA, XBTEA Very small Feistel block cipher 64-bit blocks, 128-bit key Used in Xbox (vulnerability due to equivalent keys) Configurable number of rounds Well-studied

TEA Implementation: 

TEA Implementation void encrypt(uint32_t* v, uint32_t* k) { uint32_t v0=v[0], v1=v[1], sum=0, i; uint32_t delta=0x9e3779b9; uint32_t k0=k[0], k1=k[1], k2=k[2], k3=k[3]; for (i=0; i < ROUNDS; i++) { sum += delta; v0 += ((v1<<4) + k0) ^ (v1+sum) ^ ((v1>>5) + k1); v1 += ((v0<<4) + k2) ^ (v0+sum) ^ ((v0>>5) + k3); } v[0]=v0; v[1]=v1; }

Data Compression for WSNs: 

Data Compression for WSNs S-LZW LZW compression with a “mini-cache” Compressed bit saves between 2000 and 4 million CPU cycles worth of power, depending on radio. EasiPC Eliminating redundant headers (same approach taken for HTTPS headers in Sizzle)

Joint Compression and Encryption: 

Joint Compression and Encryption OpenSSH, PGP, SSL MPEG encoders aespipe experiment:

LZW-Shuffle: 

LZW-Shuffle LZW dictionary coding Seed PRNG on sender and receiver To encode, rotate LZW symbols using PRNG before transmitting, modulo LZW-order To decode, reverse rotation Use rotation schedule to reduce overhead of PRNG

Evaluation: 

Evaluation PRNG testing (TEA) Statistical analysis STS from NIST Execution time on Tmote Sky

Evaluation Cont’d: 

Evaluation Cont’d LZW-Shuffle implementation in C Measure compression using fixed-order LZW implementation versus variable.

Security of LZW-Shuffle: 

Security of LZW-Shuffle Statistical analysis PRNG weaknesses Short period ECB attack Known-plaintext attack Others?

Other Security Concerns: 

Other Security Concerns Routing attacks HELLO floods Blackhole Sybil DoS Physical security Limited options ROM decapsulation/microprobing

Conclusions: 

Conclusions CSPRNG vs. security Heavy CSPRNG == Compression + Encryption Light CSPRNG => Security compromise Wide range of research opportunities. Lossless waveform compression based on DCTs “Good enough” (pragmatic) security Hardware-based CSPRNGs No students were harmed during the making of this presentation!