Secure Chaotic Spread Spectrum Systems Jin Yu Jin Yu WISELAB, ECE Department WISELAB, ECE Department Stevens Institute of Technology Stevens Institute of Technology Hoboken, NJ 07030 Hoboken, NJ 07030
Outline � Introduction � Chaotic SS signals � Security/ LPI performance � Intercept receivers � Binary correlating detection � “Mismatch” problem � Particle-filtering based approach � Dual-antenna approach � Numerical results � Conclusions
Introduction � LPI/ LPD-Secure/ covert communications � Spread-spectrum systems � Direct sequences � PN binary sequences � Chaotic sequences � Frequency hopping � Time hopping (UWB) � Interceptors � likelihood-ratio test � Energy detector
Chaotic Signals � Generate chaotic spreading sequences � Discrete Chaotic Map � Exponential Map, Triangular Map… . � For Example: logistic map = α − ( 1 ) 0 ≤ x n ≤ 1, 0 ≤ α ≤ 4 x x x + 1 n n n � Bipolar signaling = − 2 1 a x n n � PDF of { a n } 1 = − ≤ ≤ ( ) , 1 1 f a a n n 2 π − 1 a n
Properties of Chaotic Sequences Chaotic sequence (logistic map: α = 4, x 0 = 0.2) pdf of a n 2.5 1 2 0.8 1.5 0.6 f(a n ) x n 1 0.4 0.5 0.2 0 0 0 20 40 60 80 100 -1.2 -0.8 -0.4 0 0.4 0.8 1.2 n a n � Non-binary and non-periodic � Random-like behaviors � Good auto- and cross-correlation � Large number of available spreading sequences for multiple-access applications
System Model � Received Signals ⎧ ω + φ + ⎪ 2 ( ) cos( ) ( ) , P a t t n t H 0 1 = ≤ ≤ ( ) 0 ⎨ r t t T ( ) , ⎪ n t H ⎩ 0 where ∞ = ∑ − − τ ( ) ( ) a t a p t nT T n c c = −∞ n The chip epoch τ T c is modeled by r.v. τ , uniformly distributed in [0, 1).
Binary Correlating Method � Likelihood ratio test (Optimum Intercept Receivers) � Synchronous coherent � Synchronous noncoherent � Asynchronous coherent � Asynchronous nonherent � Gaussian approximation ⎧ ⎫ ⎡ ⎤ − − 1 1 ⎛ ⎞ Q N ⎪ + ⎪ 2 2 ( 1 ) ∏ ∑ n T P ∫ ⎜ ⎟ ⎢ ⎥ Λ = κ c ω ( ( )) E exp ( ) ( ) cos( ) ⎨ ⎬ r t r t b t t dt ε φ ⎜ ε φ ⎟ 1 , , 0 q ⎢ ⎥ N ⎪ ⎪ ⎝ ⎠ nT ⎣ ⎦ 0 ⎩ = = ⎭ c 0 0 n q
Synchronous Coherent Case � Using Gaussian approximation, we obtain − 1 = + γ δ ( )( 0 . 5 ) − γ ( ) 2 m N N T C Q P N C 0 c c , 1 λ c k FA = ( ) P Q D 2 + γ + γ 2 2 2 1 4 2 C D σ = + γ + γ δ c c ( ) ( 0 . 5 ( 2 ) ) N N T C D 0 c c c , 1 λ k (a) Binary Synchronous Coherent Detector 2 [ ] E a = C ω + φ cos( 0 ) t [ ] E a Antenna 2 1 ( ) Var a = D 2 Matched Filter [ ] E a Low Noise -1 Amplifier 1/N 0 T EXP ∫ 0
Synchronous Noncoherent Case Antenna ( ) 2 1 ω cos( ) Matched Filter ( ) 1/2 0 t Low Noise -1 Amplifier ( ) 2 I 0 ( ) ω sin( ) 0 t COMB Ln( ) EXP P(0, T c ) FILTER The mean and variance of λ is − 1 − γ ( ) Q P N C = + γ δ ( )( 1 ) c m N N T C FA ⇒ = ( ) 0 , 1 λ c c P Q k D 2 + γ + γ 1 2 0 . 5 C D 2 2 2 c c σ = + γ + γ δ ( ) ( 1 ( 2 0 . 5 ) ) N N T C D 0 , 1 λ c c c k
Asynchronous Cases � Assume chip epoch is U[0, T c ) � Coherent case − 1 2 − − τ + τ γ ( ) 2 ( 1 2 2 ) Q P N C c λ τ = FA / ) ( ) ( P Q D + γ 1 4 C c − 1 2 1 − − τ + τ γ ( ) 2 ( 1 2 2 ) Q P N C FA c ⇒ = τ ∫ ( ) P Q d D + γ 1 4 C 0 c � Noncoherent case − 1 2 − − τ + τ γ ( ) ( 1 ) Q P N C c λ τ = FA / ) ( ) P ( Q D + γ 1 2 C c − 1 2 1 − − τ + τ γ ( ) ( 1 ) Q P N C FA c ⇒ = τ ∫ ( ) P Q d D + γ 1 2 C 0 c
Performance Comparison Chaotic vs. Binary PN (Sync) P FA =0.01 and N =1000 0 10 Engergy Coherent-binary Noncoherent-binary Coherent-chaotic Noncoherent-chaotic detection -1 10 -2 10 -30 -25 -20 -15 -10 -5 SNR(chip)
Particle-Filtering Based Detector � Uncertainties in Chaotic Signals � Amplitude uncertainty (mismatch problems) � For all detection scenarios with chaotic signals � Phase uncertainty � Noncoherent detections � Delay uncertainty � Asynchronous detections
Particle-Filtering Based Detector � Design particle sets � approximate the unknown random variables � select the most likely particle statistically � combat the impact due to uncertainties � Reduce computational complexity � Updated particles for each iteration � Fixed particles for each iteration
Particle-Filtering Based Detector � LRT function with particle filtering ( | ) p r H 1 Λ = ( ( )) I r t Coherent detection ( | ) p H r 0 I ∏ ⎛ ∑ ⎞ N L ⎜ a ( | ( )) ⎟ p r a j , ⎝ I i i ⎠ = = 1 1 i j Λ = ( ( )) r t ∏ N N ( ) L p r , a I n = 1 n ( , | ) p r r H 1 I Q Λ = ( ( )) r t Noncoherent ( , | ) p r r H 0 I Q detection ∏ ∑ ∑ ⎛ ⎞ N L L φ ⎜ a p ( , | ( ), ( )) ⎟ p r r a j p , , ⎝ I i Q i i i ⎠ = = = 1 1 1 i j p Λ = ( ( )) r t ∏ N N N ( , ) L L p r r , , a p I n Q n = 1 n Notice: probability density functions p( • ) is used to select the particles a i ( j ) and φ i ( p ) which are mostly close to the actual amplitude and phase.
ω + φ cos( 0 ) t Antenna Particle-Filtering Based Detector Low Noise Amplifier INITIALIZATION: Particles a j , j = 0, P 0 = 1, P 1 = 1 Synchronous coherent receivers with P FA = 0 . 01 , L a = 50, and N = 1000 + ( 1 ) ∫ j T C jT C 0 10 CALCULATE: Binary seq.: Binary Detection p( r j | a j , H 1 ), p( r j |H 0 ) Logistic seq.: Binary Detection [6] Triangular seq.: Binary Detection [6] Logistic seq.: Particle Filter Triangular seq.: Particle Filter P 1 = P 1 p( r j |H 1 ), Detection probability P 0 = P 0 p( r j |H 0 ) -1 RESAMPLING ( a j + 1 ) 10 j = j +1 YES j < N ? NO -2 10 -30 -25 -20 -15 -10 -5 DECISION P 1 /P 0 SNR γ c (chip)
Asynchronous Detection: Multiple sampling � Obtain multiple observations by multiple sampling at τ n (combat delay uncertainty) ⎛ ⎞ ⎛ ⎞ 1 1 1 1 1 1 L L r r r r r r ⎜ ⎟ ⎜ ⎟ , 1 , 2 , , 1 , 2 , I I I N Q Q Q N ⎜ ⎟ = ⎜ ⎟ = M M O M M M O M R R I Q ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ ⎜ ⎟ N N N N N N L L r r r r r r d d d d d d ⎝ ⎠ ⎝ ⎠ , 1 , 1 , , 1 , 1 , I N I I Q Q Q N ⎛ ⎞ 1 1 1 ⎛ ⎞ 1 1 1 L n n n L ⎜ ⎟ n n n , 1 , 2 , ⎜ ⎟ Q Q Q N , 1 , 2 , I I I N ⎜ ⎟ = ⎜ ⎟ M M O M = N M M O M N Q ⎜ ⎟ I ⎜ ⎟ ⎜ ⎟ N N N ⎜ ⎟ L n d n d n d N N N ⎝ ⎠ L n n n , 1 , 1 , d d d Q Q Q N ⎝ ⎠ , 1 , 1 , I N I I
Parallel Detection Algorithm � LRT function ⎧ ⎫ ⎪ n ⎪ ( | ) p r H Λ = 1 = ( ( )) max I , 1 , 2 , , ⎨ ⎬ K r t n N d ⎪ ⎪ n ( | ) ⎩ p H ⎭ n r 0 I ⎧ ⎫ n n ( , | ) ⎪ ⎪ p r r H 1 I Q Λ = = ( ( )) max , 1 , 2 , , ⎨ ⎬ K r t n N d n n ⎪ ( , | ) ⎪ p r r H n ⎩ ⎭ 0 I Q Notice: The row having the minimum delay is automatically selected by probability density functions p( • ) to detect the presence of radio signals.
Numerical Results Asynchronous detectors with various N d , P FA = 0 . 01 , N = 1000 , and parallel detection algorithm. 0 10 ASYN-CO: PF, La = 20, Nd = 2 ASYN-CO: PF, La = 10, Nd = 2 ASYN-CO: PF, La = 20, Nd = 1 ASYN-CO: PF, La = 10, Nd = 1 ASYN-NC: PF, La = 10, Lp = 10, Nd = 2 ASYN-NC: PF, La = 10, Lp = 10, Nd = 1 Detection probability ASYN-NC: Chao. Seq - Bin. Detection -1 10 -2 10 -30 -25 -20 -15 -10 -5 SNR γ c (chip)
Dual-Antenna Approach: Synchronous coherent case Incident Wave ω + φ cos( ) t 0 ∆ = θ θ Antenna cos / d c r 1 ( t ) ψ = π θ λ 2 cos / T d ∫ 0 0 Low Noise Amplifier Decision d ω + φ + ψ cos( 0 ) t r 2 ( t ) T ∫ 0 Signal Model Detection Probability ⎛ ⎞ ⎧ − 1 = ω + φ + − γ ⎪ ( ) 2 ( ) cos( ) ( ) ( ) 2 r t P a t t n t ⎜ Q P N ⎟ 1 0 1 = FA c ⎨ P Q ⎜ ⎟ D ⎪ = ω + φ + ψ + + γ ( ) 2 ( ) cos( ) ( ) 1 4 ⎩ r t P a t t n t ⎝ ⎠ 2 0 2 c
Dual-Antenna Approach: Synchronous noncoherent case Incident Wave θ Antenna r 1 ( t ) T ∫ 0 Low Noise T ∫ Amplifier 0 ω cos( ) 0 t Decision d ω sin( ) 0 t r 2 ( t ) T ∫ 0 T ∫ 0 ⎛ ⎞ − 1 − Ω θ γ ( ) 2 ( ) ⎜ Q P N ⎟ = FA c P Q θ | ⎜ ⎟ D + γ 1 2 ⎝ ⎠ c
Dual-Antenna Approach: Asynchronous case Incident Wave Antenna θ r 1 ( t ) Low Noise Amplifier Decision T d ∫ 0 r 2 ( t ) ⎛ ⎞ − 1 − Ω θ γ ( ) ( ) Q P N ⎜ ⎟ = FA c P Q θ | ⎜ ⎟ D + γ 1 2 ⎝ ⎠ c Ω θ = π θ λ ( ) cos( 2 cos / ) d 0
Recommend
More recommend